Advanced neuroimaging of the trigeminal nerve and the whole brain in trigeminal neuralgia: a systematic review

Abstract

For trigeminal neuralgia (TN), a major role of imaging is to identify the causes, but recent studies demonstrated structural and microstructural changes in the affected nerve. Moreover, an increasing number of studies have reported central nervous system involvement in TN. In this systematic review, recent quantitative magnetic resonance imaging (MRI) studies of the trigeminal nerve and the brain in patients with TN were compiled, organized, and discussed, particularly emphasizing the possible background mechanisms and the interpretation of the results. A systematic search of quantitative MRI studies of the trigeminal nerve and the brain in patients with TN was conducted using PubMed. We included the studies of the primary TN published during 2013 to 2023, conducted for the assessment of the structural and microstructural analysis of the trigeminal nerve, and the structural, diffusion, and functional MRI analysis of the brain. Quantitative MRI studies of the affected trigeminal nerves and the trigeminal pathway demonstrated structural/microstructural alterations and treatment-related changes, which differentiated responders from nonresponders. Quantitative analysis of the brain revealed changes in the brain areas associated with pain processing/modulation and emotional networks. Studies of the affected nerve demonstrated evidence of demyelination and axonal damage, compatible with pathological findings, and have shown its potential value as a tool to assess treatment outcomes. Quantitative MRI has also revealed the possibility of dynamic microstructural, structural, and functional neuronal plasticity of the brain. Further studies are needed to understand these complex mechanisms of neuronal plasticity and to achieve a consensus on the clinical use of quantitative MRI in TN.

1. Introduction

Trigeminal neuralgia (TN) is a rare disorder characterized by recurrent, sudden, electric shock–like facial pain attacks in 1 or more divisions of the trigeminal nerve, usually triggered by innocuous stimuli.^18,45 The prevalence of TN ranges between 0.1% and 0.7%, and it is more common in women and in people aged 50 years and older.² It has been considered that the most common cause of TN is neurovascular compression (NVC), typically at the root entry/exit zone (REZ) or transition zone, which is a landmark for classic TN (75%–80%).¹⁸ Secondary TN, that is TN caused by an underlying disease such as tumors and multiple sclerosis, accounts for around 15%, and idiopathic TN, TN of unknown etiology, accounts for 5% to 10%.¹⁸ TN cases with neurovascular contract, but without evidence of morphological changes, such as atrophy or displacement, in the trigeminal nerve root are also classified as idiopathic.⁴⁵

The main purpose of neuroimaging in TN has been to identify neurovascular contact and secondary causes. Diffusion tensor imaging (DTI) analysis was introduced to TN to assess the microstructural characteristics of the affected nerve. Although the efficacy of earlier decompression operations confirms the role of NVC in TN, the fact that neurovascular contract is a common finding in asymptomatic cases and some of the clinical characteristics of TN, such as the effectiveness of the anticonvulsant (voltage-gated sodium channel [VGSC] blocker) mechanisms similar to cerebral paroxysmal discharges in epilepsy, pain attacks far larger in magnitude than the stimulus, self-sustained pain attack, and post-attack refractory period, may indicate the causes beyond those peripheral theories and support central nervous system (CNS) involvement, including hyperactivity of the trigeminal nucleus.^27,86 Diffusion tensor imaging has also been applied to the analysis of the cerebral white matter microstructure in patients with TN. Moreover, advanced imaging analysis techniques, such as voxel-based/surface-based morphometry (VBM/SBM), structural connectivity analysis, and functional magnetic resonance imaging (fMRI), have been applied to visualize and quantify multiple changes in brain structure and function because of aging and disease pathophysiology.^39,40,92

The quantitative MRI studies of TN generally provided results supporting microstructural, structural, and functional alterations of the brain; however, the results were sometimes inconsistent or contradictory, and hard to interpret. In this review, recent advanced neuroimaging studies of primary TN will be compiled and organized, particularly emphasizing the implication of the results and the association with possible pathophysiology.

2. Methods

This systematic review was planned, conducted, and reported in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses statement.⁹³ Systematic keyword-based searches were conducted using PubMed on January 1, 2024. The inclusion criteria were as follows: (1) studies of the primary TN published (including “Epub ahead of print”) during 2013 to 2023 in English; (2) describing imaging analysis performed on humans; (3) conducted for the volume, cross-sectional area, or DTI analysis of the trigeminal nerve in classic TN; (4) for the assessment of the structural and microstructural alterations of the brain; (5) for the analysis of structural connectivity of the brain; or (6) for the analysis of fMRI, including focal brain activity and functional connectivity studies of the brain, in patients with primary TN. Review, meta-analysis, and quantitative MRI studies including TN type 2 or secondary TN were excluded. The selected studies were reviewed for further evaluation independently by a reviewer (M.W.). We identified and extracted the following data from each publication: study authors, study design, study setting, eligibility criteria, participant characteristics (including age average/range, gender, TN types, and symptom duration), sample size, intervention (if applicable), outcome measures, follow-up time point(s), statistical analyses, and key findings.

3. Results

3.1. Reference search

A study flow diagram starting from database search to extraction is shown in Figure 1. Search terms ((trigeminal in the title) AND (MRI)) identified 919 records and ((trigeminal in the title) AND (brain)) identified 1182 records. After the removal of duplication (351 studies) and studies that did not meet the inclusion criteria (1612 studies), each study was reviewed for further evaluation. Review/meta-analysis (8), quantitative MRI studies including TN type 2 or secondary TN (8 studies), a quantitative MRI study of the trigeminal nerves that enrolled both patients with pre-percutaneous and post-percutaneous balloon compression for 1 analysis (1), and a study using the same dataset with another study (1) were excluded (Supplementary Table 1, available at http://links.lww.com/PAIN/C111). Of the studies reviewed, 46 were relevant for the quantitative studies of the trigeminal nerve, 13 for microstructural analysis of the brain, 28 for the structural analysis of the brain, 4 for the structural connectivity studies, and 28 for fMRI studies.

Figure 1.

Systematic search flowchart. MRI, magnetic resonance imaging; TN, trigeminal neuralgia.

3.2. Analysis of the trigeminal nerve and tract

3.2.1. Volumetry of the trigeminal nerve

The quantitative MRI studies of the affected trigeminal nerve in patients with TN consistently showed significantly smaller nerve volume or cross-sectional area compared with the nonsymptomatic side or control subjects (Table 1).^{11,14,31,35,41,48,56,58,60,62, 68,118,121} Studies have shown that smaller nerve volume was associated with better outcomes after surgery, and it correlated with the degree of compression observed during surgery.^31,58,60

Table 1.

Study characteristics and results of the structural analysis of the affected trigeminal nerve in patients with trigeminal neuralgia.

Study	Patients with TN	TN type	Age (y), average ± SD (range)	Disease duration (y), average ± SD (range)	Surgical procedure	Study design/timing of MRI acquisition	MRI scanner	Analysis	Location of measurements	HC	CSA/volume, pre-Tx CNV vs unaffected side	CSA/volume, pre-Tx CNV vs control subjects
Duan (2015)31	26 (17 Res, 9 NR)	Classic	65.3 (41–88)	1.8 (0.5–14	MVD	Cross-sectional pre-Tx	3T	CSA	Proximal and distal cisternal segments	—	↓ (proximal) P < 0.05
Li (2023) 68	58 (48 Res, 10 NR)	Primary	Res: 68.2 ± 12.1 (32–87) NR: 65.6 ± 9.5 (52–82)	Res: 4.90 ± 5.62 (0.25–30) NR: 6.70 ± 5.08 (2–15)	PBC	Cross-sectional pre-Tx	3T	CSA	Proximal and distal cisternal segments	—	↓ (REZ) P < 0.01
Hu (2019) 48	67	Classic	(22–90)	(0.17–21)	GKRS	Cross-sectional pre-Tx	1.5T	CSA	CSA: REZ 5 mm	—	↓ P < 0.001
Gunesli (2020)41	106	86 with NVC, 20 without NVC	57.6 ± 14.7 (20–89)	—	—	Cross-sectional pre-Tx	3T	Horizontal CSA	Cisternal segment	98	↓ P < 0.001
Leal (2014)58	50	Classic	56.42 (22–79)	4.50 (0.5–14)	MVD	Cross-sectional pre-Tx	3T	CSA, CNV volume	CSA: REZ, volume: cisternal segment	20	↓ P > 0.05	↓ P > 0.05
Wang Y (2016)118	42	Idiopathic	56.5 (40–68)	6.58 ± 4.43	MVD	Cross-sectional pre-Tx	3T	CSA, CNV volume	CSA: REZ, volume: cisternal segment	42	↓ P < 0.05	↓ P < 0.05
Wang Y (2019)121	40	22 major NVC, 14 minor NVC, 4 idiopathic	55.76 ± 8.23	7.08 ± 5.29	—	Cross-sectional	3T	CSA, CNV volume	Cisternal segment	40	↓ P < 0.05	↓ P < 0.05
Danyluk (2020)21	31	Primary	Res: 53 NR: 52	Res: 6.6 NR: 5.6	MVD	Cross-sectional pre-Tx	3T	CNV volume, VBM	Cisternal segment	—	↓ P = 0.46
Ge (2023) 35	79	Classic	54.05 ± 10.56	5.03 ± 4.94	—	Cross-sectional	3T	CNV volume	Cisternal segment	81	↓ P = 0.002	↓ P = 0.002
Cheng (2017)14	60	Primary	48.3 ± 9.1 (23–70)	4.5 (0.5–25)	MVD	Cross-sectional pre-Tx	3T	CNV volume	Cisternal segment	30	↓ P = 0.001	↓ P = 0.003
Kundakçı (2011)56	25	Primary	54 ± 14.06 (25–87)	3.7 ± 3.6 (0.17–15)	—	Cross-sectional	1.5T	CNV volume	Cisternal segment	25	NS P = 0.252	↓ P = 0.04
Chen ST (2016)11	43	Primary	58.8 ± 11.0	—	36 RFA (28 Res, 8 NR)	Cross-sectional pre-Tx	3T	CNV volume, DTI	Cisternal segment	—	↓ P = 0.014
Lee YJ (2019)62	14	Classic	49 (31–64)	3 (0.5–27)	—	Cross-sectional	7T	CNV volume, CSA, DTI	Cisternal, pontine segment	15	↓ P = 0.008
Leal (2019)59	10	Classic	26–68	5.7 (3.5–13)	MVD	Longitudinal pre-Tx, post-Tx 4 y	3T	CNV volume, DTI	REZ	6	↓ P < 0.05	↓ P < 0.05
Lee C (2023)60	50 (41 Res, 9 NR)	Classic	Res: 61.1 ± 10.1 NR: 55.7 ± 9.7	Res: 5.95 ± 4.64 NR: 5.15 ± 3.38	MVD	Longitudinal pre-Tx, immediate post-Tx	3T	CSA	TZ	—	Res: ↓ P = = ↓ NR: NS P = 0.95

CKRS, CyberKnife radiosurgery; CNV, trigeminal nerve; CSA, cross-sectional area; GKRS, Gamma Knife radiosurgery; HC, healthy control subjects; L, left-sided trigeminal neuralgia; MVD, microvascular decompression; NR, nonresponders; NS, not significant; NVC, neurovascular compression; PBC, percutaneous balloon compression; R, right-sided trigeminal neuralgia; Res, responders; REZ, root entry/exit zone; RFA, radiofrequency ablation; TN, trigeminal neuralgia; TZ, transition zone; Tx, treatment.

3.2.2. Diffusion tensor imaging studies of the trigeminal nerve

The key study characteristics of DTI analysis of the affected trigeminal nerve were summarized in Table 2.^{8,11,12,25,26,42,51,55,59,62,70,76,80,82,88,96,100,119,128,138} The diffusion studies of the affected nerve consistently showed a decrease in fractional anisotropy (FA) and an increase in diffusivity compared with the contralateral side or with control subjects (Table 3). In the Gaussian process classification using 36 patients with TN (28 females, average age 42.7 ± 12.0 years), the DTI features of the trigeminal nerves distinguished patients with TN from control subjects with an accuracy near 80% (affected side: 76%–86%, unaffected side: 79%–81%), and the affected nerves from the unaffected nerves with an accuracy over 70% (75% by using FA, 70% by generalized FA).⁹

Table 2.

Study characteristics of the microstructural analysis of the affected trigeminal nerve in patients with trigeminal neuralgia.

Study	Patients with TN	TN type	Age (y), average ± SD (range)	Disease duration (y), average ± SD (range)	Surgical procedure	Study design/timing of MRI acquisition	MRI scanner	Analysis	Location of measurement	HC
Chen ST (2016)11	43	Primary	58.8 ± 11.0	—	36 RFA (28 Res,8 NR)	Cross-sectional pre-Tx	3T	DTI, CNV volume	Cisternal segment	—
Liu (2013)76	16	12 classic, 4 idiopathic	50.0 ± 7.8 (35–66)	—	—	Cross-sectional	3T	DTI	REZ	6
DeSouza (2014)26	18 R	Classic	54.1 ± 17.0	—	—	Cross-sectional	3T	DTI	REZ	18
Lutz (2016)82	81	Classic	60 ± 5	13 (4.2–21.3)	MVD	Cross-sectional pre-Tx	3T	DTI	TZ	—
Pang (2019)96	25	Primary	52.75 ± 10.1 (30–75)	1.67	MVD	Cross-sectional pre-Tx	3T	DTI	Cisternal segment	25
Kumaran (2022)55	30	29 classic, 1 idiopathic	52.9 ± 13.3	—	—	Cross-sectional	1.5T	DTI	Cisternal segment	30
Qi (2023)100	40 (23 R, 17 L)	Classic 55.4 ± 13.8 (27–79)	(0.25–20)	—	Cross-sectional	3T	DTI	Cisternal segment	40	Qi (2023)100
Luo (2022)80	60	Classic	57.45 ± 8.78 (25–80)	4.22 ± 4.04	MVD or no surgery	Cross-sectional pre-Tx	3T	DSI	Cisternal segment	35
Guo (2023)42	108	Classic	59.31 ± 12.77 (27–83)	—	—	Cross-sectional	3T	DTI	Cisternal segment	—
Lee YJ (2019)62	14	Classic	49 (31–64)	3 (0.5–27)	—	Cross-sectional	7T	DTI, CNV volume	Cisternal, pontine segment	15
Moon (2018)88	14	Classic	49 (31–64)	—	—	Cross-sectional	7T	DTI	Cisternal, pontine segment	15
Lin (2016)70	50	Classic	57.98 ± 5.46 (52–82)	—	—	Cross-sectional	3T	DTI	REZ	50 control subjects w/NVC, 50 HC
Hung (2017)51	31	Classic	Res: 60.1 ± 13.7 (36–79) NR: 55.6 ± 14.9 (25–78)	—	MVD or GKRS (17 Res, 14 NR)	Cross-sectional pre-Tx	3T	DTI	Cisternal, REZ, pontine segment	16
Wang Y (2023)119	17	Primary	59.3	(1–20)	9 MVD, 2 RFT	Cross-sectional pre-Tx 7 patients: longitudinal pre-Tx, post-Tx 6 mo	3T	DTI	REZ	34
DeSouza (2015)25	14 R	Classic	—	—	MVD (9 Res, 1 NR) or GKRS (6 Res, 9 NR)	Longitudinal pre-Tx, post-Tx 5.8 ± 4 (2–6) mo	3T	DTI	REZ	14
Zhang Y (2018)138	24 R	Classic	49.2 ± 12.4	5.5 ± 3.8 (2–20)	MVD	Cross-sectional pre-Tx, 6 patients: longitudinal pre-Tx, post-Tx 1 wk and 4–6 mo	3T	DTI	REZ	28
Leal (2019)59	10	Classic	26–68	5.7 (3.5–13)	MVD	Longitudinal pre-Tx, post-Tx 4 y	3T	DTI, CNV volume, CSA	REZ	6
Chai (2019)8	34	Classic	61.47 ± 8.42 (52–81)	(0.25–10)	MVD	Longitudinal pre-Tx, post-Tx 1 wk, 6–8 mo	3T	DTI	Cisternal segment	34
Chen TY (2021)12	12	9 classic, 3 idiopathic	68.9	—	RFA and/or CKRS	Longitudinal pre-Tx, post-Tx 1, 6, 12 mo	1.5T	DTI, T2-VBM	Cisternal segment, REZ, pontine nuclear zone, pontocerebellar tracts	4
Wu (2020)128	21	Idiopathic	56.63 ± 7.98	5.3 ± 0.7 (1.5–11.5)	IN	Longitudinal pre-Tx, post-Tx 1 y	3T	DTI, DKI	Cisternal segment	20
Lee CC (2018)61	22	Classic	51–84	1.5–17 <5 y: 7 ≥5 y: 14	GKRS	Longitudinal pre, post-Tx 6 mo	3T	DTI	REZ, cisternal segment	—

