The diagnostic value of quantitative assessment of MR neurography in chronic inflammatory demyelinating polyradiculoneuropathy: a systematic review and meta-analysis

Abstract

Objectives:

The purpose of this study was to evaluate the diagnostic value of quantitative magnetic resonance neurography (MRN) in chronic inflammatory demyelinating polyradiculoneuropathy (CIDP). We also compared various MRN parameters and determined the best performing one.

Methods:

Through literature searches in PubMed, Embase, Cochrane, Ovid MEDLINE and ClinicalTtrials.gov until March 1, 2023, we selected studies with the diagnostic performance of MRN in CIDP patients. The pooled estimated sensitivity and specificity of quantitative MRN parameters were determined by a bivariate random-effects model. Subgroup analysis was performed to evaluate the proper quantitative parameters and nerve sites.

Results:

A total of 14 quantitative MRN studies with 23 results gave a pooled sensitivity of 0.73 (95% CI 0.66–0.79) and a pooled specificity of 0.89 (95% CI 0.84–0.92). The area under the curve (AUC) was 0.89 (95%CI 0.86–0.92). Subgroup analysis of quantitative parameters showed the fractional anisotropy (FA) with the highest sensitivity of 0.85 (95% CI 0.77–0.90) and cross-sectional area (CSA) with the highest specificity of 0.95 (95% CI 0.85–0.99). The pooled correlation coefficient for interobserver agreements was 0.90 (95%CI 0.82–0.95).

Conclusion:

Quantitative MRN has considerable diagnostic value in CIDP patients with accuracy and reliability. FA and CSA can be promising parameters in the future diagnosis of CIDP patients.

Advances in knowledge:

This is the first meta-analysis of quantitative MRN in the diagnosis of CIDP.We have selected reliable parameters with cut-off value and provided new insights for subsequent diagnosis of CIDP.

Introduction

Chronic inflammatory demyelinating polyradiculoneuropathy (CIDP) is an immune-mediated acquired polyneuropathy. It typically presents with proximal weakness in the upper and lower extremities, along with sensory abnormalities. ^1,2 Accurate diagnosis and precise treatment can avoid irreversible disability. However, the current diagnosis of CIDP is challenging because there are no specific diagnostic biomarkers for both typical and variant types. Neuropathies from other causes are frequently misdiagnosed as CIDP. Final diagnosis often requires a combination of clinical manifestations, electrophysiology, magnetic resonance neurography (MRN), ultrasound, and nerve biopsy.

The European Federation of Neurological Societies/Peripheral Nerve Society (EFNS/PNS) criteria included the feature that MRI showed gadolinium enhancement and/or hypertrophy of the cauda equina, or the brachial or lumbosacral plexuses as supportive criteria. ³ Previous studies have also described hyperintensity among CIDP patients, in addition to nerve enhancement and hypertrophy of the brachial or lumbosacral plexuses. ^4,5 However, the intraclass correlation coefficients (ICCs) are relatively low in the judgment of hyperintensity and hypertrophy. ^6,7 Different radiologists may provide varying qualitative results for the same image, due to diagnostic subjectivity. Therefore, several quantitative parameters based on multisequence have been applied to the diagnosis of CIDP recently and showed preferable diagnostic value with higher consistency.

Diameter, cross-sectional area (CSA) and volume are the direct indicators of nerve hypertrophy. Compared with qualitative analysis of hyperintensity on a traditional sequence like short tau inversion recovery (STIR), contrast ratio (CR), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) on new sequences are quantitative measurements of hyperintensity. ^8–11 Recently, however, there have been novel parameters in the assessments of nerves for CIDP patients, such as the intramuscular fat fraction analysis, ¹² T2 relaxation time on T2 mapping, ¹³ the contrast-enhancement ratio (CER), quantification of contrast enhancement ¹⁴ and fractional anisotropy (FA). ¹⁵ Based on diffusion tensor imaging (DTI), FA is a quantitative way to study the microstructural integrity of the nervous tissue. ¹⁶ Significantly, the FA value is positively correlated with myelin integrity, fiber denseness and parallelism, which can be useful in peripheral nerve examinations. ^17,18

For the moment, the accuracy and reliability of quantitative MRN in the diagnosis of CIDP remain unclear. Thepurpose of this systematic review and meta-analysis is to assess the diagnostic performance of various quantitative MRN parameters and identify the proper parameters for further clinical practice.

