Olfactory training affects the correlation between brain structure and functional connectivity

Abstract

Purpose

and background: Neuroimaging studies have increasingly found functional connectivity (FC) changes and structural cortical abnormalities in patients with post-traumatic anosmia (PTA). Training and repeated exposure to odorants lead to enhanced olfactory capability. This study is conducted to investigate the correlations between FC and cortical thickness on the olfaction-related regions of the brain in PTA after olfactory training (OT).

Methods

Twenty-five PTA patients were randomly divided in three groups: (1) 9 control patients who did not receive any training, (2) 9 patients underwent classical OT by 4 fixed odors, and (3) 7 patients underwent modified OT coming across 4 sets of 4 different odors sequentially. Before and after the training period, all patients performed olfactory function tests, and magnetic resonance imaging (MRI). Sniffin’ Sticks test was used to assess olfactory function. MRI data were analyzed using functional connectivity analysis and brain morphometry.

Results

Modified OT resulted in heightened activation in the medial orbitofrontal cortex and anterior cingulate cortex and increased FC between the piriform cortex (PIRC) and the caudate cortex. Conversely, classical OT induced increased activation in the insula cortex and greater FC between the PIRC and the pre-central gyrus. Furthermore, after OT, both training groups achieved significantly improved scores in the changes in brain connectivity associated with OT, which were attributable to anatomical measures.

Conclusions

This study demonstrates that intensive olfactory training can enhance functional connectivity, and this improvement correlates with structural changes in the brain’s olfactory processing areas.

Introduction

A non-obstructed nasal airway and intact neuronal pathways are essential for a functional olfactory system. Traumatic head injury (TBI) that results in the disconnection of any part of these pathways, may derive alterations in olfactory ability. ¹ TBI most often leads to total loss of function (anosmia), decreased sensitivity (hyposmia), alterations in odor quality (dysosmia), and hallucination (phantosmia).^2,3 Olfactory loss not only entails social, emotional, and behavioral consequences but also initiates reorganization processes in the brain. The olfactory system is extraordinarily plastic due to mechanisms that are under extensive investigation at both the cellular and cognitive levels. ⁴ A few published studies indicated the odor learning and odor expertise influence the human brain in adults. ⁵ These odor experts therefore showed that odor training and experience lead to functional and then structural reorganization of olfactory brain areas. In this way, training (OT), first described by Hummel et al. (2009), ⁶ seems to be the most effective behavioral therapy in patients with olfactory dysfunction. ⁷ Furthermore, the effect of OT has been proved in a large multicenter cross-over study ⁸ as well as by recent meta-analyses. ⁹ This effect seems to be even more effective when a variety of odorants are rotated over time and training is prolonged. ¹⁰

Traumatic olfactory loss is most commonly diagnosed through a psychophysical olfactory function test that primarily relies on patient self-reporting. ¹¹ Whereas smell ability is mostly assessed by using semi-objective tests such as the Sniffin’ Sticks test, ¹² magnetic resonance imaging (MRI) can also be objectively used to explain the structural and functional changes in the brain and understand olfactory loss. Using functional MRI (fMRI) to measure changes in blood oxygen level-dependent (BOLD) signal during odor administration in patients with smell disorders will contribute to a fundamental under-standing of the activity and functional connectivity (FC) of olfactory system. ¹³ Odor-stimulated or olfactory fMRI was first introduced by Yousem et al ¹⁴ and can be used to detect selective activation of the primary and higher-order olfactory region by olfactory psychophysical tasks.

Previous fMRI studies that have stimulated the olfactory system, found differences in functional activations between anosmia and healthy adults, whereas more recent studies have focused on functional connectivity networks by employing independent component analysis (ICA) and seed-based connectivity analysis (SBA). However, SBA analysis indicated that in anosmia, less functional connections of the networks’ global maxima to other brain regions were present than in controls.^15–17 On the other hand, structural MRI approaches, voxel-based morphometry (VBM) ¹⁸ and surface-based morphometry (SBM), ¹⁹ are the two most common methods that used for cortical morphology. These methods showed that cortical thickness (CTh) and cortical volume (CV) in olfactory processing areas is related to performance in olfactory tasks and that patients suffering from smell loss due to different etiologies exhibit an architectural alteration in the same structures.^20,21 So far, functional and structural changes were analyzed and interpreted mostly separately. Both functional connectivity and brain morphometry might lead to better predictions on the disease prognosis and its treatment outcome. ²² We have previously reported that a 16-week OT in post-traumatic patients led to a significant increase in CTh and CV of several brain regions. ²³