CKRS, CyberKnife radiosurgery; CNV, trigeminal nerve; DSI, diffusion spectrum imaging; DKI, diffusion kurtosis imaging; DTI, diffusion tensor imaging; GKRS, Gamma Knife radiosurgery; HC, healthy control subjects; IN, internal neurolysis; L, left-sided trigeminal neuralgia; MVD, microvascular decompression; NR, nonresponders; NVC, neurovascular compression; PBC, percutaneous balloon compression; R, right-sided trigeminal neuralgia; REZ, root entry/exit zone; RFA, radiofrequency ablation; RFT, radiofrequency thermocoagulation; TN, trigeminal neuralgia; TZ, transition zone; Tx, treatment; VBM, voxel-based morphometry.

Table 3.

Results of the microstructural analysis of the affected trigeminal nerve in patients with trigeminal neuralgia.

Study	Pretreatment trigeminal nerve vs unaffected side							Pretreatment trigeminal nerve vs control subjects
	FA	RD/λ⊥	AD/l‖I	MD/ADC	RK	AK	MK	FA	RD	AD	MD/ADC	RK	AK	MK
Chen ST (2016)11	↓ P < 0.001	↑ P < < 01	NS P = 0.536	↑ P = = .5
Liu (2013)76	↓ P < 0.005	↑ P < 0.005	NS P = 0.72	↑ trend P = 0.05				↓ P < 0.005	↑ P < 0.005	NS P > 0.005	↑ P < 0.005
DeSouza (2014)26	↓ P < 0.05	↑ trend P = 0.55	NS P > 0.05	NS P > 0.05				↓ P < 0.05	↑ P < 0.05	↑ P < 0.05	↑ P < 0.05
Lutz (2016)82	↓ P = 0.004		NS P = 0.092
Pang (2019)96	↓ P = 0.006		↑ P = 0.015					↓ P < 0.001			↑ P = 0.014
Kumaran (2022)55	↓ P < 0.001
Qi (2023)100	↓ P < 0.001	NS P > 0.05	NS P > 0.05	↑ P < 0.05	↓ P < 0.001	↓ P < 0.001	↓ P < 0.001	↓ P < 0.001	↑ P < 0.05	↑ P < 0.05	↑ P < 0.05	↓ P < 0.001	NS P > 0.05	↓ P < 0.001
Luo (2022)80	↓ P = 0.001							↓ P = 0.001		↑ P = 0.036
Guo (2023)42	↓ P < < ER	—	—	NS P > 0.05
Lee YJ (2019)62	↓ P = 0.04
Moon (2018)88	↓ P = 0.042	↑ P = 0.03	NS P = 0.37	↑ P = 0.03
Lin (2016)70	—	↑ P = 0.00	NS P = 0.16	—
Hung (2017)51	↓ (REZ) P < 0.05	↑ (REZ) P < 0.05	↑ (REZ) P < 0.05	↑ (REZ) P < 0.05				(REZ) Res: NS P = 0.93 NR:↓ P < 0.05	↑ (REZ) P < 0.05	↑ (REZ) P < 0.05	↑ (REZ) P < 0.05
Wang Y (2023)119	↓ P = 0.044	↑ P = 0.003	↑ P < 0.001	↑ P < 0.001				↓ P = 0.001	↑ P < 0.001	↑ P = 0.001	↑ P < 0.001
DeSouza (2015)25								↓ P = 0.004	↑ P = 0.001	↑ P = 0.01	↑ P = 0.002
Zhang Y (2018)138 ↓ P < 0.001	↑ P = 0.003	↑ P < 0.001	↑ P < 0.001	Zhang Y (2018)138 ↓ P < 0.001				↓ P < 0.001	↑ P = 0.002	↑ P < 0.001	↑ P < 0.001
Leal (2019)59	↓ P < 0.05			↑ P < 0.05				↓ P < 0.05			↑ P < 0.05
Chai (2019)8	↓ P < < ER			NS P = 0.21				FA ratio ↓ P = 0.001			ADC ratio NS P = 0.29
Chen TY (2021)12 ↓ (REZ) P < 0.05	↑ (REZ) P < 0.05	↑ (REZ) P < 0.05	↑ (REZ) P < 0.05	Chen TY (2021)12 ↓ (REZ) P < 0.05
Wu (2020)128	↓ P < 0.001			↑ P < 0.001				↓ P < 0.001			↑ P = 0.001
Lee CC (2018)61 NS P > 0.05	NS P > 0.05	NS P > 0.05	NS P > 0.05	Lee CC (2018)61 NS P > 0.05

AD, axial diffusivity; ADC, apparent diffusion coefficient; ADC ratio, ADCaffected nerve/ADCunaffected nerve; AK, axial kurtosis; FA, fractional anisotropy; FA ratio, FAaffected nerve/FAunaffected nerve; HC, healthy control subjects; λ‖, parallel diffusivity; λ⊥, perpendicular diffusivity; MD, mean diffusivity; MK, mean kurtosis; NR, nonresponders; NS, not significant; QA, quantitative anisotropy; RD, radial diffusivity; Res, responders; REZ, root entry/exit zone; RK, radial kurtosis.

3.2.3. Diffusion tensor imaging studies of the brainstem tract

The study characteristics of DTI analysis of the brainstem and changes in DTI indices of the trigeminal tract within the brainstem are summarized in Tables 4 and 5, respectively.^{12,51,124,125,141} Hung et al. investigated the changes in DTI parameters at the cisternal segment, REZ, and pontine segment (intrapontine tract from the REZ to the pontine nucleus) of the trigeminal tract in classic TN, and they found the most notable changes in DTI indices (lower FA and higher mean diffusivity [MD], radial diffusivity [RD], and axial diffusivity [AD]) at the REZ on the symptomatic side in both long-term responders (patients with TN with favorable outcomes after intervention) and nonresponders (with poor outcomes).⁵¹ In their study, patients with TN who achieved long-term pain relief after microvascular decompression (MVD) or Gamma Knife radiosurgery (GKRS) showed lower AD in the cisternal segment on the symptomatic side compared with the asymptomatic side,⁵¹ which was absent in nonresponders. However, nonresponders demonstrated higher AD in the pontine segment on the symptomatic side compared with control subjects.⁵¹ Another DTI study also reported DTI changes (lower FA and higher RD) in the spinal trigeminal tract in nonresponders.¹⁴¹

Table 4.

Study characteristics of the microstructural analysis of the pontine/brainstem segment in the trigeminocortical pathway in patients with trigeminal neuralgia.

Study	Patients with TN	TN type	Age (y), average ± SD (range)	Disease duration (y), average ± SD (range)	Surgical procedure	Study design/timing of MRI acquisition	MRI scanner	Analysis	Location of measurements	HC
Lee YJ (2019) 62	14	Classic	49 (31–64)	3 (0.5–27)	—	Cross-sectional	7T	DTI, CNV volume	Cisternal, pontine segment	15
Willsey (2021)125	8	Primary type 1	53	—	—	Cross-sectional	3T	DTI	Pontine segment	23 non-TN patients
Hung (2017)51	31	Classic	Res: 60.1 ± 13.7 (36–79)	NR: 55.6 ± 14.9 (25–78)	—	MVD or GKRS (17 Res, 14 NR)	Cross-sectional pre-Tx	3T	DTI	Cisternal, REZ, pontine segment
Chen TY (2021)12	12	9 classic, 3 idiopathic	68.9	—	RFA and/or CKRS	Longitudinal pre-Tx, post-Tx 1, 6, 12 mo	1.5T	DTI, T2-VBM	Cisternal segment, REZ, pontine nuclear zone, pontocerebellar tracts	4
Moon (2018)88	14	Classic	49 (31–64)	—	—	Cross-sectional	7T	DTI	REZ, trigeminal nucleus	15

CKRS, CyberKnife radiosurgery; CNV, trigeminal nerve; DTI, diffusion tensor imaging; GKRS, Gamma Knife radiosurgery; HC, healthy control subjects; MD, mean diffusivity; MVD, microvascular decompression; NR, nonresponders; Res, responders; REZ, root entry/exit zone; RFA, radiofrequency ablation; TN, trigeminal neuralgia; Tx, treatment; VBM, voxel-based morphometry.

Table 5.

Results of the microstructural analysis of the pontine/brainstem segment in the trigeminocortical pathway in patients with trigeminal neuralgia.

Study	Pre-Tx DTI of the pontine segment or trigeminal nucleus vs unaffected side				Pre-Tx DTI of the pontine segment vs control subjects
	FA	RD	AD	MD/ADC	FA	RD	AD	MD/ADC
Lee YJ (2019)62	NS P = 0.13
Willsey (2021)125	↓ P = 0.044	↑ P = 0.025	NS P > 0.05	↑trend P = 0.066	NS P > 0.05	↑P = 0.0046	NS P > 0.05	↑P = 0.048
Hung (2017)51	NS P = 0.14	Res: NS P > 0.05 NR: ↑P = 0.034	NS P = 0.95	NS P > 0.05
Chen TY (2021)12	↓ (pontine) P < 0.05	NS (pontine) P > 0.05	↑ (pontine) P < 0.05	↑ (pontine) P < 0.05
Moon (2018)88	NS (nucleus) P = 0.26	NS (nucleus) P = 0.10	NS (nucleus) P = 0.78	NS (nucleus) P = 0.26

AD, axial diffusivity; ADC, apparent diffusion coefficient; DTI, diffusion tensor imaging; FA, fractional anisotropy; MD, mean diffusivity; NR, nonresponders; NS, not significant; RD, radial diffusivity; Res, responders; Tx, treatment.

3.2.4. Predictive parameters for postinterventional outcomes

Predictive findings in the pretreatment quantitative MRI indices of the affected nerve were summarized in Table 6. In the study by Hung, presurgical DTI characteristics, lower pontine AD and MD for responders and lower cisternal FA for non-responders, predicted surgical outcomes with high accuracy (separating 85% of nonresponders from responders).⁵¹ Willsey et al. reported that higher presurgical RD of the pontine segment combined with longer symptom duration predicted poor MVD outcome.¹²⁵ Similarly, Zhang et al. reported that higher RD of the spinal trigeminal tract and lower NVC severity were predictive of poor MVD outcome.¹⁴¹

Table 6.

Predictive structural and microstructural characteristics of the affected trigeminal nerve in patients with trigeminal neuralgia.