Methods

This systematic review and meta-analysis was formulated according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Supplementary Material 1). ¹⁹ The protocol was registered with INPLASY, the international Platform of Registered Systematic Review and Meta-analysis (INPLASY202350088).

Search strategy and study sources

Two investigators independently performed electronic English-language literature searches in PubMed, Ovid MEDLINE, Embase, Cochrane and ClinicalTrials.gov databases from September 1988 to March 1, 2023. Also, the reference lists of included studies were investigated for other related studies.Information pertaining to the literature search approach is available in the Supplementary Methods section.The keywords were as follows (1) “CIDP” OR “chronic inflammatory demyelinating polyneuropathy”; (2) “MRI” OR “magnetic resonance imaging”; (3) “diagnosis” OR “evaluation”.

Inclusion and exclusion criteria

Firstly, we screened qualitative studies that gave the rate range of CIDP patients with the characteristics of hypertrophy, hyperintensity, or contrast-enhancement. Regarding the diagnostic segment, we specifically selected cohort and cross-sectional studies that contained a CIDP group and a healthy control group to measure the diagnostic accuracy of the quantitative MRN. Almost all the CIDP patients from the includedstudies met the 2010 EFNS/PNS criteria, with one exception that met the EFNS/PNS criteria in 2005. ²⁰

Data extraction and quality assessment

Data were extracted into a designed format containing (1) author, country and year of publication; (2) study design; (3) study population and participants; (4) MRI sequences and selected nerve sites; (5) cut-off value; (6) the number of CIDP patients with abnormal MRI and total CIDP patients; (7) true positive (TP), true negative (TN), false positive (FP), and false negative (FN) values.

We extracted the parameters containing diameter, CSA, volume, FA, SNR, CNR, CR, and CER as results of the included studies. The results with other parameters, namely, nT2w (T2w nerve/T2w muscle), T2 relaxation time and apparent diffusion coefficient (ADC) from four studies were excluded for their low sensitivity and unreliability. ^10,11,14,21 The nerve site that yielded the largest area under thecurve (AUC) was recorded if the same parameter was used in several nerve sites (e.g., brachial plexus or lumbosacral plexus). The quality of the studies included was evaluated using a modified 15-item Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2). ²² Each potentially eligible study was independently assessed by two reviewers. Four domains and 14 questions were assessed to evaluate risk bias from the diagnostic validity.

Statistical analyses

Sensitivity and specificity for each MRN parameter were calculated using two-by-two tables. The study results were visually presented through a forest plot, displaying estimates of sensitivities and specificities for each parameter. We used a bivariate random-effects model to analyze pairs of sensitivity and specificity, incorporating the correlation between them. Measures of the diagnostic accuracy, including sensitivity, specificity, positive likelihood ratio (LR+), negative likelihood ratio (LR-) and diagnostic odds ratios (DORs), were reported as point estimates with 95% confidence interval (CI). Study statistical heterogeneity was assessed using the Higgins inconsistency index(I ² )and the Cochrane Q tests based on the random-effects analysis. ²³ p < 0.05 or I²>50% generally indicated significant heterogeneity. The threshold effect was evaluated by visual assessment of the summary receiver operating characteristic (SROC) plot and Spearman correlation effect r ≥ 0.6 generally indicated a considerable threshold effect. ²⁴ To assess publication bias, we utilized the effective sample size funnel plot and conducted a regression test of asymmetry. ²⁵ Subgroup analysis was performed into four groups: diameter, CSA, signal-related parameters (CR, CNR, SNR), and FA. We conducted a sensitivity analysis using the leave-one-out method, wherein one study was excluded at a time. We applied the random-effects model on the basis of Fisher’s Z transformation of correlations to compute the pooled correlation coefficients with 95% CIs. The values of ICC between 0.75 and 0.9 are interpreted as good reliability, and above 0.9 are indicative of excellent reliability. ²⁶ The statistical analyses were conducted using STATA 16.0 and R v4.2.3 software programs.