Now, we conducted a study on patients who lost their sense of smell due to a traumatic event. ²³ Our investigation focused on the effects of 16 weeks of olfactory training on the activation of the whole brain, which we analyzed using GLM. We also examined the connectivity patterns in the networks related to the sense of smell using SBA. Then, based on our previous report on the SBM and VBM findings, we examined the correlations between brain FC and CTh on the olfaction-related regions of the brain, with two training methods. To do so, we performed fMRI in the PTA patients to determine the impact of OT on the neuronal activation pattern and FC in two olfactory training groups (1: classical OT group, 2: modified OT group) and healthy controls before and after OT. Since, we used our previous results on these patients that reported the effects of OT on structural brain measures with a focus on CTh and CV. Finally, we hypothesized that a long-term olfactory training (1) leads to an alteration in neuronal activation patterns in olfactory and other brain areas for both trained groups, (2) results in changes in FC in whole brain areas in both OT groups, (3) leads to the correlations between FC and CTh on the olfaction-related regions of the brain, and (4) use of more odors in the modified training method that may increase the effectivity of OT on the FC and structural brain measures.

Materials and methods

Study design

A total of 30 patients were informed about the aims of the study, and written consent was obtained from them. Five patients included in this study were excluded from the analysis due to discontinued training program; 25 subjects with an age range 20 to 45 years (mean age ±SD = 28.00 ± 7.56 years, 21 males, 16 right-handed) were included in the final analysis (see Table 1). Moreover, the mean duration of smell loss was 9.53 ± 6.1 months. Furthermore, none of the participants had significant health problems (e.g., a history of neurological diseases such as Parkinson’s and Alzheimer’s, congenital anomalies (Kalman Syndrome), pregnancy, diabetes, alcohol addiction, another olfactory disorder, etc.) All patients were evaluated based on two testing sessions. First, patients were examined by an ENT doctor to determine the cause of olfactory dysfunction. Also, the olfactory performance test, using the Sniffin’ Sticks test, was applied to estimate the severity of olfactory dysfunction. Next, fMRI measurement was performed, using an odor-stimulated task-based fMRI. Following the fMRI session, all patients were randomly divided into three groups: (1) The control group (Control-G), 9 patients who received no training; (2) The classical OT group (COT-G), 9 patients undergone OT using four constant odors; (3) The modified OT group (MOT-G), 7 patients undergone OT using four different odorants in each month. After completing the training period, subjects were followed up by examining for olfactory performance and the second fMRI session (Figure 1).

Table 1.

Demographic characteristics of participants.

	Control–G, n = 9 mean (SD)	COT-G, n = 9 mean (SD)	MOT-G, n = 7 mean (SD)
Age (year)	26.67 ± 7.59	29.00 ± 7.28	29.29 ± 8.67
Sex (female/male)	1/8	1/8	2/5
Handedness (right/left)	6/3	5/4	5/2
BMI	25.7 ± 3.4	24.5 ± 5.3	25.4 ± 4.0
Disease duration (month)	10.00 ± 8.2	9.56 ± 3.67	9.57 ± 5.35
Education (year)	10.6 ± 2.9	13.1 ± 3.1	12.6 ± 2.9
Smoking (yes/no)	0/8	0/8	0/7

Data represent mean ± SD, except for sex, handedness, and smoking. Control-G = Control group, COT-G = Classical olfactory training group, MOT-G = Modified olfactory training group, BMI = Body mass index, n = Number of subjects.

Figure 1.

Schematic description of the study protocol. Control-G: Control group; COT-G: Classical olfactory training group; MOT-G: Modified olfactory training group; OT: Olfactory training.