Study	Patients with TN	TN type	Age (y), average ± SD (range)	Disease duration (y), average ± SD (range)	Surgical procedure	Analysis	Location of measurements	HC	Correlation with outcome/differences between res vs NR/other key findings
Duan (2015)31	26 (17 Res, 9 NR)	Classic	65.3 (41–88)	1.8 (0.5–14)	MVD	CSA	Proximal and distal cisternal segments	—	Atrophy at the distal site correlated with poor outcome (P < 0.05); greater atrophy of the proximal trigeminal nerve in Res did not achieve statistical significance
Leal (2014)58	50	Classic	56.42 (22–79)	4.50 (0.5–14)	MVD	CSA, CNV volume	CSA: REZ, volume: cisternal segment	20	Smaller ipsilateral pre-Tx CSA in pain-free patients at the 2-y follow-up compared with partially relieved patients or patients with treatment failure (P = 0.02)
Wang (2016)118	42	Idiopathic	56.5 (40–68)	6.58 ± 4.43	MVD	CSA, CNV volume	CSA: REZ, volume: cisternal segment	42	NS CSA difference between res and NR (P = 0.80)
Danyluk (2020)21	31	Primary	Res: 53 NR: 52	Res: 6.6, NR: 5.6	MVD	CNV volume	Cisternal segment	—	NS differences in ipsilateral CNV volumes between Res and NR (P = 0.83); larger contralateral CNV volume in NR (P = 0.009).
Cheng (2017)14	60	Primary	48.3 ± 9.1 (23–70)	4.5 (0.5–25)	MVD	CNV volume	Cisternal segment	30	Degree of nerve atrophy correlated with long-term surgical outcome (OR = 1.18, P = 0.035); trigeminal nerve indentation associated with long-term outcome (OR = 2.968, P = 0.022)
Zhang (2023)141	97 (73 Res, 24 NR)	Primary	Res: 55.2 ± 9.5 NR: 53.2 ± 11.4	Res: 6.6 ± 1.8 NR: 6.0 ± 2.0	MVD	DTI	Spinal trigeminal tract	—	Significantly lower FA (P < 0.001), higher RD (P < 0.001) in the spinal trigeminal tract in NR compared with Res
Willsey (2021)125	22	Primary type 1	60.1 ± 3.4	0.5–12.5	13 MVD (Res: 6/13, NR: 7/13) 3 RFA 4 other Tx	DTI: RD histogram	Pontine segment	6 type 2 15 non-TN patient controls	Pre-Tx normalized peak RD (normalized by the mean background level of the pons) of the pontine segment and pre-Tx symptom duration predicted post-MVD outcome with 85% accuracy, 83% sensitivity, and 86% specificity (leave-one-out cross-validation; P = 0.029)
Hung (2017)51	31 (17 Res, 14 NR)	Classic	Res: 60.1 ± 13.7 NR: 55.6 ± 14.9	—	MVD or GKRS	DTI	Cisternal, REZ, pontine segment	16	Res: Decrease in AD and MD of the ipsilateral cisternal segment compared with the asymptomatic side (P ,0.05); NR: declined FA in the ipsilateral REZ and increased AD in the ipsilateral pontine segment compared with HC (P < 0.05)
Lee C (2023)60	50 (41 Res, 9 NR)	Classic	Res: 61.1 ± 10.1 NR: 55.7 ± 9.7	Res: 5.95 ± 4.64 NR: 5.15 ± 3.38	MVD	CSA	TZ	—	20% of the CSA change rate: most appropriate cutoff value to predict surgical outcomes
Hu (2019)48	67	Classic	22–90	0.17–21	GKRS	CSA	CSA: REZ	—	Smaller CSA correlated with initial pain relief (P = 0.03) and subsequent recurrence (P = 0.02)
Tohyama (2018)113	37	Primary	68.3 ± 13.1	—	GKRS	DTI	Cisternal segment	—	Res: higher FA (P 5 0.013), lower MD (P = 0.046), and lower RD (P = 0.009) in the post-GKRF CNV compared with the unaffected side; NR: NS difference in diffusion indices in the post-GKRF CNV compared with the unaffected side (P > 0.05); Res vs NR: lower post-GKRS FA (P = 0.013) and trend toward higher RD (P = 0.057) in Res compared with NR; NS group difference in post-GKRF MD and AD
Pikis (2021)99	16	Classic	65.8 ± 13.0 (39–85)	—	GKRS	DTI	Cisternal, REZ, pontine segment	—	Lower mean FA at the pontine segment (P = 0.04) and increased mean RD at the REZ (P = 0.032) of the treated CNV at post-Tx 3 mo in Res (BNI I-III) as compared with the NR (BNI IV or V)
Chen ST (2016)11	43	Primary	58.8 ± 11.0	—	36 RFA (28 Res, 8 NR)	DTI, CNV volume	Cisternal segment	—	NS differences in DTI metrics (FA:P = 0.072, AD: 0.842, RD: 0.450, ADC: 0.552) or volume (P = 0.95) between Res and NR
Lee YL (2019)63	51 (41 Res, 10 NR)	Classic	—	Res: 6.96 ± 6.63 NR: 6.48 ± 7.45	RFA	DTI histogram	Cisternal segment	—	NS in FA, ADC, AD, RD between Res and NR; decreased kurtosis of FA (P = 0.0004) and ADC (P = 0.015) in NR compared with Res
Chen ST (2019)10	37 (25 Res, 12 NR)	Primary	59.8 (43–87)	7.7 ± 7.5	RFA	DTI, CNV volume	Cisternal segment	—	NS differences in volume or DTI metrics between Res and NR before RFA (P > 0.05); larger post-RFA nerve volume in NR compare with Res (P = 0.016); NS differences in the post-RFA DTI indices between the groups (P > 0.05)
Li (2023)68	58 (48 Res, 10 NR)	Primary	Res: 68.2 ± 12.1 NR: 65.6 ± 9.5	Res: 4.90 ± 5.62 NR: 6.70 ± 5.08	PBC	CSA	Proximal and distal cisternal segments	—	Significantly smaller CSA ratio (NVC/non-NVC) at REZ in Res (P < 0.05); NS between-group difference in CSA ratio of the distal segment

AD, axial diffusivity; ADC, apparent diffusion coefficient; BNI, Barrow Neurological Institute; CNV, trigeminal nerve; CSA, cross-sectional area DTI, diffusion tensor imaging; FA, fractional anisotropy; GKRS, Gamma Knife radiosurgery; HC, healthy control subjects; IN, internal neurolysis; MD, mean diffusivity; MVD, microvascular decompression; NR, nonresponders; NS, not significant; NVC, neurovascular compression; PBC, percutaneous balloon compression; RD, radial diffusivity; Res, responders; REZ, root entry/exit zone; RFA, radiofrequency ablation; TN, trigeminal neuralgia; TZ, transition zone.

3.2.5. Changes in quantitative parameters after intervention

Changes in quantitative MRI indices of the nerve after MVD surgery, radiofrequency ablation (RFA), and GKRS/CyberKnife radiosurgery (CKRS) were summarized in Table 7.^{8,10,12,25,50,59,61,99,107,113,119,128,138} The studies demonstrated intervention-specific changes in DTI metrics after procedures. Recovery in DTI indices was correlated with pain relief after treatment,^25,119,128 whereas nonresponders demonstrated different postinterventional trajectories in DTI analysis.^{8,10,50,60,107}

Table 7.

Treatment-related structural and microstructural changes of the affected trigeminal nerve in patients with trigeminal neuralgia.

Study	Patients with TN	TN type	Surgical procedure	Study design/timing of MRI acquisition	Analysis	HC	Changes between before and after treatment/correlation between treatment-related quantitative MRI changes and clinical scores
Lee C (2023)57	50 (41 Res, 9 NR)	Classic	MVD	Longitudinal pre-Tx, immediate post-Tx	CSA	—	Significant increase in post-Tx CSA of the symptomatic CNV in responders (P < 0.01) NS CSA changes in NR(P = 0.69)
Zhang Y (2018)138	24 R	Classic	MVD	Cross-sectional pre-Tx, 6 patients: longitudinal pre-, post-Tx 1 wk and 4–6 mo	DTI	28	Post-Tx 1 wk: DTI recovery mainly because of a decrease in λ‖ (consistent with axonal membrane stabilization) Post-Tx 4–6 mo: DTI recovery because of predominant reduction in λ⊥ (consistent with remyelination) with increase in FA
Leal (2019)59	10	Classic	MVD	Longitudinal pre-, post-Tx 4 y	DTI, CNV volume, CSA	6	4 y after MVD: FA in the affected side remained significantly lower compared with FA in the unaffected side (P < 0.05) and mean FA of control subjects (P < 0.05) ADC in the affected side increased to the level of ADC in the unaffected side (P >0.05) and mean ADC of control subjects (P > 0.05) Nerve volume and CSA was smaller compared with the unaffected side (P < 0.05) and control subjects (P < 0.05)
Chai (2019)8	34 (33 Res, 1 NR)	Classic	MVD	Longitudinal pre-Tx, post-Tx 1 wk and 6–8 mo	DTI	34	Res: significant increase in FA ratio at 7 d (P = 0.02) and 6–8 mo (P = 0.002) after MVD, decrease in ADC ratio in 6–8 mo (P = 0.004) NR: no change in FA ratio or ADC ratio
Wang (2023)119	17	Primary	9 MVD, 2 RFT	Cross-sectional pre-Tx, 7 patients: longitudinal pre-Tx, post-Tx 6 mo	DTI	34	An increase in FA (F(48.000) = 78.867, P = 1.063 × 10−11) and a decrease in AD (F(48.000) = 210.99, P <2.200 × 10−16), RD (F(48.000) = 178.04, P < 2.200 × 10−16), and MD (F(48.000) = 282.24, P < 2.200 × 10−16) 6 mo after surgery
DeSouza (2015)25	14 R	Classic	MVD (9 Res, 1 NR) or GKRS (6 Res, 9 NR)	Longitudinal pre-, post-Tx 5.8 ± 4 (2–6) mo	DTI	14	Res: Recovery from FA reduction in the affected trigeminal REZ and bilateral MD, RD, and AD abnormalities NR: Persistent DTI abnormalities (decreased FA, increased MD and RD, further increase in AD) More pain relief associated with greater resolution of MD (r = −0.53, P=0.02), RD (r = −0.51, P = 0.031), and AD (r = −0.54, P = 0.02) abnormalities after treatment; greater pain relief associated with larger increase trends in FA toward the level of control subjects at REZ (r = 0.43, P = 0.06)
Tohyama (2018)113	37	Primary	GKRS	Cross-sectional post-Tx 16 patients: longitudinal pre, post-Tx 6 mo	DTI	—	Res: significant decrease in FA (P = 0.023) and increase in RD (P = 0.044) from pre-GKRS topost-GKRS in the affected CNV NR: NS pre-GKRS and post-GKRS differences in the DTI metrics of the affected CNV (P > 0.05)
Hung (2019)50	55 (31 R, 24 L)	Classic	GKRS	Cross-sectional post-Tx 6 mo (28 Res, 21 NR), 12 mo (11 Res, 11 NR), 24 mo (14 Res, 7 NR)	DTI	—	Res: Significant lower FA at the GKRS target zone compared with the unaffected side 12 and 24 mo after GKRS (P = 0.023, P = 0.017) (reduction rates: 9.9% at 6 mo, 29.7% at 12 mo, and 32.3% at 24 mo) NR: NS changes in diffusion metrics at the GKRS target zone NS at REZ for both Res and NR
Pikis (2021)99	16	Classic	GKRS	Longitudinal pre-Tx, post-Tx 3 mo	DTI	—	Significantly reduced mean FA (P = 0.0026) and elevated mean RD (P = 0.0025) at the REZ of the affected CNV as compared with the contralateral unaffected nerve at the 3-mo follow-up
Song (2014)107	20 (15 Res, 5 NR)	Primary	GKRS	Longitudinal pre-Tx, post-Tx 6, 12, 24, 36 mo	CSA	—	Decreased CSA of the treated nerve compared to pre-GKRS (P < 0.01) and to the unaffected nerve (P < 0.01) Smaller post-GKRS CSA and lower post-GKRS T1 signals over time were seen in patients with greater pain relief and less facial numbness recurrence
Lee CC (2018)61	22	Classic	GKRS	Longitudinal pre-Tx, post-Tx 6 mo	DTI	—	A reduction in cisternal FA (P = 0.04) and an increase in cisternal λ2 (P = 0.02) in patients with ,5 y of TN history after GKRS
Chen TY (2021)12	12	9 classic, 3 idiopathic	RFA and/or CKRS	Longitudinal pre-Tx, post-Tx 1, 6, 12 mo	DTI, T2-VBM	4	REZ: FA increased (side difference no longer significant) at post-Tx 1 mo, but decreased after 6 mo; NS posttreatment changes in higher pre-Tx RD, AD, and MD of the affected nerve until post-Tx 6 mo Bilateral pontine nuclei: Significant decrease in FA (P > 0.05) 1 mo after treatment and increase in FA (P > 0.05) at post-Tx 6 mo Pontocerebellar tracts: post-Tx increase in MD (P = 0.00069)
Chen ST (2019)10	37 (25 Res, 12 NR)	Primary	RFA	Longitudinal pre-Tx, post-Tx 2 wk	DTI, CNV volume	—	An increase in volume (P 5 0.014) and FA (P = 0.015), and a decrease in AD (P = 0.016), RD (P = 0.001), and ADC (P = 0.001) after RFA compared with pre-Tx; post-RFA RD and ADC recovered to the normal level; post-RFA FA (P = 0.012) and AD (P = 0.001) lower than the unaffected side
Wu (2020)128	21	Idiopathic	IN	Longitudinal pre-Tx, post-Tx 1 y	DTI, DKI	20	Post-Tx increase in FA: side difference no longer significant, but still lower than FA in control subjects (P < 0.001) Post-Tx decrease in ADC: still higher than ADC on the contralateral side (P < 0.001), but difference no longer significant compared with control subjects Strong negative correlation (r = 0.5237, P = 0.0148) between decrease in ADC on the affected side and the surgical outcomes in BNI score

AD, axial diffusivity; ADC, apparent diffusion coefficient; BNI, Barrow Neurological Institute; CKRS, CyberKnife radiosurgery; CNV, trigeminal nerve; CSA, cross-sectional area; DKI, diffusion kurtosis imaging; DTI, diffusion tensor imaging; FA, fractional anisotropy; GKRS, Gamma Knife radiosurgery; HC, healthy control subjects; IN, internal neurolysis; L, left-sided trigeminal neuralgia; λ‖, parallel diffusion; λ⊥ perpendicular diffusion; MD, mean diffusivity; MVD, microvascular decompression; NR, nonresponders; NS, not significant; NVC, neurovascular compression; R, right-sided trigeminal neuralgia; RD, radial diffusivity; Res, responders; REZ, root entry/exit zone; RFA, radiofrequency ablation; RFT, radiofrequency thermocoagulation; TN, trigeminal neuralgia; TZ, transition zone; Tx, treatment; VBM, voxel-based morphometry.

3.3. Diffusion analysis of the whole brain

For quantitative MRI analysis of the brain in patients with TN, the study characteristics of the microstructural analysis of the brain and a summary of the results of those studies are demonstrated in Tables 8 and 9, respectively.^{26,32,44,66,67,73,97,112,117,123,128,130} Multiple DTI studies in TN reported microstructural changes in the superior longitudinal fasciculus, corona radiata, internal capsule, thalamic radiation, external capsule, corpus callosum, cingulum, uncinate fasciculus, and fornix, on the ipsilateral, contralateral to TN or both sides. Lower FA or increased diffusivity (mainly RD and MD) were observed in those white matter pathways, with increased AD and decreased kurtosis (a broadening of the peak) in some of the studies. The DTI study of the subdivisions of the white matter tracts in 23 right-sided and 14 left-sided patients with TN demonstrated changes in DTI metrics localized in the middle (P = 0.001) and posterior cingulum (P < 0.0001) and the medial forebrain bundle near the ventral tegmental area (P = 0.001) with laterality, which were accompanied with decreased volumes in the associated gray matter structures, including the hippocampus, posterior cingulate, nucleus accumbens, and ventral diencephalon.⁴⁴

Table 8.

Key characteristics of the studies for microstructural analysis of the brain.