Results

Study selection

A total of 961 articles were comprehensively retrieved: 356 from PubMed and 349 from Embase, 250 from Ovid MEDLINE, six from ClinicalTrials.gov. Among these, 213 records with duplicated titles were excluded. Additionally, 453 studies were excluded after screening the title and abstract, and 44 studies remained. Sixteen studies were excluded due to difficulties in extracting data of sensitivity and specificity. Two studies were discarded due to lack of healthy controls. ^13,27 We excluded 12 studies with only qualitative evaluation. Finally, 14 quantitative MRN studies with 23 results were eligible for inclusion (Figure 1). ^{8–11,14,15,20,21,28–33} Three studies provided both quantitative and qualitative results and were included twice. ^14,30,31 We also calculated the rate range of abnormal patterns from total 15 studies with qualitative results. ^{4–6,13,14,30,31,34–41}

Figure 1.

Flow of information through the different phases of the meta-analysis.

Study description

A total of 462 participants from 14 studies were enrolled in the quantitative analysis. Fourteen studies, yielding 23 results, were evaluated and the characteristics of these results are presented in Table 1. Ten studies were prospective and four studies were retrospective. For MRI sequences, four studies used three-dimensional nerve-sheath signal increased with inked rest-tissue rapid acquisition of relaxation imaging (3D SHINKEI), two studies usedthree-dimensional sampling perfection with application optimized contrast using different flip angle evolution (3D SPACE) sequences on the coronal plane, and five studies applied DTI. For the selection of parameters in quantitative analysis, there were seven results with diameter, four with CSA, five with FA, two with CR and SNR respectively.The parameters, including CER, CNR, and volume, each yielded only one result.There was no threshold effect as indicated by the Spearman correlation coefficient of −0.546, between sensitivity and the false-positive rate. Deek’s funnel plot asymmetry tests were presented to assess the publication bias (Supplementary Figure 1). It showed no significant risk of publication bias (p = 0.16). Supplementary Table 2 showed the characteristics of studies with qualitative results.

Table 1.

Characteristics of the total 14 quantitative studies

Included studies	Country	No.(male)	Age	MRI sequence	Study design	Nerve site	Specific location		Parameter	Cut-off value
Su et al. 2021a 14	China	72(50)	50.5 (19–71)	MIP 3D SPACE	P	L4-S1 roots	1.0 cm away from the DRG		Diameter	Unknown
Su et al. 2020a 31	China	52(40)	45.8 (18–67)	MIP 3D SPACE	P	L5 nerve root	1.0 cm away from the DRG		Diameter	≥7.0 mm
Shah etal. 2020a 30	UK	20(14)	51(19-69)	MPRAGE	P	L5 nerve root	Right intraforaminal segment		Diameter	≥5.8  mm
Hiwatashi et al. 2019 11	Japan	18(15)	(11-80)	3D SHINKEI	R	C5-T1 roots			Diameter	≥4.6 mm
Hiwatashi et al. 2018 10	Japan	15(11)	(11-68)	3D SHINKEI	R	L3-S1 roots			Diameter	≥5.0 mm
Hiwatashi et al. 2017a 9	Japan	36(15)	(14-81)	3D SHINKEI	R	T12-L5 ganglions			Diameter	Unknown
Hiwatashi et al. 2017b 8	Japan	23(16)	(19-72)	3D SHINKEI	R	C5-T1 roots			Diameter	≥3.99 mm
Wu et al. 2022a 32	China	40(27)	38.3	3D T2WI FSE	P	L3 to S1 roots	Middle of the corresponding intervertebral disc		CSA	≥28.15 mm2
Wu et al. 2020 29	China	42(26)	51.2 (24–68)	3D MERGE	P	L3 to S1 roots	Middle of the corresponding intervertebral disc		CSA	≥19.2 mm2
Shah et al. 2020 b 30	UK	20(14)	51(19-69)	3D PD-weighted	P	Sciatic nerve	Mid-thigh level		CSA	≥35 mm2
Sinclair et al. 2010 20	UK	19	52.9	STIR	P	Sciatic nerve	Maximum area		CSA	≥64.4 mm2
Wu et al. 2022a 32	China	40(27)	38.3	DTI	P	L4-S1 roots	Average at intraspinal, intraforaminal, and extraforaminal levels		FA	≤0.278
Wu et al. 2022b 15	China	36(20)	40.1	DTI	P	L4-S1 roots	Average at intraspinal, intraforaminal, and extraforaminal levels		FA	≤0.27
Su et al. 2021b 14	China	70(50)	50.5 (19–71）	DTI	P	L4-S1 roots	1.0 cm away from the DRG		FA	≤0.361
Kronlage et al. 2017 13	Germany	36(14)	56.9	DTI	P	Sciatic +Tibial+ Ulnar +Median+Radial nerve			FA	≤0.47
Markvardsen et al. 2016 21	Denmark	28(20)	59.5 (40–69)	DTI	P	Sciatic nerve	20%of the distance from the patella to the anterior superior iliac spine		FA	≤0.446
Hiwatashi et al. 2017a 9	Japan	36(17)	(14-81)	3D SHINKEI	R	T12-L5 roots			CR	Unknown
Hiwatashi et al. 2017b 8	Japan	23(16)	(19-72)	3D SHINKEI	R	C5-T1 roots			CR	≥0.66
Hiwatashi et al. 2017a 9	Japan	36(15)	(14-81)	3D SHINKEI	R	T12-L5 ganglions			SNR	Unknown
Hiwatashi et al. 2017b 8	Japan	23(16)	(19-72)	3D SHINKEI	R	C5-T1 roots			SNR	≥7.81
Hiwatashi et al. 2017a 9	Japan	36(16)	(14-81)	3D SHINKEI	R	T12-L5 roots			CNR	Unknown
Ishikawa et al. 2016 28	Japan	25(16)	61.9 (38–84)	DWIBS	P	Brachial plexus	From the lower rim of the vocal cords through the lung apices		Volume	≥11.6 cm3/m2
Su et al. 2021b 13	China	72(50)	50.5 (19–71）	VIBE	P	L4-S1 roots	1.0 cm away from the DRG		CER	Unknown