Olfactory training

The first olfactory training method, named classical OT was introduced by Hummel et al. ⁶ In this method, patients received four constant odors over the training period. These odors are Rose, Eucalyptus, Lemon, and Thyme. The second method, known as the modified OT method was introduced by Altundag et al. ¹⁰ In this method, the training approach aims to stimulate olfactory receptors by different odors over a certain period. We performed the training by four different packages including the Rose, Lemon, Thyme, and Eucalyptus for the first month; Narcissus, Strawberry, Cardamom, and Peppermint for the second month; Saffron, Banana, Cinnamon, and Garlic for the third month; Orange Blossom, Orange, Vanilla, and Vinegar for the fourth month. All odorants were obtained from Adonis Gol Darou Company, Tehran and selected based on the four “odor prism” of flowery, fruity, aromatic, and resinous. Further, these odors filled in the dark brown-colored glass jars with 1 mL of the respective odorant (soaked in cotton pads to prevent spilling). All patients were conducted to expose themselves twice daily to four odorants. Furthermore, all patients were informed to fill a checklist form for all training sessions after daily training.

Olfactory testing

The olfactory performance was assessed using a clinical battery test, Sniffin’ Sticks test, with three sub-tests including odor threshold, odor discrimination, and odor identification. ¹² This test was done by pen-like devices for exposing odorants at 2 cm of the participant’s nose for 2–3 s. In the first step, the odor threshold is performed by a single-staircase, 3-alternative, forced-choice procedure. In the next step, the odor discrimination is obtained by 16 triplets of pens in which one of the odors was different from the other two odors in each row. Last, to achieve the odor identification task, 16 common odors are presented in a multiple-choice answering format, including a list of four descriptors for each odor. Finally, the scores obtained from all three sub-tests were reported as TDI (Threshold-Detection-Identification) score. The TDI score ≥30.5 points defines as the normosmia, 16.5 < TDI <30.5 as hyposmia, and TDI <16.5 as anosmia. ¹¹

MRI acquisition and fMRI task paradigm

MRI measurements were performed on a 3 T MRI scanner (Siemens Magnetom Prisma Model) using a full-head 20-channel receiver coil. The tasked-based fMRI was acquired in the axial plane (oriented parallel to the planum sphenoidal to minimize artifacts) using a gradient-echo T2*-sensitive echo-planar imaging (GE-EPI) sequence (TR = 3000 ms; TE = 30 ms; image matrix = 64 × 64; voxel size = 3 × 3 × 3 mm³; field of view (FOV) = 256 mm × 256 mm; number of slices = 45; scan of volume: 109). To overlay functional data, anatomical images were acquired using a high-resolution T1-weighted MPRAGE (magnetization prepared-rapid gradient echo) sequence, with the following parameters: TR = 1800 ms; TE = 3.53 ms; inversion time = 1100 ms; image matrix = 256 × 256; voxel size = 1 × 1 × 1 mm³ (in total, 160 contiguous slices of 1-mm thickness).

Synchronous with functional imaging, olfactory-stimulants were delivered in a block task design, using a computer-controlled MR-compatible olfactometer, ²⁴ according to Figure 2. This paradigm was composed of interleaved time intervals 30 and 10 s which were, respectively, switched between rest (odorless) and stimulant (odor) conditions. ²⁵ In stimulant conditions, two rose and vanilla odors were used, respectively, whilst in rest conditions, the odorless airflow was delivered to the participant’s nose through a respiratory mask. The paradigm (odorless-vanilla-odorless-rose) was repeated four times and the entire paradigm lasted for approximately 320 s. During fMRI scanning, all participants were instructed to breathe through the nose naturally and keep their eyes closed. Also, a pair of sponge pads were placed in the ears to reduce the sounds received from the MRI system.

Figure 2.