Study	Patients with TN	TN type	Age (y), average ± SD (range)	Disease duration (y), average ± SD (range)	Surgical procedure	Study design/timing of MRI acquisition	Analysis	Target structures	HC
DeSouza (2014)26	18 R	Idiopathic	54.1 ± 17.0	—	—	Cross-sectional	DTI	Whole brain	18
Liu (2018) 73	29 R	Classic	59.4 ± 1.9 (35–77)	10.2 ± 9.6(1–30)	MVD	Cross-sectional pre-Tx	DTI	Whole brain	35
Wang (2017)117	38	Classic	55.87 ± 8.38	7.05 ± 5.32		Cross-sectional	DTI, VBM, fMRI	Whole brain	38
Tian (2016)112	20	Primary	52.6 (36–65)	1.76 ± 1.35(0.17–1.33)	—	Cross-sectional	DKI, FCD	Whole brain	22
Wu (2020)127	45 rh (23 wNVC, 22 woNVC)	23 classic, 22 idiopathic	wNVC: 53.30 ± 8.66, woNVC: 47.77 ± 9.24	wNVC: 5.74 ± 3.35, woNVC: 4.97 ± 2.09	23 MVD, 22 IN	Cross-sectional pre-Tx	DTI, SC, VBM, SBM	Whole brain	45
Li R (2021)67	34	Primary	53.53 ± 11.0	—	—	Cross-sectional	DTI, AFQ	20 fiber tracts	34
Wang(2018)123	20	Primary	56.0 ± 11.75	6.25 ± 5.89		Cross-sectional	SBM, DTI, fMRI	Tracts correlated with left insular cortex	21
Parise (2014)97	24	Primary	55.8 ± 8.5	1.0–17	4 with MVD	Cross-sectional/rescan for 4 patients 6 mo after MVD	DTI, SBM	Whole brain	24
Xiong (2023)130	36 L	Classic	53.71 ± 7.63	3.42 ± 3.20	—	Cross-sectional	DKI	Whole brain	36
Li R (2021) 66	22 R	Classic	52.6 ± 9.0	9.2 ± 8.6	—	Cross-sectional	DTI	CC	22
Filimonova (2023)32	60 (36 R, 23 L, 1 B)	Primary	57.8 ± 10.5(30–75)	8.3 ± 5.3	—	Cross-sectional	DTI	Whole brain	24
Hayes (2017)44	37(23 R, 14 L)	Primary	R: 47 ± 12 (23–67), L: 55 ± 10 (36–70)	0.5–12		Cross-sectional	DTI, VBM	Fornix, cingulum, medial forebrain bundle	28

AFQ, automated fiber tract quantification; B, bilateral; DKI, diffusion kurtosis imaging; DTI, diffusion tensor imaging; FCD, functional connectivity density; fMRI, functional MRI; HC, healthy control subjects; IN, internal neurolysis; L, left-sided trigeminal neuralgia; MVD, microvascular decompression; NVC, neurovascular compression; R, right-sided trigeminal neuralgia; SBM, surface-based morphometry; SC, structural connectivity; TN, trigeminal neuralgia; Tx, treatment; VBM, voxel-based morphometry; wNVC, with NVC; woNVC, without NVC.

Table 9.

Summary of major brain structures with significant changes in diffusion indices in patients with trigeminal neuralgia compared with control subjects.

Study (reference number)	26				73		117		112	127				67		123	130				32
										wNVC		woNVC
DTI parameters	FA	RD	AD	MD	FA	RD	FA	MD	AD	FA	RD	FA	RD	FA	MD	FA	FA		AD	MD	FA
CC genu	↓	↑		↑	↓		↓	↑									↓
Callosum forceps minor														↓
CC body	↓	↑		↑	↓	↑	↓	↑		↓
CC splenium	↓	↑	↑	↑	↓	↑	↓	↑		↓	↑		↑					↑	↑	↑
Parietal, posterior CC
Forceps major									↑
Fornix					↓ C	↑ C
Cingulum	↓ B	↑ B		↑ B	↓ C	↑ C				↓ R							↓	↑	↑	↑
Cingulate gyrus									↑ C
Corticospinal tract						↑								↓ L	↑ L
Internal capsule																					↓ B
IC anterior limb					↓ B	↑ B		↑ B		↓ R	↑ R		↑ L
IC genu										↓ L
IC posterior limb					↓ C	↑ B		↑ B								↓ L
Retrolenticular portion						↑ C
Corona radiata
Superior corona radiata					↓ B	↑ B	↓ B	↑ B		↓ B			↑ B			↓ L	↓	↑	↑	↑	↓ B
Anterior corona radiata					↓ B	↑ B	↓ B	↑ B				↓ B
Posterior corona radiata	↓ B	↑ B		↑ B			↓ B	↑ B			↑ B
External capsule					↓ C	↑ B		↑ B					↑ L			↓ L
Anterior thalamic radiation									↑ B				↑ R
Posterior thalamic radiation											↑ L						↓	↑	↑	↑
Thalamic sensory tract														↓ L	↑ L
Sensorimotor
Thalamus								↑ B													↓ B
Uncinate fasciculus									↑ B			↓ B		↓ R
SLF	↓ C	↑ C			↓ C			↑ B	↑ B						↑ B		↓ B	↑ B	↑ B	↑ B
ILF									↑ B
IFOF									↑ B
Internal sagittal stratum					↓
Optic radiation																					↓ B
Brainstem CST								↑ B
Pontine crossing tract						↑
Cerebral peduncle					↓ C	↓ C
Superior cerebellar peduncle								↑ B													↓ B
Middle cerebellar peduncle																					↓ B
Inferior cerebellar peduncle								↑ B

AD, axial diffusivity; AK, axial kurtosis; B, bilateral; C, contralateral; CC, corpus callosum; CST, corticospinal tract; DTI, diffusion tensor imaging; FA, fractional anisotropy; I, ipsilateral; IC, internal capsule; IFOF, inferior fronto-occipital fasciculus; ILF, inferior longitudinal fasciculus; L, left; MD, mean diffusivity; MK, mean kurtosis; NS, not significant; NVC, neurovascular compression; R, right; RD, radial diffusivity; RK, radial kurtosis; SLF, superior longitudinal fasciculus; TN, trigeminal neuralgia; wNVC, with NVC; woNVC, without NVC; ↑, significant increase; ↓, significant decrease.

Disease duration was correlated with decreased FA in the corona radiata, genu of the corpus callosum, external capsule and cerebral peduncle, and with increased RD in the corona radiata, external capsule, fornix cerebri and cerebral peduncle, on mostly contralateral side.^73,142 Pain severity was correlated with decreased FA in the corona radiata, thalamic radiation, external capsule and cerebral peduncle, increased RD in the corona radiata and external capsule on the contralateral side, and increased MD in the thalamic radiation.^67,73

3.4. Structural analysis of the brain

For quantitative MRI analysis of the brain in patients with TN, the study characteristics of the structural analysis of the brain and a summary of the results of those studies were demonstrated in Tables 10 and 11, respectively.^{3,12,19,21,25,28,35,44,64,65,71,84,87,90, 91,97,103,115–117,121,123,127,131,133,139}

Table 10.

Key characteristics of the studies for structural analysis of the brain.

Study	Patients with TN	TN type	Age (y), average ± SD (range)	Disease duration (y), average ± SD (range)	Surgical procedure	Study design/timing of MRI acquisition	Analysis	Target structures	HC
DeSouza (2013)28	24 R rh	Idiopathic	48.5 ± 12.7 (24–68)	6.3 ± 3.0(1–13)	—	Cross-sectional	CT, subcortical VBM	Whole brain	24
DeSouza (2015)25	25 R (8 Res, 6 NR)	Classic	57.6 ± 11.5 (37–71)	(1–30)	MVD or GKRS	Longitudinal pre- Tx, post-Tx5.8 ± 4 (2–6) mo	CT, subcortical VBM	Cortical GM ROIs (right ventral anterior insula, PCC, left S1, bilateral M1, posterior insula, FP, ACC, and OFC) and subcortical GM	14
Hayes (2017)44	37 (23 R, 14 L)	Primary	R: 47 ± 12(23–67) L: 55 ± 10(36–70)	(0.5–12)	—	Cross-sectional	DTI, VBM	GM ROIs: areas immediately innervated by the analyzed WM structures, the cingulum (ACC, MCC, PCC, and isthmus), MFB (NAc and VD, substantia nigra, hypothalamus, and mammillary bodies), and fornix (hippocampus)	28
Vaculik (2019)116	22 R	Classic	47.6 ± 13.5 (24–76)	F: 4.5 ± 2.7 M: 3.1 ± 2.3	—	Cross-sectional	VBM	Hippocampal subfields	22
Noorani (2022)90	61 (47 Res, 18 NR) (26 L, 35 R)	Primary	64.9 ± 12.0	7.0 ± 7.8	GKRS	Longitudinal pre- Tx, post-Tx 6 mo	VBM	12 hippocampal subfields	61
Danyluk (2020)21	34 (23 Res, 11 NR)	32 classic, 2 idiopathic	Res: 52.59 ± 9.50 NR: 51.86 ± 16.7	Res: 6.61 ± 5.07 NR: 5.55 ± 4.63	MVD	Cross-sectional pre-Tx	CNV volume, structural analysis of the brain	Hippocampus, amygdala, and thalamus
Danyluk (2021)19	23 (17 Res, 6 NR)	Primary	Res: 58.6 ± 9.7 NR: 47.4 ± 10.3	Res: 4.6 ± 3.3 NR: 8.8 ± 4.7	MVD/IN	Cross-sectional for VBM, longitudinal for MRS (pre-Tx, post-Tx 1 wk)	VBM, MRS	Thalamic volume and shape, thalamic MRS	20
Obermann (2013)91	60	Classic	62 ± 13.2 (3186)	8.3 ± 6.7(1–29)	—	Cross-sectional	VBM	Whole brain	49
Tsai (2018) 115	62 (36 R, 26 L)	Classic	R: 58.0 ± 7.7 L: 59.0 ± 6.6	R: 5.80 ± 6.29 L: 5.27 ± 4.92	—	Cross-sectional	VBM, fMRI	Whole brain	19
ChenTY (2021)12	12	9 classic, 3 idiopathic	68.9		RFA and/or CKRS	Longitudinal pre- Tx, post-Tx 1, 6, 12 mo	T2-VBM, CNV- DTI	ROIs: pain-related regions including PAG, PFC, PCC, ACC, insula, amygdala, thalamus, PPC, S1, S2, SMA, and cerebellum	4
Liu (2022) 71	34	Classic	53.06 ± 10.91	4.63 ± 3.53	—	Cross-sectional	VBM, fMRI	Whole brain	29
Wang Y (2017)117	38	Classic	55.87 ± 8.38	7.05 ± 5.32	—	Cross-sectional	VBM, DTI, fMRI	Whole brain	38
Wang Y (2019)121	40	22 major NVC, 14 minor NVC, 4 idiopathic	55.76 ± 8.23	7.08 ± 5.29	—	Cross-sectional	VBM, CNV volume	Whole brain	40
Li (2017)64	28	Primary	45.86 ± 11.17	8.43 ± 3.65 (218)	—	Cross-sectional	VBM	Whole brain	28
Li (2021)65	30	Primary	51.63 ± 8.16	4.72 ± 3.69	—	Cross-sectional	SBM (CT, LGI, sulcal depth)	Whole brain	30
Wang Y (2018)123	20 (7 L, 13 R)	Primary	56 ± 11.75	6.25 ± 5.89	—	Cross-sectional	SBM (LGI, CT, SA), DTI, fMRI	Whole brain for LGI, ROI for CT and SA	21
Shen (2022)103	49 rh	Idiopathic	58.4 ± 11.5	TN1 (0–2 y): 1.1 ± 0.7 (n = 16) TN2 (3–7 y): 4.8 ± 1.5 (n = 17) TN3 (10–30 y): 15.1 ± 5.5 (n = 16)	—	Cross-sectional	VBM	Whole brain	18
Wu (2020)127	45 rh (23 wNVC, 22 woNVC)	23 classic, 22 idiopathic	wNVC: 53.30 ± 8.66 woNVC: 47.77 ± 9.24	wNVC: 5.74 ± 3.35 woNVC: 4.97 ± 2.09	23 MVD, 22 IN	Cross-sectional pre-Tx	DTI, SC, VBM, SBM	Whole brain	45
Ge (2023)35	79 rh	Classic	54.05 ± 10.56	5.03 ± 4.94	—	Cross-sectional	VBM, DBM, SBM	Whole brain	81
Mo (2021)84	43 (21 L, 22 R)	Classic	59.05 ± 9.95	5.41 ± 4.71	MVD	Cross-sectional pre-Tx	SSA	Whole brain	40
Parise (2014)97	24 (13 L, 11 R)	Classic	55.8 ± 8.5	(1–17)	—	Cross-sectional	SBM, DTI	Whole brain	24
Zhang (2018)139	29	Classic	48.1 ± 11.9	6.02 ± 4.35	MVD	Longitudinal pre- Tx, post-Tx 4–6 mo	VBM, fMRI	Whole brain	34
Albano (2022)3	30 (12 R, 18 L)	18 classic, 12 idiopathic	66.24 ± 10.57 (31–74)	6.84 ± 6.81 (0.83–10)	GKRS	Cross-sectional pre-Tx	SBM	Whole brain	15
Moon (2018)87	15 (7 R, 8 L)	Classic	49.0 (36–64)	5.46 ± 1.72	—	Cross-sectional	SBM	ROI: rostral and caudal ACC, PCC whole brain vs clinical features	16
Xu (2023)131	50 (28 R, 22 L)	Primary	55.90 ± 9.52	7.12 ± 5.47	—	Cross-sectional	SBM (CMC, CT, CV, SA)	Whole brain	48
Xu (2022)133	48 (25 R, 23 L)	Classic	54.96 ± 10.40	7.12 ± 4.94		Cross-sectional	Volumetry, vertex-wise shape analyses	subcortical structures	46

ACC, anterior cingulate cortex; CC, corpus callosum; CKRS, CyberKnife radiosurgery; CMC, cortical mean curve; CNV, trigeminal nerve; CT, cortical thickness; CV, cortical volume; DBM, deformation-based morphometry; DTI, diffusion tensor imaging; FG, fusiform gyrus; fMRI, functional MRI; FP, frontal pole; GKRS, Gamma Knife radiosurgery; GM, gray matter; HC, healthy control subjects; IC, internal capsule; IFG, inferior frontal gyrus; IN, internal neurolysis; L, left-sided trigeminal neuralgia; MCC, middle cingulate cortex; MFB, medial forebrain bundle; LGI, local gyrification index; MRS, MR spectroscopy; MVD, microvascular decompression; NAc, nucleus accumbens; NR, nonresponders; NVC, neurovascular compression; OFC, orbitofrontal cortex; PAG, periaqueductal gray matter; PCC, posterior cingulate cortex; PFC, prefrontal cortex; R, right-sided trigeminal neuralgia; Res, responders; RFA, radiofrequency ablation; rh, right-handed; ROI, region of interest; SA, surface area; SBM, surface-based morphometry; SMA, supplementary motor area; SSA, surface-based shape analysis; S1, primary somatosensory cortex; S2, secondary somatosensory cortex; TN, trigeminal neuralgia; Tx, treatment; VBM, voxel-based morphometry; VD, ventral diencephalon; WM white matter; wNVC, with NVC; woNVC, without NVC.

Table 11.

Summary of major brain structures with significant changes in gray matter volume and/or cortical thickness in patients with trigeminal neuralgia compared with control subjects.