^aHiwatashi et al.2017# is for reference 9; Wu et al.2022# is for reference 32.

^bHiwatashi et al.2017 * is for reference 8;Wu et al.2022* is for reference 15;

Methodological quality assessment

According to the Quality Assessment of Diagnostic Accuracy Studies 2(QUADAS-2) tool, the risk of bias and applicability concerns were shown in Figure 2. All included studies showed a low risk of bias in at least four of the whole seven domains. Most studies were at high risk of index test because the cut-off value was not set in advance, but rather determined through ROC analysis.

Figure 2.

Summary of risk of bias and applicability concerns according to the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2).

Heterogeneity test of individual studies

We conducted a meta-regression analysis to investigate the possible origins of heterogeneity among the studies included (Supplementary Table 3). Our findings revealed that study design was the only significant factor contributing to the observed heterogeneity. No significant variations were detected through the sensitivity analysis.

Qualitative MRI in CIDP patients

Three common abnormal MRI patterns of CIDP including hypertrophy, hyperintensity, and contrast enhancement have been shown in previous studies. Hypertrophy ranged from 23 to 91% of CIDP patients in cohorts. ^{5,6,13,14,31,34,37–41} Hyperintensity was found in 14–91% of patients. ^{4–6,13,30,37,39–41} Contrast enhancement was reported in 0–89% of patients (Supplementary Table 2). ^{4,5,31,34–37,40,41}

Diagnostic accuracy of quantitative MRN

For the 14 quantitative studies (23 results), the pooled sensitivity was 0.73 (95% CI 0.66–0.79) with heterogeneity (I² = 47.96%, p = 0.01) and the pooled specificity was 0.89 (95% CI 0.84–0.92) with heterogeneity (I² = 5.97%, p = 0.38) (Figure 3). The LR + was 6.4 (95%CI 4.4–9.3) and the LR- was 0.31 (95%CI 0.24–0.39). The DOR was 21 (95% CI 12–37). The SROC curve demonstrated the AUC was 0.89 (95%CI 0.86–0.92) (Figure 4).

Figure 3.

Forest plot showing pooled sensitivity and specificity of quantitative MRN in CIDP. Numbers are pooled estimates with 95% confidence interval (CIs) in brackets. Horizontal lines represent 95% CIs.

Figure 4.

Summary receiver operating characteristic (SROC) curve for quantitative MRN in CIDP, indicating an AUC of 0.89.

The assessment of quantitative MRN parameters

We did a subgroup meta-analysis for the following parameters containing more than three studies: Diameter, CSA, FA, and signal-related parameter (CR, SNR, CNR). The quantitative parameter CER only has been reported by one study so far, as well as Volume. Both of them were analyzed separately (Table 2).

Table 2.