Schematic overview of the olfactory paradigm for the tasked-based fMRI session. During each block, participants experienced alternating conditions of odor and odorless air. The intervals for each condition were set at 10 s for either vanilla or rose odor and 30 s for odorless air, respectively.

fMRI data pre-processing

fMRI images were processed using tools developed by the University of Oxford’s Functional Magnetic Resonance Imaging of the Brain (FMRIB) division, Software Library (FSL) version 6.0.1. ²⁶ Briefly, preprocessing steps are included high pass filtering (80 Hz), slice-timing correction, motion correction,^15,27 spatial smoothing using a 5 mm full-width at half-maximum (FWHM), ²⁷ linear registration of functional images to the standard of Montreal Neurological Institute (MNI) space. ²⁸ Anatomical images were skull-stripped using Brain Extraction Tool in the FSL, ²⁹ and co-registration with functional images using boundary-based registration. Statistical analysis for fMRI was performed using a generalized linear model (GLM) with convolving of double-gamma hemodynamic response function and experimental design matrix. To provide a descriptive overview, activation patterns, were obtained with z-threshold >2.3 and corrected cluster significance level of p < .05. ²⁷ In the first-level analysis, the whole brain activation analysis was performed using FSL’s FEAT tool.^30,31 Then, second-level analysis was run based on one sample t-tests to evaluate neuronal activation pattern in each stimulation condition (vanilla and rose) with a false discovery rate (FDR)–corrected threshold of p < .05. To assess the effect of the olfactory training on the neuronal activation, a paired two-sample t test analysis was tested for the patient group (before OT vs after OT), using the same statistical threshold. In a last step, group differences were done using one-way ANOVA (FDR-corrected p < .05).

Further to explore the FC, seed-based analysis (SBA) was performed using the CONN toolbox (https://www.nitrc.org/projects/conn), implemented in MATLAB framework (MATLAB 9.5, The MathsWorks Inc., Natick, MA, USA). ³² The piriform cortex (PIRC), one of the structures of the primary olfactory cortex, was selected to consider as seed regions for further functional connectivity analysis. The seed regions, in the left and right PIRC, were manually created with a 10 mm sphere on a standard brain using the WFU PickAtlas software. ³³ The center of the seed region located at the following MNI coordinates, left PIRC (−14, 0, −22) and right PIRC (−12, 2, 22), reported in previous studies. ³⁴ The preprocessing steps are included functional volume realignment, unwarping and segmentation, slice timing correction, segmentation, and normalization of structural volumes to the MNI template, normalization of functional volumes, and smoothing of functional volumes with a 5 mm FWHM kernel. Additionally, the artifact detection tools software package was used to detect the motion outliers. Then, the nuisance parameters extracted from the motion-correction step (CSF, GM, and WM) were used for regressing out. In the first-level analysis, the mean time series from the preselected seed region was correlated with the time series of all other voxels in the brain for each subject and for each condition. In the second-level analysis, significant differences of FC between groups were calculated using one-way ANOVA in each stimulation condition (vanilla and rose), with an FDR-corrected threshold of p < .05. In addition, to assess the effect of the olfactory training on the FC, a paired 2-sample t test analysis was tested for the patient group (before OT vs after OT), using the same statistical threshold.

Correlations between functional connectivity and cortical measurements

The possible correlations between changes in functional connectivity and brain morphometry related to training were tested in both olfactory training groups. First, structural MRI data were analyzed using the CAT12 toolbox (https://dbm.neuro.uni-jena.de/vbm/) implemented in SPM12 software (https://www.fil.ion.ucl.ac.uk/spm) and MATLAB (The MathWorks, Natick, MA). First, using the longitudinal segmentation tool, the T1-weighted images from the before- and after-OT period were realigned. The images were segmented into GM, WM, and CSF. Subsequently, the segmented GM images were spatially normalized to a template in MNI space using the high dimensional Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra (DARTEL). ³⁵ The segmented normalized GM images from the before- and after-OT period are again realigned and underwent quality control using CAT12 quality control tools. Then, cortical thickness data was resampled and smoothed with a 15 mm FWHM kernel. ¹⁹ Last, whole-brain analyses were performed by testing the association between changes in morphometry measures (after OT > before OT) and changes in FC measures (after OT > before OT). An uncorrected threshold (P-uncorrected <0.001) and a voxel size greater than 10 were used for analysis.

Statistical analysis

All statistics were performed using the Statistical Package for the Social Sciences, version 20.0 (SPSS, Chicago, IL, USA). For all olfactory performance results, the mean and standard error of the mean (SEM) were calculated. To compare TDI scores and its constituent tests, the between-groups comparison was evaluated using analysis of variance (ANOVA), with post hoc Bonferroni tests, before OT as well as after OT. Also, the comparison between before and after olfactory training was performed using within-groups analysis with paired t-tests. The p < .05 was considered statistically significant.