Study (reference number)	28	25	44	116	90	91	115	71	117	121	64	65	123	103			127		35	84	97	139	3	87	131
														TN1	TN2	TN3	W NVC	Wo NVC
Thalamus	↑ B	↑ B				↓ L	↓ B								↓ L		↓ L			↓
MD	↑ B
VPM	↑ B
VL	↑ B
Pulvinar	↑ B
Hypothalamus							↓ C
Diencephalon			↓ B
Putamen	↑ B	↑ B				↓ L	↓ I													↓
NAc	↑ I		↓ B				↓ I													↓
Caudate nucleus	↑ I					↓ L					↓ L
Pallidum																				↓
Cingulate
ACC	↓ B					↓ B			↓ B	↓ R	↓ L					↓ R				↓			↓ L
MCC									↓ B						↓ R					↓
PCC	↓ I	↓ I	↓ B																	↓			↓ L	↓ B
Isthmus																							↓ B
Corpus callosum
Insula	↓ I	↓ B				↓ B		↓ L	↓ R	↓ L						↓ B						↓ R	↓ R		↓ L
Frontal lobe
PFC							↓															↓ B
OFC	↓ B	↓ C				↓ R						↓ L			↓ B	↓ B						↓ R
Frontal pole	↓ B	↑ B
DLPFC						↓ R
SFG							↓ C							↓ L	↓ R	↓ L						↓ R
MFG																↓ R	↓ L
IFG							↓ I					↓ L				↓ B
M1		↑ B	↑ C				↓ L		↓ L						↓ R				↓ R
PMA									↓ L
Parietal lobe
S1	↑C	↑C				↓ L									↓ B				↓ L				↓ B
S2						↓ L			↓ R	↓ L
SPL																							↓ B
IPL																							↓ R
PPC																↓ B
Angular gyrus						↓ L				↓ B						↓ L	↓ L
Precuneus																							↓ B
Temporal lobes										↓ B
Temporal pole																			↓ R ↑ L
STG								↓ L	↓ L																↓ R
											↓ B												↓ B
MTG								↓ L		↓ L	↓ B						↓ B						↓ L
ITG								↓ R	↓ L	↓ R								↓ B
Fusiform gyrus								↓ R			↓ R										↓ L		↓ R		↓ R
Hippocampus			↓B	↓ I	↓ B			↓ R		↓ R						↓ L				↓
PHG											↓ B
Amygdala	↑ I																					↓ B
TTG																									↓ L
Occipital cortex															↓ R		↓ B						↓ B		↓ B
SOG
Cuneus																					↓ L
Midbrain/PAG	↑	↑															↓					↓
Ventral pons	↓
Cerebellum						↓ B					↓ R			↓ R			↓ R
Tonsil							↓ B
Pyramis							↓ I										↓ R
Culmen							↓ C

ACC, anterior cingulate cortex; B, bilateral; C, contralateral; DLPFC, dorsolateral prefrontal cortex; I, ipsilateral; IFG, inferior frontal gyrus; IPL, inferior parietal lobule; ITG, inferior temporal gyrus; L, left; MCC, mid-cingulate cortex; MD, medial dorsal nucleus; MFG, medial frontal gyrus; MTG, middle temporal gyrus; M1, primary motor cortex; NAc, nucleus accumbens; NS, not significant; NVC, neurovascular compression; OFC, orbitofrontal cortex; PAG, periaqueductal gray; PCC, posterior cingulate cortex; PFC, prefrontal cortex; PHG, parahippocampal gyrus; PMA, premotor area; PPC, posterior parietal cortex; R, right; SFG, superior frontal gyrus; SOG, superior occipital gyrus; SPL, superior parietal lobule; STG, superior temporal gyrus; S1, primary sensory cortex; S2, secondary somatosensory cortex; TN, trigeminal neuralgia; TN1, TN with disease duration 0 to 2 years; TN2, TN with disease duration 3 to 7 years; TN3, TN with disease duration 10 to 30 years; TTG, transverse temporal gyrus; VL, ventral lateral nucleus; VPM, ventral posteromedial nucleus; wNVC, with NVC; woNVC, without NVC; ↑, significant increase; ↓, significant decrease.

There were mixed results in the thalamic volumetry; some studies reported larger volume in the bilateral thalami²⁸ or the contralateral thalamus,^19,21 and others reported decreased volume in the thalamus,⁹¹ bilaterally^84,115 or on the left.^103,127 Voxel-based morphometry studies of the subsegments of the thalamus demonstrated increased or decreased volume in the ventrodorsal areas (the ventroposterior medial and pulvinar subnucleus) and the central/medial areas of the thalamus compared with the control subjects.^28,115 The vertex-wise shape analysis demonstrated significantly less presurgical contralateral thalamus volume in nonresponders compared with responders in an axially oriented band spanning the outer thalamic circumference (peak P value = 0.019).¹⁹ Changes in volumes were also observed in the basal ganglia (caudate, putamen, nucleus accumbens, and palladium).^{25,28,44,64,84,91,115} Other brain structures that have shown changes in gray matter volume/cortical thickness included the prefrontal cortex, insula, anterior cingulate cortex, hippocampus, primary/secondary somatosensory cortex, precuneus, fusiform gyrus, and cerebellum.^{3,12,19,21,25,28,35,44,64,65,71,84,87,90,91,97,103,115–117,121,123,127,131,133,139}

Disease duration was correlated with decreased gray matter volume in various brain areas, including the superior, middle, and inferior temporal gyrus, fusiform gyrus, hippocampus, para-hippocampus, medial frontal cortex, orbitofrontal cortex, superior and inferior frontal cortex, thalamus, caudate, insula, anterior cingulate cortex, and corpus callosum.^{28,35,64,65,71,87,90,91,97,103,116,117,121} The study by Shen et al. showed the effects of disease duration on the different regions of the brain; patients with TN and with shorter duration (within 2 years) showed reduced cerebellar gray matter volume compared with the control subjects; in patients with TN suffering longer period (3–7 years) reduced gray matter volume was observed in the left thalamus, bilateral postcentral cortex, bilateral orbitofrontal cortex, and left superior frontal cortex; long-lasting TN (10–30 years) affected gray matter volume in the right anterior cingulate cortex, left hippocampus, bilateral insular cortex, and orbitofrontal cortex.¹⁰³ Pain severity was associated with reduced gray matter volume in the inferior temporal gyrus, hippocampus, prefrontal cortex, insula, and the head of the caudate.^65,71,117 The study of the machine leaning models using pretreatment regional brain gray matter morphology in 51 patients with TN (35 responders and 16 nonresponders) revealed successful prediction for the treatment outcomes by the regional surface area model with an accuracy of 96.7%, 100.0% sensitivity, and 89.1% specificity and by the regional cortical thickness model with an accuracy of 90.5%, 93.5% sensitivity, and 83.7% specificity, more accurate than the nerve-based models.⁵² The best predictors were the surface area of the contralateral superior frontal gyrus and the cortical thickness of the contralateral isthmus cingulate gyrus, the regions associated with the default mode network (DMN).⁵²

3.5. Analysis of the structural connectivity of the brain

The structural connectivity studies were summarized in Table 12.^{127,132,137,142} Zhong et al. reported that the normalized streamline counts classified classic patients with TN from healthy control subjects with an accuracy of 88%, driven by lower normalized streamline counts of the connections from the bilateral hemisphere with greater weight in the left and interhemispheres (including subcortical [thalamus, caudate, and amygdala], precuneus, temporal pole, and entorhinal and insular regions), and by higher normalized streamline counts between the medial orbitofrontal cortex and pars triangularis and between the thalamus and hippocampus in the left hemisphere.¹⁴² Another structural connectivity study of patients with classic TN revealed (1) the rich-club reorganization with fewer hubs in the subcortical network but more in the sensorimotor network (SMN) (right inferior parietal lobule and left postcentral gyrus); (2) a broad disruption in their structural network, with reduced global efficiency and decreased nodal efficiency in the non–rich-club regions, along with enhanced connectivity strength in feeders, mostly involving the cortical-subcortical connections; and (3) strengthened structural-functional network coupling.¹³⁷ The study of structural connectomics demonstrated that classic TN was associated with decreased efficiency, coupled with dysfunction in regional information integration involving the thalamus, insula, angular gyrus, posterior parietal gyrus, inferior occipital gyrus, middle temporal gyrus, and middle and superior frontal gyrus, whereas idiopathic patients with TN demonstrated increased efficiency with dysfunction in the hub organization in the posterior cingulate cortex, hippocampus, and superior and inferior temporal gyrus.¹²⁷ Xu et al. investigated structural connectivity of the tracts that showed significant group differences in functional connectivity study and found decreased FA values in the white matter tract connecting left anterior insula and left anterior cingulate cortex (P < 0.0001) but no DTI differences across other tracts (angular gyrus–posterior cingulate cortex, precentral gyrus–postcentral gyrus).¹³²

Table 12.

Key characteristics of the studies for structural connectivity analysis of the brain in patients with trigeminal neuralgia.

Study	Patients with TN	TN type	Age (y), average ± SD (range)	Disease duration (y), average ± SD (range)	Surgical procedure	Study design/timing of MRI acquisition	Analysis	Target structures	HC
Xu (2022)132	48	Classic	54.48 ± 10.35	7.10 ± 5.43	—	Cross-sectional	FC, SC (mean FA)	Tracts interconnecting brain regions that showed significant TNHC group differences in FC study	46
Zhang (2022)137	29	Classic	54.59 ± 10.82	4.12 ± 3.25	17 MVD, 2 RFA, 10 none	Cross-sectional, pre-Tx	SC (connection density), FC	Brainnetome Atlas	34
Wu (2020)127	45 rh (23 wNVC, 22 woNVC)	23 classic, 22 idiopathic	wNVC: 53.30 ± 8.66 woNVC: 47.77 ± 9.24	wNVC: 5.74 ± 3.35 woNVC: 4.97 ± 2.09	23 MVD, 22 IN	Cross-sectional, pre-Tx	VBM, SBM, DTI, SC (graph theoretical analysis)	Whole brain	45
Zhong (2018)142	23 R	Primary	49 ± 13 (23–69)	6 ± 6 (1–30)	—	Cross-sectional	VBM, SBM, DTI, SC (normalized streamline count)	Whole brain	23

DTI, diffusion tensor imaging; FA, fractional anisotropy; FC, functional connectivity; HC, healthy control subjects; IN, internal neurolysis; MVD, microvascular decompression; NVC, neurovascular compression; R, right-sided trigeminal neuralgia; RFA, radiofrequency ablation; rh, right-handed; SBM, surface-based morphometry; SC, structural connectivity; TN, trigeminal neuralgia; Tx, treatment; VBM, voxel-based morphometry; wNVC, with NVC; woNVC, without NVC.

3.6. Analysis of the brain activity and functional connectivity

3.6.1. Brain activity and functions of the node

Task-based fMRI study of 6 TN subjects (4 females, average age: 73.3 ± 10.1 years, age range: 57–81 years) comparing before (2–3 weeks) and after (1–2 weeks) percutaneous radiofrequency rhizotomy reported that reduced preinterventional blood-oxygen-level-dependent (BOLD) activation and post-interventional recovery were observed in the right primary motor cortex, bilateral primary and secondary somatosensory cortices, anterior cingulate cortex, superior and middle temporal gyrus, right amygdala, and right cerebellum for the jaw clenching task.⁷⁴ For the air sensation stimulation, aside from activation in the bilateral secondary somatosensory cortices, pretreatment patients with TN demonstrated extra BOLD activation in the right amygdala, bilateral hippocampus, right thalamus, right insula, brainstem, and bilateral cerebellum, which was subsided after the intervention.⁷⁴ The resting-state fMRI studies for activity and connectivity measures of the whole brain are listed in Table 13, and a summary of the results of those studies are provided in Table 14.^{13,30,36–38,71,72,110,112,120,122,129,134,135,140,143} Whole-brain fMRI analysis of patients with TN demonstrated alterations in spatiotemporal BOLD signal property and degree of functional connectivity in the various brain structures, including the thalamus, basal ganglia, precentral and postcentral gyrus, medial and superior frontal gyrus, precuneus, middle and inferior temporal gyrus, fusiform gyrus, hippocampus, middle occipital gyrus, and cerebellum, highlighting impairments in major brain areas that are part of the “pain matrix,” where volumetric changes were observed in VBM (Table 14). Conventional and artificial intelligence-inspired analyses of percept-related (momentary pain rating) fMRI of 39 patients with TN extracted the similar brain regions including the precentral gyrus, superior frontal gyrus, supramarginal gyrus, superior temporal cortex, fusiform gyrus, insula, dorsal anterior cingulate cortex, thalamus, lateral occipital gyrus, inferior temporal cortex, postcentral gyrus, lingual gyrus, and inferior parietal gyrus.⁶⁹

Table 13.

Key characteristics of the resting-state functional MRI studies of the brain in patients with trigeminal neuralgia.

Study	Patients with TN	TN type	Age (y), average ± SD (range)	Disease duration (y), average ± SD (range)	Study design/timing of MRI acquisition	Analysis	Target structures	HC
Wang Y (2015)122	17 rh	Idiopathic	63.41 ± 7.25	6.98 ± 5.64	Cross-sectional	ReHo	Whole brain	19
Dou (2016)30	31 rh R	Classic	60.5 ± 11.6	6.4 ± 7.2	Longitudinal pre-1 wk, post-PRT 6 mo	ReHo	27 voxels	—
Xiang (2019)129	28 rh	Classic	51.39 ± 9.37	3.73 ± 4.10	Cross-sectional	ReHo	Whole brain	28
Tan (2022)110	27 (13 R, 14 L)	Classic	57.6 ± 7.5	4.95 ± 3.10	Cross-sectional	ReHo, DTI	Hippocampus	27
Yan (2019)134	28 rh	Primary	37.4 ± 9.0	4.5 ± 13.3	Cross-sectional	Dynamic ReHo	Whole brain	28
Yuan (2018)135	23 rh	Idiopathic	56.9 ± 12.5	5.69 ± 3.33	Cross-sectional	ReHo, fALFF	Whole brain	23
Wang Y (2017)120	17 rh	Classic	63.41 ± 7.25	6.98 ± 5.64	Cross-sectional	ALFF	Whole brain	19
Chen (2019)13	28	Classic	51.39 ± 9.37	3.73 ± 4.10	Cross-sectional	ALFF	Whole brain	28
Liu (2022)71	34 rh (20 R, 14 L)	Classic	53.06 ± 10.91	4.63 ± 3.53	Cross-sectional	VBM, zALFF	Whole brain	29
Zhang Y (2019)140	29 R	Classic	48.1 ± 11.9	6.02 ± 4.35	Cross-sectional	ALFF, fALFF, sbFC	Whole brain for ALFF and fALFF, 6 ROIs with significant ALFF/fALFF results for sbFC	34
Ge (2023)38	77 rh	Classic	54.08 ± 10.78	4.96 ± 4.90	Cross-sectional	Static and dynamic ALFF, ReHo, and DC in classic, slow-4, and slow-5 frequency band	Whole brain	73
Ge X (2023)37	43	Classic	55.14 ± 11.59	5.14 ± 5.94	Cross-sectional	Static and dynamic DC (baseline, 5 s, 30 s after triggering)	Whole brain	—
Ge X (2022)36	48	Classic	55.65 ± 11.41	3.0 (1.225–6.375)	Cross-sectional	Static and dynamic ALFF (baseline, 5 s, 30 s after triggering)	Whole brain	—
Zhu (2020)143	28 rh	Primary	51.39 ± 9.37	3.73 ± 4.10	Cross-sectional	DC	Whole brain	28
Liu H (2022)72	34 rh (20 R, 14 L)	Classic	53.06 ± 10.91	4.63 ± 3.53	Cross-sectional	DC, sbFC	Whole brain for DC analysis, ROIs with significant difference in DC for sbFC analysis	29
Tian (2016)112	20	Primary	52.6 (36–65)	1.76 ± 1.35 (0.17–1.33)	Cross-sectional	DKI, FCD	Whole brain	22

ALFF, amplitude of low-frequency fluctuations; DC, degree centrality; DKI, diffusion kurtosis imaging; DTI, diffusion tensor imaging; fALFF, fractional ALFF; FC, functional connectivity; FCD, FC density; HC, healthy control subjects; L, left-sided trigeminal neuralgia; PRT, percutaneous radiofrequency thermocoagulation; R, right-sided trigeminal neuralgia; ReHo, regional homogeneity; rh, right-handed; ROI, region of interest; sbFC, seed-based FC; TN, trigeminal neuralgia; VBM, voxel-based morphometry; zALFF, Fisher z transformation ALFF.