Subgroup analysis according to the parameter and nerve site in quantitative studies

	Sensitivity (%)	I 2 (%)	p value	Specificity (%)	I 2 (%)	P value	LR+	I 2 (%)	P value	LR-	I 2 (%)	p value
Parameter
D	66.9	71.0	0.002	87.1	35.0	0.161	3.812	26.2	0.229	0.428	70.2	0.003
CSA	76.3	1.2	0.386	95.0	3.2	0.376	10.29	0.0	0.623	0.288	19.0	0.295
FA	84.8	0.0	0.789	91.4	0.0	0.977	9.632	0.0	0.981	0.176	0.0	0.832
CR,CNR, SNR	61.5	0.0	0.967	77.8	0.0	0.993	2.743	0.0	0.991	0.498	0.0	0.974
Nerve site
LSP	74.8	39.7	0.047	89.4	18.8	0.234	5.471	31.2	0.107	0.314	46.5	0.018
BP	59.8	54.5	0.067	82.1	0.0	0.703	3.320	0.0	0.957	0.524	40.5	0.151

D, diameter; CSA, cross-section area; LSP, lumbosacral plexus; BP, brachial plexus;FA,fractional anisotropy; CR,contrast ratio; SNR,signal-to-noise ratio; CNR,contrast-to-noiseratio;LR +, positive likelihood ratio;LR-, negative likelihood ratio.

Diameter

Seven studies applied the diameter in the diagnosis of CIDP with a pooled sensitivity (67%) and specificity (87%). ^{8–11,14,30,31} Two studies chose the brachial plexus and set the cut-off value to 3.99 mm and 4.6 mm for the mean value of C5–T1 roots. ^8,11 The remaining five studies defined the lumbosacral plexus as the proper nerve site. ^{9,10,14,30,31} Two studies set 5.8 mm and 7.0 mm for the L5 nerve root, respectively. ^30,31 In particular, a cut-off value ≥5.8 mm for the intraforaminal diameter of the L5 nerve root yielded both 100% sensitivity and specificity. ³⁰ Nonetheless, the pooled sensitivity (67%) is not satisfactory from other four studies, which applied 3D SHINKEI and came from the same author. ^8–11

CSA

CSA was applied in four studies with the pooled sensitivity (76%) and highest specificity (95%). ^20,29,30,32 Two studies chose the sciatic nerve as the measuring site and defined the cut-off value of 35 mm² at the mid-thigh level and 64.4 mm² at the point of maximum area. ^20,29 The other two studies chose the L3 to S1 roots and defined the mean CSA of 19.2 mm² and 28.2 mm^2, respectively, which could exhibit the best diagnostic value.

FA

Five studies adopted DTI-related parameter FA and yielded the highest pooled sensitivity (85%) and second highest specificity (92%). Three studies selected the L4-S1 nerve roots and defined the cut-off value as being below 0.361, 0.278, and 0.27, separately for CIDP. ^14,15,32 The other two studies reported FA cut-off value 0.45 for the sciatic nerve with sensitivity of 83 and 93% separately. ^21,33 However, one of the two studies used the average cut-off value 0.47 for sciatic, tibial, ulnar, median, and radian nerves with a higher AUC than 0.45 for the sciatic nerve. Consequently, we included the results with cut-off value of 0.47 in this meta-analysis. ³³

Signal-related parameter

Based on 3D SHINKEI, two studies with five results from the same author published in the same year reported these parameters. ^8,9 They were different quantification methods of hyperintense nerves. Thus, we combined these three parameters in subgroup analysis, giving relatively low pooled sensitivity (62%) and specificity (78%).Two results from one study chose the CR and SNR of C5-T1 roots ⁸ and three results from the other study applied CR, SNR, and CNR of T12–L5 roots or ganglion. ⁹

Volume

Only one study with DWIBS (diffusion-weighted whole-body imaging and background body signal suppression) reported the parameter Volume, a 3D measurement of nerve hypertrophy. ²⁸ The brachial plexus was chosen as the measuring site with a cut-off value 11.6 cm³/m² and good sensitivity (85%) but relatively low specificity (82%).

CER

Only one study with VIBE (volumetric interpolated breath-hold examination) applied this novel parameter CER, the quantification of contrast-enhancement of peripheral nerves in CIDP patients. ¹⁴ It was defined as the ratio of contrast enhancement nerve SNR to non-contrast enhancement nerve SNR. Regarding the measuring site, lumbosacral plexus showed both higher sensitivity (73%) and specificity (96%) than brachial plexus.