Results

Olfactory performance

Table 2 presents the olfactory performance of all patients. Comparing the impact of OT among three groups revealed significant within-subject effects for odor identification (F2,23 > 4.38, p = .026) and TDI score (F2,23 > 4.74, p = .02). Post hoc t-tests indicated higher odor identification in the MOT-G and a superior TDI score in the COT-G compared to the control group (p = .044, p = .023), respectively. No pre-training differences were observed among the three patient groups (p > .05). Within-group comparisons (after OT > before OT) using paired t-tests demonstrated a significant improvement in TDI score (COT-G, p = .002, MOT-G, p = .001) and odor identification score (COT-G, p = .004, MOT-G, p < .001). No significant differences were noted for the odor detection threshold task and odor discrimination task in both training groups (p > .05). In the Control group, neither the TDI score nor its constituent tests exhibited differences before and after OT (p > .05).

Table 2.

Olfactory performance measurements before and after olfactory training.

	Olfactory performance measures
	Before OT				After OT
	TDI	T	D	I	TDI	T	D	I
Control-G; n = 9	12.41 ± 1.61	2.19 ± 0.73	6.44 ± 0.81	3.77 ± 0.86	13.61 ± 1.42	1.61 ± 0.61	6.11 ± 0.58	5.88 ± 1.21
COT-G; n = 9	12.33 ± 0.95	1.00 ± 0.00	6.33 ± 0.44	5.00 ± 0.74	19.05 ± 1.25 a, b	2.83 ± 0.94	7.22 ± 0.59	9.00 ± 0.66 a
MOT-G; n = 6	12.62 ± 1.69	1.29 ± 0.29	6.5 ± 0.88	4.83 ± 0.90	17.95 ± 1.42 a	1.25 ± 0.12	7.00 ± 0.51	9.83 ± 0.45 a, b

Results are shown as mean ± SEM. (a) Shows significant differences of results (after OT > before OT) by the within-group analysis (p < .05). (b) Shows significant differences of results in comparison with the control group by the between-groups analysis (p < .05). T: Threshold, D: Discrimination, I: Identification; Control-G: Control group, COT-G: Classical olfactory training group, MOT-G: Modified olfactory training group.

Neuronal activation patterns

Vanilla Odor Stimulus: Table 3 outlines neuronal activation patterns in response to the vanilla odor stimulus before and after 16 weeks of OT using GLM. Prior to OT, no activation was observed in primary and secondary olfactory regions. After 16 weeks of OT, the MOT-G showed activation in the right anterior cingulate cortex (ACC) and right amygdala in response to vanilla, while no response was observed in the COT-G and control group (Figure 3). After OT, the COT-G and control group showed no significant changes compared to before OT. In contrast, the MOT-G exhibited alterations in brain regions, including the right inferior frontal gyrus (IFG, triangularis part), right ACC, and left medial orbitofrontal cortex (OFC) (Figure 4, (a)). Furthermore, between-group analysis after OT demonstrated neuronal activation in the right ACC and left medial OFC in the MOT-G compared to the control group (Figure 4, (c)).

Table 3.

Neuronal activation patterns in response to vanilla odor before and after olfactory training.

State (contrast)	Anatomical areas	Voxel	MNI (mm)			Z-MAX	p-value
			ZMAX	YMAX	XMAX
Control-G
pre	No suprathreshold cluster
post	No suprathreshold cluster
post>pre	No suprathreshold cluster
pre>post	No suprathreshold cluster
COT-G
pre	No suprathreshold cluster
post	No suprathreshold cluster
post>pre	No suprathreshold cluster
pre>post	No suprathreshold cluster
MOT-G
pre	No suprathreshold cluster
post	R ACC	256	14	32	26	3.66	0.0074
	R amygdala	280	32	0	−22	4.26	0.0207
post>pre	R IFG (triangularis)	675	36	22	12	3.92	<0.0001
	R ACC	509	14	40	24	3.32	0.0004
	L medial OFC	230	−10	36	−10	1.32	0.04
pre>post	No suprathreshold cluster
Post MOT-G > post Control-G
-	R ACC	1931	8	48	20	4.39	<0.0001
	L medial OFC	475	−8	50	−12	4.38	0.0039