Table 14.

Summary of major brain structures with significant changes in brain activity or connectivity parameters observed in the resting-state functional MRI analysis of the brain in patients with trigeminal neuralgia compared with control subjects.

Study (reference number)	122	129	110	134	135		120	13	71	140	38						37				36				143	72	112	75
Analysis	ReHo	ReHo	ReHo	dReHo	ReHo	fALFF	ALFF	ALFF	zALFF	ALFF/fALFF	sReHo	dReHo	sALFF	dALFF	sDC	dDC	sDC	sDC	dDC	dDC	sALFF	sALFF	dALFF	dALFF	DC	DC	FCD	BEN
																	5 s	30 s	5 s	30 s	5 s	30 s	5 s	30 s
Thalamus	↑ R			↑					↑ L												↓ R	↑ R					↑ B	↑
Striatum																										↑ R
Putamen						↑			↑B	↓ L
NAc
Caudate									↑ B						↑ R		↑ R	↓ R
Pallidum									↑ R		↑ R															↑ R
Cingulate
ACC					↑
MCC							↑ R																			↓ L
PCC										↓											↓ L	↓ L
Insula					↓					↓ L
Frontal lobe
PFC										↓ L							↑ R	↓ R									↓ B
SFG		↑ R			↑								↑B	↑B		↑ R	↓ B	↑ B			↓ B	↓ B
MPFC					↑		↓ L					↑ B
MFG							↑ L										↑ R	↓ R
M1	↑ L	↑ R		↓ L				↑ R
PCL																									↑ L	↓ L
Parietal lobe
S1	↑ L
SPL				↓ L
IPL	↑ R
SMG																											↓ R
AG																											↓ R
Precuneus					↑								↓ R	↓ R													↓ R
Temporal lobe							↑ B			↓ B
STG																					↑ L	↓ L
MTG				↓ L	↑						↑B						↑ R	↓ R			↓ R	↑ R
ITG	↑ R	↑ R					↓ R	↑ L
FG		↑ R						↑ B	↑ B								↑ R	↓ R					↑ R	↑ R
HC			↑ R						↑ B																			↑ L
PHG	↓ R								↑ B
Amygdala	↓ L
Lingual															↓ L				↑ R	↑ R			↑ B	↑ B	↑ R	↓ R
Occipital lobe						↓
SOG								↑ R
MOG		↑ R					↑ L				↓ L										↑ B	↑ B
IOG								↑ L													↑ R	↑ R	↑ R	↑ R
Cuneus																										↓ R
Calcarine											↑ R				↓ R													↑
Brainstem
PAG									↑
Pons									↑
Cerebellum					↓		↑ R		↑ L		↑ B				↑ R
Superior								↑ R
Inferior		↑ B						↑ B																	↑ B			↓ R
Posterior					↑

ACC, anterior cingulate cortex; AG, angular gyrus; ALFF, amplitude of low-frequency fluctuations; B, bilateral; BEN, brain entropy; DC, degree centrality; dALFF, dynamic ALFF; dDC, dynamic DC; dReHo, dynamic ReHo; fALFF, fractional ALFF; FCD, functional connectivity density; FG, fusiform gyrus; HC, hippocampus; IOG, inferior occipital gyrus; IPL, inferior parietal lobule; ITG, inferior temporal gyrus; L, left; MCC, mid-cingulate cortex; MFG, middle frontal gyrus; MOG, middle occipital gyrus; MPFC, medial prefrontal cortex; MTG, middle temporal gyrus; M1, primary motor cortex; NAc, nucleus accumbens; PAG, periaqueductal gray; PCC, posterior cingulate cortex PCL, paracentral lobule; PFC, prefrontal cortex; PHG, parahippocampal gyrus; R, right; ReHo, regional homogeneity; sALFF, static ALFF; sDC, static DC; SFG, superior frontal gyrus; SMG, supramarginal gyrus; SOG, superior occipital gyrus; SPL, superior parietal lobule; sReHo, static ReHo; STG, superior temporal gyrus; S1, primary sensory cortex; TN, trigeminal neuralgia; zALFF, Fisher z transformation ALFF; ↑, significant increase; ↓, significant decrease.

Disease duration was correlated with brain activity or functional connectivity measures in the brain structures like thalamus, prefrontal cortex, amygdala, and inferior occipital gyrus,^{13,20,72,114,115,120,134,139} whereas pain intensity was correlated with those parameters in the insula, cingulate cortex, prefrontal cortex, middle temporal gyrus, inferior occipital gyrus, precuneus, postcentral gyrus, and posterior cerebellar lobe.^{13,117,132,135,140,143}

3.6.2. Functional connectivity studies

The key characteristics and findings of the studies for functional connectivity of the brain were summarized in Table 15. Functional connectivity analysis between selected region of interests (ROIs) revealed increased or decreased functional connectivity in the DMN, salience network (SN), SMN, visual networks, connections between those networks, or connections involving the thalamus and limbic system.^{20,72,114,115,117,123,132,136,137,139,140}

Table 15.

Key characteristics of the functional connectivity studies for resting-state functional MRI analysis of the brain in patients with trigeminal neuralgia.

Study	Patients with TN	TN type	Study design/timing of MRI acquisition	Analysis	Target structures	HC	Key findings vs HC
Zhang Y (2019)140	29 R	Classic	Cross-sectional	ALFF, fALFF, sbFC	6 ROIs with significant ALFF/fALFF results	34	ALFF/fALFF: Reduced ALFF/fALFF in the PCC, left insula, left DLPFC, left putamen and bilateral temporal lobe, exclusively in the frequency of the slow-5 band sbFC: 2 weaker circuits including the PCC-medial PFC and dorsolateral PFC-hippocampus circuits, indicating abnormal interactions with the DMN.
Liu (2022)72	34 rh (20 R, 14 L)	Classic	Cross-sectional	DC, sbFC	ROIs with significant difference DC results	29	Significantly greater DC values in the right pallidum and putamen, and lower DC values in the right lingual gyrus, right calcarine sulcus, left paracentral lobule, and left MCC sbFC: the right lingual gyrus showed decreased FC within the visual network and with other core brain networks, including the SMN, DMN, SN; and the left MCC exhibited decreased FC within the MCC and VN.
Danyluk (2021)20	22 (16 Res, 6 NR)	Primary	Cross-sectional	sbFC	6 pain-relevant ROIs: ACC, PCC, hippocampus, amygdala, thalamus, insula	19	Increased FC between the right insular cortex and bilateral thalami in patients with TN compared with HC.
Tsai (2018)115	62 (36 R, 26 L)	Classic	Cross-sectional	VBM, sbFC	ROIs with significant VBM results	19	R: significantly lower spatial extension of the motor network and lower connections of the right SFG with the right middle frontal gyrus. L: lower spatial extension of the SN and lower connection between the left precentral gyrus with the left SFG and significantly higher connection between the pulvinar of bilateral thalamus.
Tsai (2019)114	25	Classic	Longitudinal pre-1 wk, post-RFA 1 wk	Graph theory, modularity analysis, FC	268 ROIs on Shen Brain Atlas	20	Lower SMN and DMN activities in patients with TN and it increased after surgery. Higher pre-RFA FC between the DMN module and other modules was associated with better treatment response. A lower connection between the DMN and subcortical modules was associated with a better treatment response and the thalamus and MCC were the major connectors within the subcortical module.
Wang Y (2017)117	38	Classic	Cross-sectional	DTI, VBM, sbFC	ROIs with significant VBM results	38	Enhanced FC between the right insula/S2 and ACC, medial PFC, PCC, and bilateral dorsolateral PFC.
Zhang Y (2018)139	29 R	Classic	Longitudinal pre-, post-MVD 4–6 mo	VBM, sbFC	ROIs with significant VBM results	34	Similar resting-state FC patterns between patients with TN and HC. Weaker connectivity strengths between the left amygdala, left thalamus and putamen. A weaker connectivity between the left amygdala and left DLPFC, and enhanced connectivity between the right amygdala and right PFC. NS in the PAG or right insula. TN mainly associated with abnormalities in the fronto-limbic circuits.
Xu (2022)132	48	Classic	Cross-sectional	FC, SC	8 RSN components: DMN, SMN, SN DAN, FPN, VN, LN, CN	46	Reduced FC in DMN (bilateral PCC and right angular gyrus), SMN (bilateral precentral and postcentral gyrus), and SN (bilateral ACC and left anterior insula).
Wang Y (2018)123	20	Classic	Cross-sectional	SBM, VBM, DTI, sbFC	ROIs with significant SBM results	21	Increased insula FC to the left PCC and thalamus.
Zhang P (2021)136	41	Classic	Cross-sectional	Dynamic FC	59 independent components	43	Among 2 distinct states identified, increased expression of the state 2, which was characterized by complicated coupling between DMN and cognitive control network and tight connections within DMN. Patients switched less frequently between states than HCs. Regarding the dynamic topological analysis, disruptions in global graph-theoretical properties (including network efficiency and small-worldness) observed in patients, coupled with decreased variability in nodal efficiency of ACC in the SN and the thalamus and caudate nucleus in the SCN.
Zhang P (2022)137	29	Classic	Cross-sectional	SC, FC	Brainnetome Atlas	34	The rich-club reorganization with less hubs in subcortical network but more in SMN. Broad disruption in SC, more vulnerable than FC. The non-hub regions mainly contributed to the declined global efficiency. SC-FC coupling strengthened as a compensation.

ACC, anterior cingulate cortex; ALFF, amplitude of low-frequency fluctuations; BC, percutaneous balloon compression rhizotomy; CN, cerebellar network; DAN, dorsal attention network; DC, degree centrality; DLPFC, dorsolateral prefrontal cortex; DMN, default-mode network; DTI, diffusion tensor imaging; fALFF, fractional ALFF; FC, functional connectivity; FPN, fronto-parietal network; HC, healthy control subjects; L, left-sided trigeminal neuralgia; LN, language network; MCC, middle cingulate cortex; MVD, microvascular decompression; NR, nonresponders; NS, not significant; PAG, periaqueductal gray; PCC, posterior cingulate cortex; PFC, prefrontal cortex; R, right-sided trigeminal neuralgia; Res, responders; RFA, radiofrequency ablation; rh, right-handed; ROI, region of interest; RSN, resting-state network; sbFC, seed-based FC; SBM, surface-based morphometry; SC, structural connectivity; SCN, subcortical network; SFG, superior frontal gyrus; SMN, sensorimotor network; SN, salience network; S2, secondary somatosensory cortex; TN, trigeminal neuralgia; VBM, voxel-based morphometry; VN, visual network.

4. Discussion

4.1. Histological characteristics of the trigeminal nerve and pathology of the trigeminal neuralgia

Classic TN is characterized by NVC at the REZ. It is considered that the transition zone is particularly vulnerable to mechanical irritation. Fundamentally, peripheral nerves are structurally resistant to mechanical stresses because they are physiologically exposed to those stresses, such as developmental growth, stretches associated with movements of the limbs, and mechanical compressions from daily activities.⁵ Histologically, the peripheral nervous system (PNS) segment of the cranial nerve is characterized by parallel-aligned organization of the axons and regular and rounded myelin sheaths around the axons, and it is rich in connective tissue.²³ Consisted of undulating nerve fibers, it is protected from traction by its elasticity.²³ Moreover, Schwann cells themselves are structurally resistant to mechanical stresses, like tension and compression.⁵ By contrast, oligodendrocytes are highly vulnerable to stress, such as trauma, ischemia, active oxygen, excitotoxicity, cytotoxicity, and autoimmune attack, because of its energy consuming myelination process and the associated high metabolic turnover.⁵⁴ The CNS segment of the cranial nerve is characterized by less pronounced parallel organization of the axons, and irregular and thicker myelin sheaths with connective tissue to a far lesser extent, outlined by astrocyte processes.²³ If demyelination happens, newly differentiated oligodendrocytes derived from an adult oligodendrocyte progenitor cell pool can replace deceased oligodendrocytes, which can reinstate the myelin sheath around demyelinated axons (remyelination), and existing oligodendrocytes also contribute to remyelination by reensheathing nearby axons.⁵⁴ Regenerated myelin would be thinner than the original myelin sheath, and it does not go back to the previous state.⁵⁴ The ability of efficient remyelination declines with age,⁵⁴ and it may be one of the contributing factors associated with higher incidence of TN in the older population. The transition zone, also called the Obersteiner-Redlich zone, is a boundary transitioning from the PNS to the CNS located at the proximal cisternal segment of the trigeminal nerve. Microscopically, the CNS–PNS interfaces of the individual nerves overlap in the rootlets or at the CNS surface, and therefore, a length of rootlet contains both tissue types, forming the transition zone.³³ The length of the transition zone of the trigeminal nerve is typically around 2 to 6 mm (the length of the medial aspect: range 0.1–2.5 mm, the lateral aspect: range 0.17–6.75 mm).^23,98 At the CNS–PNS interface, the thick glia limicans are traversed by individual myelinated fibers and by bundles of small nonmyelinated axons.³³ For myelinated neurons, there is a transitional node, the area of a sharp discontinuity of tissue types, which is covered by astrocyte processes (continuously from the CNS) with the basal lamina (continuously from the PNS) on top to bind them.³³ The study of animal model demonstrated that mechanical compression injury activated various glial cells, including oligodendrocytes, astrocytes, Schwann cells, and microglia/macrophages at the transition zone of the trigeminal REZ, which would trigger various neural and immune mediators and inflammation, leading further sensitizing the neurons and resulting in central sensitization to induce neuropathic pain.⁷⁹ These unique anatomical and neurophysiological features of the trigeminal REZ make it more susceptible to exogenous stimulation or injury⁷⁹; therefore, chronic compression injury would induce irreversible alteration in local microenvironment of the REZ, resulting in malfunction of the trigeminal pathway.