Reliability of quantitative MRN

Two studies reported intra-observer variability ^29,31 and five studies with eight results reported inter-observer variability. ^{8,15,29,31,32} The ICCs for intra-observer variability were 0.935 for CSA and 0.94 for diameter in lumbosacral nerve roots, which indicated excellent intra-observer consistencies. The pooled ICC for inter-observer agreement was 0.90 (95%CI 0.82–0.95) with a random-effects model (Figure 5). An ICC value above 0.9 showed excellent reliability. ²⁶

Figure 5.

Pooled correlation coefficient for interobserver variability in the quantitative MRN.

Discussion

CIDP is still primarily a diagnosis of exclusiondue to its high clinical heterogeneityand insufficient pathophysiological understanding. ⁴² More ancillary tools are clinically needed to help with the diagnosis. Compared with ultrasound and nerve conduction, MRI can sensitively detect deeper proximal nerve roots (brachial or lumbosacral plexus), where demyelination is more pronounced than distal nerve sites. ²

CIDP patients typically manifest hypertrophy, hyperintensity, and contrast enhancement in nerves on MRN. Hypertrophy nerve forms as a result of the pathological proliferation of Schwann cells and overlapping whorls of enlarged nerve fibers due to episodes of demyelination and remyelination. ⁴³ Hyperintensity in STIR can be attributed to interstitial edema and subendothelial inflammatory cell infiltration, ⁴ which can be seen to varying degrees in nerve fiber biopsies of patients with CIDP. Contrast enhancement can be attributed to the increased permeability and derangement of the blood-nerve barrier (BNB). ³⁴ These abnormal patterns are usually symmetrical and more prominent in lower limbs, which can be differentiated from other immune-mediated neuropathies like multifocal motor neuropathy (MMN). ^39,44,45 Therefore, MRI is useful in diagnosing CIDP while the current diagnostic value remains limited because of the high variability between judgments of radiologists and the lack of objective criteria. Former studies provided evidence that interrater reliability of brachial plexus MRI in chronic immune-mediated neuropathies was low with κ ranging from 0.3 to 0.491 for hypertrophy and 0.101 to 0.304 for signal intensity. ^6,7 Our results also showed a wide rate range of hypertrophy, hyperintensity, and contrast enhancement in CIDP patients, which indicated low reliability of qualitative assessments of brachial and lumbosacral plexus MRI. In comparison, quantitative MRN could show pretty good interobserver consistencies in the diagnosis of CIDP patients with a pooled ICC of 0.90, as well as excellent intraobserver consistencies. In light of this, quantitative MRN can be superior to qualitative analysis with objective criteria and certain cut-off values.

Apart from good reliability, our systematic review showed that quantitative MRN could be a promising tool in diagnosing CIDP with the sensitivity of 73% and specificity of 89%.In terms of specificparameters, SNR, CR, and CNR were associated with hyperintensity and showed a poor diagnostic value. It was reported that signal-related parameter T2 relaxation time on T2 mapping slightly increased in CIDP compared with healthy people, but no statistical difference was found. ¹² As a result, signal-related quantitative parameters are not recommended in the diagnosis of CIDP. Based on DTI and tractography, FA had the highest sensitivity (85%) for the diagnosis of CIDP. Chronic demyelinating lesions and mononuclear cell infiltration in the endoneurium, fasciculus, and epineurium reduced the structural integrity and directional coherence of nerve fibers, leading to decreased anisotropy. ⁴⁶ FA can indirectly reflect a range of changes, especially the loss of myelin in the nerve roots of CIDP patients. ⁴⁷ Kronlage et al found that decreased FA values were also strongly correlated with decreased NCV and increased F-wave delay. ³³ The decrease in FA was mainly due to an increase perpendicular to fiber orientation (RD), independent of diffusion in fiber orientation (AD). It is consistent with the current hypothesis that an increased RD may serve as a specific biomarker of demyelinating neuropathy. ⁴⁸ Decreased FA accompanied by increased RD is a solid finding in CIDP patients. As for the FA cut-off value, two studies set it to 0.446 and 0.45 for sciatic nerve ^21,33 and two studies set it to 0.27 and 0.278 respectively for L4-S1 rootsby ROC analysis. ^15,32 In conclusion, the cut-off value of 0.28 for average measurements of the L4 to S1 nerve roots at intraspinal, intraforaminal, and extraforaminal levels and 0.45 for the sciatic nerve with slabs acquired over the middle third of the thigh could distinguish between CIDP and healthy individuals.In consonance with this, the mean FA values of the sciatic nerve also increased distally from the S1 root level(0.314) to the femoral head level (0.567) in healthy people. ⁴⁹ It is vital to measure the specific nerve site for subsequent clinical practice and more studies on CIDP patients with FA in DTI are needed to verify these cut-off values. Moreover, the combined analysis of FA and other parameters can improve the diagnostic value of MRN. The combination of CER and FA could yield the largest AUC (0.973), sensitivity (92.2%) and specificity (96.3%) in differentiating between CIDP and healthy controls. ¹⁴