The result shows a statistically significant cluster (p < .05). All coordinates are given in Montreal Neurological Institute (MNI) space. Control-G: Control group; COT-G: Classical olfactory training group; MOT-G: Modified olfactory training group; pre: Before olfactory training; post: After olfactory training; R: right; L: Left; OFC: Orbitofrontal cortex; ACC: Anterior cingulate cortex; IFG: Inferior frontal gyrus.

Figure 3.

Representation of the mean of neuronal activation patterns (z-threshold >2.3 and cluster significance level of p < .05). The cluster of neuronal activation is represented in Tables 3 and 4. The top-down order of groups is Control-G, COT-G, and MOT-G. Control-G: Control group; COT-G: Classical olfactory training group; MOT-G: Modified olfactory training group; pre: Before olfactory training; post: After olfactory training; OT: Olfactory training.

Figure 4.

Two-sample t test (z-threshold >2.3 and cluster significance level of p < .05) of the neuronal activation patterns. The cluster of neuronal activation is represented in Tables 3 and 4. (a) The within-group comparison for the vanilla stimulus in the MOT-G (post > pre). (b) The within-group comparison for the rose stimulus in the COT-G (post > pre) (c) The between-group comparison for the vanilla stimulus (post MOT-G > post Control-G). (d) The between-group comparison for the rose stimulus (post COT-G > post Control-G). Control-G: Control group; COT-G: Classical olfactory training group; MOT-G: Modified olfactory training group; pre: Before olfactory training; post: After olfactory training.

Rose Odor Stimulus: Neuronal responses to the rose odor stimulus before and after OT were analyzed using GLM (Table 4). No activation was observed in the study groups before OT. After OT, the COT-G exhibited activation in the right insula cortex (IC), and the MOT-G showed activation in the left superior temporal gyrus. No significant clusters were identified in the control group (Figure 3). Within-group analyses after OT revealed differences only in the right IC in the COT-G compared to before OT (Figure 4, (b)). Additionally, between-group analysis after OT showed differences in the right IC in the COT-G compared to the control group (Figure 4, (d)). However, no activation changes were observed in the control and MOT-G groups for the comparison-group analysis.

Table 4.

Neuronal activation patterns in response to rose odor before and after olfactory training

State (contrast)	Anatomical areas	Voxel	MNI (mm)			Z-MAX	p-value
			ZMAX	YMAX	XMAX
Control-G
pre	No suprathreshold cluster
post	No suprathreshold cluster
post>pre	No suprathreshold cluster
pre>post	No suprathreshold cluster
COT-G
pre	No suprathreshold cluster
post	R insula cortex	1044	38	22	0	2.93	0.0044
post>pre	R insula cortex	996	44	2	−6	3.49	<0.001
pre>post	No suprathreshold cluster
MOT-G
pre	No suprathreshold cluster
post	L superior temporal gyrus	282	−34	14	−38	3.58	0.0362
post>pre	No suprathreshold cluster
pre>post	No suprathreshold cluster
Post COT-G > post Control-G
-	R insula cortex	883	48	16	−6	3.22	0.0034

The result shows a statistically significant cluster (p < .05). All coordinates are given in Montreal Neurological Institute (MNI) space. Control-G: Control group; COT-G: Classical olfactory training group; MOT-G: Modified olfactory training group; pre: Before olfactory training; post: After olfactory training; R: right; L: Left.

Seed based functional connectivity

Table 5 outlines differences within and between groups before and after OT through SBA method. The within-group analysis on the COT-G and the control group did not reveal significant differences after OT, while the MOT-G displayed several alterations. After OT, the between-group analysis using the same parameters as the within-group analysis indicated connectivity changes in trained patient groups compared to the control group.

Table 5.

Comparison of functional connectivity of the piriform cortex with other brain areas before and after olfactory training.