The autopsy reports or the studies of surgical specimens of patients with TN reported changes observed at the site of NVC, usually at the transition zone. The findings include focal demyelination, degeneration of myelin, a few scattered fibers with abnormally thin myelin sheaths, and aberrantly myelinated axon clusters observed in the CNS segment,^29,78,83,101 and demyelination was more extensive in TN with severe NVC.^29,101 The pathogenesis of the demyelination is not well understood; It can be a direct consequence of the pulsatile compression or secondary to focal endoneurial vascular injury and resulting ischemia.⁷⁸ It was also demonstrated that stretching of the plasma membrane leads to entry of calcium ions that cause disruption and loss of microtubules.¹⁷ Axonal changes, such as degeneration of the fibers, axonopathy and axonal loss, were observed occasionally.^29,83,101 Immunohistochemistry analysis revealed signs of oligodendrocyte degeneration and damaged astrocytes sporadically at the site of NVC and a sign of a reactive astrocyte proliferation around the central nerve root.⁸³ In the PNS segment near NVC, PNS myelin was better preserved than the CNS myelin, and most of the myelinated axons in the PNS segment had a normal appearance.^29,83 However, there was some nerve atrophy or hypertrophy, swelling, or dystrophy observed with their myelin sheath varied in shape and thickness, sporadically with larger number of collagen fibers, and those microscopic changes were more profound in patients with TN with long-lasting disease.^29,83 It was hypothesized that segmental demyelination as well as focal and parafocal axonal damage, possibly with retrograde degradation of the axons and sensory neurons in the ganglion, may be responsible for TN pathology.⁸³ Patients with TN and without NVC showed minimal pathological findings.^29,46 Proposed mechanisms of TN pain attacks include dysfunctional neuronal hyperactivity, ectopic impulse discharge, spontaneous and triggered afterdischarge, and cross-excitation among neighboring afferents.^29,78 In addition, it is also believed that some individual susceptibility, cytokine-mediated inflammatory background, and altered proteome in serum and cerebrospinal fluid might predispose to the development of focal demyelination at the REZ.¹⁰⁹ There is accumulating evidence that dysregulation of the VGSCs plays a crucial role in the generation of ectopic activity in trigeminal afferents. Therefore, VGSCs have been considered the main target for pain control in TN.³⁴ Surgical treatment includes MVD and stereotactic radiosurgery (GKRS/CKRS). Microvascular decompression can achieve pain relief with a high initial success rate and can eliminate TN entirely in some cases, 20% to 30% of patients do not show improvement 1 year postsurgery, and recurrence is seen in approximately 30% of patients.¹⁰⁵ The failure of MVD may be attributable to severe local depletion of oligodendrocytes and astrocytes within the zone of chronic compression or irreversible axonal degeneration because it is most common in patients with TN with very long-standing disease. Aberrant remyelination, that is occasionally seen in the compressed nerve root, prevents the separation of groups of apposed axons and, therefore, would also contribute to MVD failure, relapse, and persistent TN.⁷⁷

Surgical observation revealed that the location of the compression is not limited to the REZ; NVC was observed (including multiple sites) in the proximal third segment (around the REZ) in 52.3% of patients with TN, in the mid-third segment in 54.3%, and in the distal third segment (near the exit from Meckel cave) in 9.8%.¹⁰⁴ However, the incidence of TN was higher when NVC was close to the REZ.¹⁰⁸ Direct vascular compression at the trigeminal transition zone is considered to be the common cause of TN, whereas other etiologies, such as mechanical pulsatile stretching of the root, thickened arachnoid adhesions, and smaller posterior fossa (especially in females), may also contribute to damage at the transition zone.¹⁰⁹

4.2. Structural and microstructural analysis of the trigeminal nerve and tract

Simple neurovascular contract is a common finding in the aging population; however, not all the cases develop TN symptoms. Moreover, treatment outcomes do not always correlate with disease duration, although patients with long-lasting TN tend to have unfavorable outcomes. Therefore, in vivo imaging analysis of the affected trigeminal nerve would contribute to the assessment of underlying pathophysiology in the individual cases, leading to proper treatment strategies and prediction of the outcomes.

Smaller nerve volume or cross-sectional area of the affected nerve seen in the structural analysis may relate to nerve atrophy or nerve indentation because of severe NVC. It is assumed that, for the cases with severe NVC, TN symptoms are likely directly caused by NVC and are likely to be relieved if the cause is resolved by MVD. Severe NVC with distortion and indentation of the nerve root predict the probability of long-term pain relief after MVD,¹⁰⁵ and those findings were reflected in prognostic preoperative TN evaluation using the severity of the vascular compression (Grading system⁶ and Scoring system⁹⁵). However, smaller nerve volume at the distal segment (near the exit from the Meckel cave) was associated with poor outcomes (nonresponders).³¹ It may be reflecting chronic axonal degeneration extending to the distal segment of the trigeminal nerve, which was also observed in the microscopic studies. However, it is also noted that other studies reported no significant differences in the trigeminal nerve volume between the responders and nonresponders.^11,21

One of the characteristic findings in the microstructural analysis of the affected trigeminal nerve is increased RD, the magnitude of water diffusion parallel to the tract, which is reflective of demyelination when AD remains the same. It comes with a decline in FA, which indicates a loss of neuronal integrity. Reduced AD, the magnitude of water diffusion perpendicular to the tract, might reflect axonal degeneration, whereas the greater extracellular water content because of the vasogenic edema or the atrophy of the nerve might dilute AD changes within the axons.¹²⁶ MD is the overall diffusion within a voxel, and an increase in MD is typically consistent with increased extracellular water content because of acute conditions (eg, edema and inflammation) or chronic injury.

Damages were seen further into the trigeminal tract within the brainstem (Table 5). In the DTI analysis at the cisternal segment, REZ and pontine segment of the trigeminal tract in classic TN,⁵¹ and lower AD in the cisternal segment on the symptomatic side seen only in the long-term responders may reflect the direct or indirect effects of NVC on the trigeminal nerve axons. Thus, surgical removal of NVC would contribute to the recovery of axonal integrity. However, characteristic findings in nonresponders was demonstrated in the pontine segment (higher AD compared with control subjects) on the symptomatic side⁵¹ or in the spinal trigeminal tract (lower FA and higher RD).¹⁴¹ With the fact that patients with TN without successful MVD demonstrated neuronal alterations reaching the pontine segment of the trigeminal pathway in the presurgical DTI studies, it can be predicted that intervention for the cisternal segment of the trigeminal nerve would not contribute to relieving TN symptoms in those patients. In summary, those presurgical DTI characteristics would help identify the location of the TN pathology and may be useful for the selection of surgical candidates. Interestingly, there were location-specific characteristics in DTI indices along the trigeminal pathway at the cisternal segment, REZ, and pontine segment of the unaffected nerves,⁵¹ which might be reflecting histological features of each segment. Although in the central hypothesis of TN suggested by Trousseau a similarity between TN attack and seizure disorders may indicate hyperactivity in the trigeminal nucleus as TN pathology,^27,86 DTI analysis of the pontine nucleus did not reveal significant findings.⁸⁸

After effective MVD, normalization or recovery trend in DTI metrics were observed (Table 7).^{8,25,59,119,128,138} At the REZ, effective treatment reversed FA, MD, RD, and AD abnormalities, and recovery in DTI indices was correlated with pain relief after treatment.^25,119,128 The previous study suggested that recovery in diffusion parameters at the early postoperative phase (1 week) was mainly because of a decrease in λ‖ and recovery at the chronic phase (4–6 months) was predominantly because of a reduction in λ⊥.¹³⁸ Possible mechanisms for immediate postoperative pain relief derived by release of the nerve root from NVC include cessation of the ectopic impulse discharge, termination of the ephaptic transmission among adjacent fibers, rapid reversal of conduction block, and recovery of the plasma membrane. Recovery from edema might also explain an immediate decrease in diffusivity. Remyelination is a most likely explanation for a reduction in λ⊥ with increased FA at the later phase after MVD. Chai et al. also reported a significant increase in FA ratio (affected side/unaffected side) demonstrated at the acute phase (7 days) after MVD and a greater increase in FA ratio with a significant decrease in ADC ratio at the chronic phase (6–8 months). However, persistent abnormalities (decline in FA) 4 years after the surgery were also reported in the past literature.⁵⁹ The microscopic findings also suggested that the affected nerve would not have the same integrity as the original state. Furthermore, nonresponders demonstrated no significant changes in DTI metrics after MVD surgery,⁸ possibly implying there were preexisting chronic, irreversible myelin and axonal damages in the affected trigeminal nerve.

Diffusion tensor imaging studies of patients with TN after RFA and GKRS/CKRS revealed different outcomes (Table 7). Although pre-interventional DTI values of the trigeminal nerve did not predict treatment response,^10,11,63 changes in DTI metrics after intervention may be useful for predicting long-term outcomes. Chen et al. investigated acute posttreatment changes in nerve volume and DTI metrics of the trigeminal nerve.¹⁰ Two weeks after RFA, the patients without sufficient pain relief showed increased nerve volume, whereas there was no significant change in volume between pre-RFA and post-RFA in responders.¹⁰ In this study, DTI metrics in both responders and nonresponders demonstrated post-RFA changes, including a decrease in RD and ADC to normal level, a decrease in AD (lower than the unaffected side), and an increase in FA 2 weeks after RFA; however, those changes in DTI metrics did not correlate with treatment outcomes.¹⁰ Decreased diffusivity here could be because of nonselective intracellular edema after RFA. Therefore, postinterventional acute (weeks) reactions in DTI metrics of the trigeminal nerve may not be useful for predicting the outcomes of RFA. Moreover, the DTI tractography study revealed that, even if the DTI indices were showing the same results (eg, reduced FA), the successful treatment may depend on the extent of the damaged and preserved fibers within the trigeminal nerve.⁴⁷ However, DTI demonstrated different responses between successful and unsuccessful cases in the later phase (months) after intervention. It was reported that, compared with the preintervention DTI metrics, responders exhibited significant FA reduction and RD increase in the post-GKRS trigeminal nerve at 6 months after intervention, whereas nonresponders did not exhibit significant differences in pre-GKRS and post-GKRS DTI metrics.¹¹³ Moreover, there was an FA reduction at the target zone of GKRS over time during 24 months after intervention in responders, whereas there was no significant differences in DTI metrics over the course of 24 months in nonresponders.⁵⁰ Other studies also support those post-interventional DTI changes.^12,99 Chen et al. reported that preoperatively declined FA at the REZ no longer showed significant decline compared with the contralateral side 1 month after RFA or CKRS (acute reaction), but a significant difference reappeared after 6 months (chronic response). In their study, RD, AD, and MD at the REZ of the treated trigeminal nerve remained significantly higher compared with the contralateral side until 6 months after treatment, but there was no significant difference in diffusivity a year later.¹² There was nerve atrophy over time (months-years) after GKRS, and smaller post-GKRS cross-sectional area and lower T1 signals of the treated trigeminal nerve were correlated with pain relief after treatment.¹⁰⁷ Those post-GKRS microstructural changes were more apparent in patients with TN with shorter (less than 5 years) disease duration, who demonstrated higher success rates, compared with patients with long-lasting TN (more than 5 years).⁶¹ Because the goal of RFA and GKRS/CKRS is to damage the nerve to stop the transmission of pain signals, the pretreatment nerve damage might be less important in predicting the outcomes. In fact, the study of radiofrequency-induced rat sciatic nerve injury reported the nonselective structural damage to the axons and collagen and inflammatory response.^10,15 It is expected to see the signs of nerve degeneration (decreased FA) and nerve atrophy over time after intervention in patients with TN with successful outcome. However, a lack of those responses in DTI metrics would indicate unsuccessful intervention. In addition, in patients with long-lasting TN, who demonstrate lower success rates, the trigeminal nerves with chronic damages may not respond well to those treatments. Although the histopathological changes of the trigeminal nerve after those interventions are still debated, changes in DTI metrics after these interventions might suggest microscopic changes, such as inflammation (increased diffusivity), neuronal degeneration (lower FA), fibrosis (decreased diffusivity), and atrophy, depending on the phase after the intervention. Those postinterventional changes in DTI metrics could be useful as a biomarker for predicting long-term pain relief.

Interestingly, not consistently but some studies demonstrated lower FA or significantly higher diffusivity (AD, RD, and MD) on the nonsymptomatic side in patients with TN,^26,80 which might suggest individual susceptibility.

4.3. White matter integrity of the whole brain

Diffusion tensor imaging indices of the cerebral white matter have been studied by exploratory analysis of the whole brain or hypothesis-driven analysis of the selected tracts and demonstrated that TN involves microstructural changes in various areas of the brain (Table 9). Observed changes in DTI metrics of the white matter tracts included lower FA or increased diffusivity, mainly RD and MD, with increased AD and decreased kurtosis (a broadening of the peak) in some of the studies. Possible mechanisms for the neural plasticity include alterations in axonal organization (such as packing density, axonal sprouting/branching, axon size/number, crossing fibers, and myelin remodeling) and change in other factors such as astrocyte morphology and angiogenesis.²⁶ An increase in RD and MD might be also associated with demyelination, inflammation, or edema. Diffusion tensor imaging alterations in the trigeminal somatosensory and modulatory pain pathways are suggestive of CNS plasticity as a result of hyperactivity or inhibitory reaction. Alterations in those pathways were also reported in the studies of other chronic pain disorders, such as migraine.²²

The sensory information of the trigeminal nerve conducted through the trigeminal nuclei, which descends from the level of the lower pons down to the spinal cord (C2-C4 level), travels to the thalamus and ultimately projects to the somatosensory cortex.^106,111 The second-order neurons arise from the ipsilateral caudal spinal nucleus, mostly cross the midline, and project to the thalamus (80% contralaterally and 20% ipsilaterally) through the trigeminothalamic tract.^85,106,111 Diffusion tensor imaging was applied to extract the trigeminal pathway within the brainstem and it showed abnormalities in merged tractography.⁹ However, as identification of the tracts within the brainstem is challenging, they have not been fully explored.