Parameters that reflect nerve hypertrophy in CIDP patients can also aidin the diagnosis. Compared with diameter, CSA had a better diagnostic performance for sensitivity and specificity because the nerve roots were not always circular. As a two-dimensional parameter, the CSAcan be easily measured and record more information than one-dimensional parameter diameter. The CSA is the direct and accurate reflection of the hypertrophic nerve. Recently, it has been reported to differentiate CIDP from axonal polyneuropathy (PNP) with a sensitivity and specificity of 75 and 100% while FA value may decrease in both CIDP and axonal PNP. ³² Significantly, inherited peripheral neuropathy like CMT1A could show larger CSA than CIDP.While it is not difficult to discriminate them according to the clinical features that CMT1A is characterized by a longer disease course and high-arched feet. Furthermore, the nerve volume measurement, a 3D reflection of hypertrophyadjusted by the body surface, showed considerable sensitivity (85%) with a cut-off value of 11.6 cm³/m². ²⁸ But so far, only one study has applied this parameter and more studies are needed to verify the validity.

Subgroup analysis for the nerve site (brachial plexus and lumbosacral plexus) showed that the higher sensitivity (75%) and specificity (89%) were related to the lumbosacral plexus than brachial plexus, suggesting that the lumbosacral plexus may be more likely to be affected in patients with CIDP. ⁸ Consistently, CIDP patients usually have walking difficulties earlier in clinical practice with the lower limbs more severely affected. ⁵⁰

This review, however, is subject to several limitations. Firstly, the number of included studies was small with not many individuals. Secondly, due to the different parameters applied in these studies, the meta-analysis showed a high degree of heterogeneity. But the heterogeneity was low when subgroup analysis was performed. Finally, we only included studies comparing CIDP patients with healthy controls because few diagnostic studies contained the disease control of CIDP mimics currently.

We conclude that there is limited evidence in the literature regarding quantitative MRN evaluation of CIDP. Our systematic review provides clinicians and researchers with the available evidence. Hypertrophy, hyperintensity and contrast enhancement are typical abnormal patterns of CIDP on qualitative MRI evaluation. Based on these qualitative assessments, quantitative MRN parameters can considerably help the identification of CIDP patients with higher sensitivity and reliability and a combination of parameters can increase the diagnostic accuracy. The lumbosacral plexus isa more suitable nerve site for measurements than brachial plexus on MRN. We recommend FA and CSA in the lumbosacral plexus as promising diagnostic parameters for CIDP patients. The FA value of sciatic nerve below 0.45 and L4-S1 nerve roots below 0.28 can indicate CIDP.

3D SPACE,three-dimensional sampling perfection with application optimized contrast using different flip angle evolution; MPRAGE, Magnetization prepared - rapid Gradient Echo;3D SHINKEI, three-dimensional nerve-sheath signalincreased with inked rest-tissue rapid acquisition of relaxation imaging; PD,proton density; DWIBS,diffusion-weighted whole-body imaging and background body signal suppression; VIBE,volumetric interpolated breath-hold examination; LS, Lumbosacral; P, Prospective; R, Retrospective; CSA, Cross-Sectional Area; DRG,Dorsal Root Ganglion; DTI, Diffusion Tensor Imaging; FA, fractional anisotropy; CER, contrast-enhancement ratio; CR, contrast ratio;SNR signal-to-noise ratio;CNR, contrast-to-noise ratio.