Seed	Brain region	Voxels	MNI (mm)			T-Value	P-FDR
			XMAX	YMAX	ZMAX
Post MOT-G > pre-MOT-G (VANILLA)
R PIRC	R middle frontal gyrus
	R superior frontal gyrus	86	34	12	62	−10.03	0.001
	L middle frontal gyrus
	Pre-central gyrus	37	−50	12	32	−7.21	0.04
L PIRC	R superior frontal gyrus	57	28	36	54	−9.47	0.016
	R lateral occipital cortex	42	48	−70	2	9.18	0.03
Post COT-G > post Control-G (ROSE)
L PIRC	R pre-central gyrus	70	36	−10	56	6.77	0.005
Post MOT-G > post Control-G (VANILLA)
R PIRC	L caudate	76	−12	24	2	8.49	0.006

The result shows a statistically significant cluster (p < .05). All coordinates are given in Montreal Neurological Institute (MNI) space. Control-G: Control group; COT-G: Classical olfactory training group; MOT-G: Modified olfactory training group; pre: Before olfactory training; post: After olfactory training; R: right; L: Left; PIRC: Piriform cortex.

Right PIRC Connectivity Map: Within-group comparison using the right PIRC as a seed region revealed significantly higher functional connectivity with the right piriform in MOT-G after OT for the vanilla stimulus. This increase was observed in several regions, including the middle frontal gyrus in two clusters of 86 and 37 voxels (Figure 5(a)). The between-group analysis showed stronger FC between the right PIRC and a cluster in the caudate cortex with 76 voxels in MOT-G compared to control after OT for the vanilla stimulus (Figure 5(b)).

Figure 5.

Two-sample t test (P-FDR <0.05) of the piriform functional connectivity map. (a) The within-group comparison (post > pre) for the vanilla stimulus in the MOT-G using the R PIRC as seed revealed negative correlation in the left middle frontal gyrus (light-blue arrow) and right middle frontal gyrus (blue arrow). (b) The between-group comparison (post MOT-G > post Control-G) for the vanilla stimulus using the R PIRC as seed revealed positive correlation in the left caudate (red arrow). (c) The within-group comparison (post > pre) for the vanilla stimulus in the MOT-G using the L PIRC as seed revealed positive correlation in the right lateral occipital cortex (red arrow) and negative correlation in the right superior frontal gyrus (blue arrow). (d) The between-group comparison (post COT-G > post Control-G) for the rose stimulus using the L PIRC as seed revealed positive correlation in the right pre-central gyrus (red arrow). Control-G: Control group; COT-G: Classical olfactory training group; MOT-G: Modified olfactory training group; pre: Before olfactory training; post: After olfactory training RH: Right hemisphere, LH: Left hemisphere.

Left PIRC Connectivity Map: Using the left PIRC as a seed region, within-group comparison in MOT-G after OT for the vanilla stimulus demonstrated significantly increased FC with the right superior frontal gyrus (cluster size = 57) and right lateral occipital cortex (cluster size = 42) compared to pre-OT (Figure 5(c)). Comparing with pre-OT, the between-group analysis showed stronger FC between the left PIRC and the right pre-central gyrus, with a cluster size of 70 voxels in COT-G compared to control group after OT for the rose stimulus (Figure 5(d)).

Brain function—structure correlations

To assess whether any of the changes in brain connectivity associated with OT were attributable to anatomical measures, a series of exploratory correlations between the statistically significant brain connections and brain structure measures was conducted, resulting in 5 correlations that reached a significance level of p < 0 .001, uncorrected for multiple tests (Table 6). In the COT-G, there was a positive correlation between the CTh in the right inferior occipital cortex and the global efficiency of FC related to the vanilla odor (Figure 6(a)). Additionally, a positive correlation was observed between the CTh in the left middle cingulate gyrus and left paracentral gyrus and the global efficiency of FC related to the scent of rose (Figure 6(b)). In the MOT-G, a positive correlation was found between the CTh in the right postcentral gyrus and right OFC and the global efficiency of FC related to vanilla scent (Figures 6(c) and (d)).

Table 6.

Correlation between functional connectivity and structural brain measures in response to scent before and after 16 weeks of olfactory training.