The thalamus sends third-order neuronal projections to the primary and secondary somatosensory cortex (from ventroposterior medial thalamus through the posterior limb of the internal capsule and corona radiata) and to the cingulate and insula (from mediodorsal ventral caudal thalamus through the anterior limb of the internal capsule and anterior corona radiata).⁸⁵ Those cortical regions are responsible for interpreting sensory information in terms of location, intensity, and duration.³⁴ Projections from the somatosensory cortices to the sensory/motor nuclei of the trigeminal nerve in the brainstem (mostly from the contralateral side) also have inhibitory function and can regulate facial pain perception.^34,106 Outputs from the thalamus can be directed to the limbic areas and other cortical and subcortical structures that are responsible for processing the cognitive, affective, and emotional components of the pain.^34,106,111 In DTI studies, the changes in DTI metrics were observed particularly in the superior longitudinal fasciculus, corona radiata, internal capsule, thalamic radiation, external capsule, corpus callosum, cingulum, uncinate fasciculus, and fornix, on the ipsilateral, contralateral to TN, or both sides. Those white matter tracts are associated with the projection and inhibitory pathway for nociceptive information and the processing circuit for pain perception. The cingulum is the complex major white matter tract interconnecting frontal, parietal, and medial temporal areas, as well as linking subcortical nuclei to the cingulate cortex.⁷ Imaging studies implicate the role of the cingulum bundle in executive control, emotion, pain (dorsal cingulum), and episodic memory (parahippocampal cingulum).⁷ The uncinate fasciculus (connecting the temporal lobe with the medial orbitofrontal cortex) and fornix (major output tract of the hippocampus) are also key regions controlling memory and executive functions. In addition, some DTI studies of TN reported abnormalities in the medial forebrain bundle, brainstem, and cerebellar tracts. There are numerous intranuclear and internuclear connections within the nuclei of the trigeminal nerve and between the trigeminal nuclei and other brainstem nuclei.^106,111 The activation of mesencephalic and bulbar structures, such as the periaqueductal gray matter and the rostral ventromedial medulla, can modulate nociceptive processing and work as a main inhibitory descending pathway.³⁴ The medial forebrain bundle, the tract connecting the anterior nuclei of the hypothalamus with the mesencephalon, is probably associated with nociceptive hypothalamic response. The trigeminal nuclei also project to the hypothalamus,¹¹¹ which will lead to neuroendocrine corticotropin response. The roles of the cerebellum in pain processing remain poorly understood; however, it is known that the cerebellum is connected to the various areas of the cerebrum, including motor, somatosensory, visuo-spatial, and cognitive cortices, through pontine nuclei and the inferior olive,⁸⁹ and that the trigeminal sensory nuclei are directly connected to various posterior fossa structures (superior colliculus, cerebellar cortex, and deep nuclei) and using the tegmental interneurons.¹¹¹ Studies suggested that the cerebellum plays an important role not only in motor response to pain but also in sensorimotor nociceptive processing, affective processing, and pain modulation.⁸⁹

Disease duration and intensity were associated with the DTI changes mainly in the corticotrigeminal pathways on the contralateral side. It may imply modification of the inhibitory pathway for nociceptive information in patients with TN with long-suffering and intense pain. The fornix, tracts involved in memory, were also affected in the patients with long-term TN.

4.4. Structural changes of the brain

Structural analysis of the brain using statistical parametric mapping techniques, such as VBM and SBM, has played an increasingly greater role in brain research in recent years. Voxel-based morphometry allows voxel-wise comparison of the local concentration of gray matter between groups of subjects, whereas SBM is a brain morphometric technique to construct and analyze surfaces that represent structural boundaries of the brain and provides the indices of cortical morphology, including surface area, cortical thickness, cortical volume, sulcal depth, and gyrification index.³⁵ Those statistical parametric mapping can be exploratory whole-brain analysis or ROI based.

Multiple volumetric studies of TN reported alteration of the thalamic volume (Table 11). The thalamus is a relaying station of the corticotrigeminal pathway, but it also plays a central role in processing nociceptive information before transmitting it to various parts of the cortex. The animal studies have shown microscopic neuronal changes, including neurogenesis, cell loss, gene expression, and structural remodeling/reorganization of synapses, cells and circuits, and non-neuronal changes, such as the biochemistry and thalamic blood flow, responding to neuropathic pains.^1,57 In the thalamic volumetry, the results were often contradictory. It is hypothesized that an increase in volume is because of adaptive changes attributable to constant activity and may be associated with higher pain thresholds, whereas a decrease in volume is assumed to be associated with generalized compensatory deactivation of the thalamic-somatosensory system or maladaptation with low pain thresholds, which was also observed in the patients with chronic pain caused by disorders other than TN.^43,57 The extent of the TN pathology, disease duration, and pain intensity might also affect the outcomes of the structural alterations.

Furthermore, higher magnetic field MRI, such as 7- or 3-T MRI, enables us to explore the volumetry of subsegments of the small structures. The trigeminothalamic tract projects to the various parts of the thalamic subnuclei, including the ventroposterior medial, mediodorsal ventral caudal, intralaminar, and ventroposterior inferior and ventromedial posterior thalamus.⁸⁵ The groups of thalamic subnuclei are involved in different components of processing pain; for example, the ventrobasal complex for the sensory discriminative component of the pain and the intralaminar nuclei for the affective–motivational component of the pain.¹ Using higher magnetic field MRI, VBM studies of patients with TN have detected volume alterations in those subsegments of the thalamus.

In addition to the thalamus, there were changes in volumes in the basal ganglia (caudate, putamen, nucleus accumbens, and palladium) as well. The collective evidence suggests critical roles of the basal ganglia in nociception, including pain processing/modulation, cognitive/affective aspect of nociception, and sensory gating of nociceptive information to higher motor areas, which may affect responses in acute and chronic pain.¹⁶ It is noted that the basal ganglia is rich in various neuroactive chemicals that might be involved in the modulation of nociceptive information.¹⁶ Structural changes were also seen in other brain regions associated with pain processing/modulation and emotional networks, including the prefrontal cortex, insula, anterior cingulate cortex, hippocampus, primary/secondary somatosensory cortex, precuneus, fusiform gyrus, and cerebellum (Table 11).

Disease duration was correlated with morphological changes of the gray matter structures. The effects of TN were seen in the cerebellum in the early phase, and then the thalamic somatosensory system was affected later. As discussed in the previous section, the cerebellum plays a key role in perception and processing of the nociceptive information. The somatosensory system would be affected both by hyperactivity and compensatory/inhibitory response as seen in the DTI studies. As disease intensifies and becomes longer, changes were seen in the areas associated with affection, cognition, visuospatial processing, and modulation of pain, and it may be related to the negative emotional response, such as anxiety and depression, seen in patients with long-lasting TN. Those structures were also affected in patients with other disorders causing chronic pain, such as migraine, fibromyalgia, and neuropathic pain.^24,53,57,94 Morphological gray matter changes in the DMN, the brain areas affected by severe, long-term TN, may be the key for the treatment response.⁵²

Those changes in gray matter morphology can be reversible. Prior studies reported normalization in the thickness of the anterior ventral insular cortex after effective treatment²⁵ or the hippocampal volume after surgery⁹⁰ in TN, or in the thalamic volume after arthroplasty for osteoarthritis of the hip,⁴³ whereas other studies reported no significant changes in gray matter volume in TN when compared before and after treatment.^12,139 Further studies are needed to confirm this phenomenon.

Voxel-based/surface-based morphometry enables subjective quantification of the structural changes of the brain; however, it is still unknown how microscopic and functional changes observed in animal models are reflected in imaging-detectable level observations of volume increase or reduction in human brains, or how volumetric changes in those specific brain structures actually play a role in pathophysiology of TN.

4.5. Alterations in the structural connectivity of the brain

Structural connectivity analysis is the diffusion MRI technique that estimates the physical white matter interconnection between brain regions by measuring the direction and magnitude of water diffusion, using DTI data. Measures of structural connectivity include mean FA, MD, streamline count, and edge weight.⁴⁹

The studies of structural connectivity in patients with TN revealed decreased global efficiency and reorganized connectivity in the brain areas affiliated to sensory, affective, and cognitive dimensions of pain, including the SMN, subcortical network, insula, anterior cingulate cortex, posterior cingulate cortex, hippocampus, prefrontal cortex, precuneus, and inferior and superior parietal lobules.^{127,132,137,142} Structural connectivity strength, that is, white matter fiber density, is thought to be reflective of synaptic pruning in the cortex, myelination/demyelination of nerve fibers, and the addition/reduction of neurons¹⁴²; therefore, those results also support central plasticity in TN.

4.6. Alterations in the brain activity and functional connectivity

Although task-based fMRI can directly visualize pathophysiology of TN, resting-state fMRI may demonstrate accumulated pathological alterations that affect normal brain activity.⁹² Resting-state fMRI has been increasingly used for brain mapping to evaluate regional interactions under task-negative states using BOLD blood flow signals. Exploratory whole-brain analysis investigates local spontaneous activity using techniques such as amplitude of low-frequency fluctuation, fractional amplitude of low-frequency fluctuation, and regional homogeneity. Amplitude of low-frequency fluctuation measures low-frequency oscillations of the BOLD signals, characterizing local spontaneous activity during the resting state, whereas regional homogeneity reflects regional synchronization of the BOLD signals. Functional MRI is also useful for assessing the functional networks of the brain. Functional connectivity is defined by the statistical dependencies between neural signals acquired from different brain areas. Degree of centrality is a measure of local network connectivity represented as the number of edges connected to a node, reflecting the importance of the node. Functional connectivity density quantifies the number of global and local functional connections for each voxel of the brain as long-range and short-range functional connectivity densities, which reflect interregional and intraregional functional interactions, respectively. Seed-based functional connectivity analysis requires the selection of the ROIs, usually from the nodes that showed significant changes in the whole-brain analysis (eg, VBM, SBM, and focal BOLD signal activity) or in previous studies. It measures correlations between the seed and the voxels of the whole brain (seed-to-voxel) or between the selected seeds (seed-to-seed).

Task-based fMRI study that investigated differences in preinterventional and postinterventional brain functions in patients with TN demonstrated subsided activity in the sensory-motor system and its postinterventional recovery for the jaw clenching task.⁷⁴ As for the air sensation stimulation, there was extra activity in various brain areas, such as the brainstem, cerebellum, thalamus, insula, hippocampus, and amygdala, which was deactivated after the intervention.⁷⁴ Multiple resting-state fMRI studies also demonstrated increased activity in the thalamus, basal ganglia, cerebellum, hippocampus, and fusiform gyrus. Increased brain entropy in the thalamus and brainstem may indicate disorder or an exploratory attempt for adaptation against nociception.⁷⁵ In addition to involvement of structures in the corticotrigeminal pathways, there were alterations in the functional activity in the brain areas related to cognitive and memory functions. In many fMRI studies of TN, there was hyperactivity in the cerebellum, which was also observed in most of fMRI studies of acute and chronic pain.⁸⁹ With increased degree of centrality, indicating the increased importance of the node, those results have proposed that the cerebellum is one of the key hubs in functional networks for TN pathology. Compared with structural analysis, characteristic findings in fMRI studies were hyperactivity and lower-degree centrality in the occipital lobe and calcarine area. With increased activity in the middle and inferior temporal gyrus and fusiform gyrus, brain areas associated with visual perception, attention and recognition may be malfunctioning in patients with TN. It has been proposed that pain affects visual perception. For example, in patients with chronic lower back pain, the judgment of sensorimotor aspects of a visually displayed movement was specifically affected.⁸¹ Patients with TN may become more sensitive to the recognition of the actions that possibly cause TN pain attacks.

Functional connectivity analysis identified disruption of the major brain networks, such as DMN, SN, SMN, and visual networks, and the connections involving the thalamus and limbic system (Table 15). The sensorimotor network, which primarily includes somatosensory (postcentral gyrus) and motor (precentral gyrus) regions and extends to the supplementary motor areas, is responsible for motor controls. Alterations in the SMN would be compatible with the results of the microstructural and structural analysis mentioned in the previous sections. Core areas of the DMN include the medial posterior cortex (specifically the posterior cingulate cortex and parts of the precuneus), medial prefrontal cortex, as well as bilateral inferior parietal lobules expanding to posterior temporal areas around the temporo-parietal junction. The DMN represents areas that are active during periods of self-directed thought or introspection, and plays a critical role in attention, memory, prospection and self-processing.⁴ Changes in functional connectivity of the DMN have been reported in acute and chronic pain and may be associated with alterations in attentional processes under the pain.⁴ The SN, also known as the ventral attention network, primarily consists of the anterior insula and dorsal anterior cingulate cortex and functions as a dynamic switch between concentration on self and the inner world, deciding which information is deserving of our attention. Dysfunction of the SN is associated with a broad spectrum of deficits and maladaptive behavioral patterns in a variety of clinical disorders, such as chronic pain, anxiety disorders, insomnia, post-traumatic stress disorder, schizophrenia, and autism.¹⁰² In TN, depression and anxiety ratings were correlated with functional connectivity involving the SN.^117,139 With heterogeneity in study designs and selected ROIs, the results were not always consistent; however, these functional connectivity analyses have suggested that TN is associated with dysfunction of brain networks responsible for pain perception/modulation and cognitive, affective, and emotional components of the pain.

Although it is intriguing to see how diversity in analysis methods would enable us to investigate brain networks from various angles, it still requires extensive work to unveil the whole picture of complex mechanisms of brain network alterations in patients with TN.

4.7. Limitations

Limitations related to the method of this systematic review include that we used a single database (PubMed) and conducted searches with limited keywords. Therefore, there is a possibility that we might have missed some of the articles. It is also noted that the articles were assessed by a single reviewer. For study-related limitations, there could be significant diversity in patient demographics (age range, disease duration, disease severity, right-sided vs left-sided TN, right-handed subjects only vs handedness not specified, etc.) and in treatment outcomes among the institutions. Each study has a relatively limited number of subjects, and there may be subject duplications in the studies from the same institution. Although studies of the trigeminal nerve specifically investigated patients with classic TN to assess the damages caused by NVC, studies of the whole brain did not always specify or differentiate classic and idiopathic TN. It should be noted that patients with secondary TN were excluded in all studies. Because TN is a rare disorder and a spectrum of diseases with a variety of characteristics (etiologies, disease severity, duration, and individual vulnerability), conducting a large-scale study is challenging, and the golden standard of the processing pipeline for quantitative MRI analysis has not been established. Therefore, there were a variety of analysis methods for each category. Differences in methodological approaches, statistical analysis, and thresholds might affect the detection of the brain regions with statistical significance. Finally, the most notable limitation is that, for structural, microstructural, and connectivity analysis of the brain, the results have not been confirmed pathologically.

5. Summary

Quantitative MRI in TN demonstrated structural and microstructural abnormalities of the trigeminal nerve compatible with microscopic findings, and with accumulating quantitative MRI evidence, the pathophysiology of NVC and posttreatment changes of the nerve have become more comprehensible. It has been shown that it could be a tool to assess the extent of trigeminal nerve injury and predict treatment outcomes. Moreover, multispectral quantitative MRI analysis has been showing the evidence of dynamic microstructural, structural, and functional neuronal plasticity of the brain at various levels in the nociceptive pathway in the earlier stage of TN and in the brain regions processing cognitive, affective, and emotional aspects of the pain in patients with more advanced, long-lasting TN, possibly as a stronger modulator especially for chronic neuropathic pain. Further studies are needed to understand these complex mechanisms of neuronal plasticity and to achieve a consensus on the clinical use of quantitative MRI in TN. High-field MRI, such as 7T, has an advantage in analyzing small structures, including the subnuclei of the thalamus and hippocampus, and in investigating the subtle microstructural changes in TN. It would contribute to the understanding of complex TN pathophysiology. Novel techniques such as machine learning enable us to conduct data-driven analysis and create multivariate models for predictive biomarkers, and it would contribute to identifying etiologies and finding suitable solutions for each TN patient. Those advanced technologies might compensate for some of the technical difficulties, such as limited patient enrollment and lack of pathological confirmation, and would hopefully promote further understanding of this disorder to achieve more favorable outcomes in the future.

Supplementary Material

Supplemental digital content

Supplemental digital content associated with this article can be found online at http://links.lww.com/PAIN/C111.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.painjournalonline.com).