Brain region	Cluster size	MNI				T-value
		Side	X	Y	Z
COT-G (VANILLA)
Inferior occipital cortex	16	R	44	−65	−18	6.52
COT-G (ROSE)
Middle cingulate gyrus	47	L	−2	−13	38	11.61
Paracentral gyrus	19	L	−14	−43	57	6.50
MOT-G (VANILLA)
Postcentral gyrus	88	R	25	−39	68	23.46
Orbital frontal cortex	51	R	11	32	−26	22.98
MOT-G (ROSE)
No suprathreshold cluster

The result shows a statistically significant cluster (P-uncorrected <0.001). All coordinates are given in Montreal Neurological Institute (MNI) space. Control-G: Control group; COT-G: Classical olfactory training group; MOT-G: Modified olfactory training group; pre: Before olfactory training; post: After olfactory training; R: right; L: Left.

Figure 6.

Correlation between cortical thickness and functional connectivity related to olfactory training (P-uncorrected <0.001 and cluster size greater than 10). (a) Positive correlation of CTh and FC with vanilla scent in classic group. (b) Positive correlation of CTh and FC with rose scent in COT group. (c) and (d) Positive correlation of CTh and FC with vanilla scent in the MOT group. COT-G: Classical olfactory training group; MOT-G: Modified olfactory training group; OFC: Orbital frontal cortex.

Discussion

We have previously reported a comprehensive analysis of brain structure in post-traumatic patients, which showed that olfactory training significantly increases the cortical thickness and density of several brain regions. ²³ For the first time, we combined the examination of structural and functional data from the brain, which carries significant potential for decoding the complex mechanisms of the human brain. Three significant discoveries underscore the impact of olfactory training on both the functional neural pathways and the structural integrity of the brain in patients diagnosed with Post-Traumatic Anosmia. Firstly, a 16-week modified olfactory training regimen resulted in an enhanced neuronal activation in the orbitofrontal cortex (OFC) and insula. Secondly, the olfactory training induced an increased functional connectivity (FC) between the posterior insular cortex (PIRC) and both the pre-central gyrus and the caudate cortex. Lastly, following the olfactory training, both training groups demonstrated significantly improved scores in the changes in brain connectivity associated with olfactory training, which were attributable to anatomical measures.

In this study, we initiated an exploration of neuronal activation and the connectivity of functional networks using task-based fMRI, stimulated by rose and vanilla scents. Our findings indicate the modified OT brain activation in the left medial OFC and right ACC in response to the vanilla stimulant. The OFC, a component of the secondary olfactory cortex is recognized as the central hub for cognitive processing and memorization of odors.^36–39 The left OFC is more involved in the pleasantness and unpleasantness processing of one odor. ⁴⁰ Prior research has indicated that the medial OFC is activated in response to pleasant odors.^41,42 The fMRI studies have corroborated that the ACC plays a pivotal role in attention, with its activation being contingent on the hedonic valence of the olfactory stimulus.^43–45 Activation was observed in the right ACC and right amygdala following a period of 16 weeks of olfactory training in the MOT group. The amygdala plays a crucial role in the encoding of higher-order representations pertaining to the quality, identity, and familiarity of odors. It is significantly associated with the processes of learning and memory retention of these odors.^36,46 Furthermore, the amygdala is specifically activated during the intensity coding of both pleasant and unpleasant odors. ³⁶ Our findings indicate an upsurge in activity within the left OFC, right ACC, and right IFG in the MOT group. It appears that when an odor could be readily identified, its retention in working memory was manifested as sustained activity, predominantly in the IFG region. ⁴⁷

Upon exposure to the scent of a rose, classical OT elicited neuronal activation within the right insular cortex. The insular cortex, a critical region in the brain, plays a significant role in olfactory processing. This region has been extensively studied and is associated with various aspects of odor perception. These include, but are not limited to, the discernment of an odor’s quality, the assessment of its intensity, and the subjective evaluation of its pleasantness.^48–50 Furthermore, the insular cortex’s involvement in olfactory processing underscores its integral role in our sensory perception and highlights the complex interplay between our senses and brain regions. ⁵¹ Following a period of 16 weeks, we observed neuronal activation in the superior temporal gyrus as a result of the MOT group. The superior temporal gyrus, a critical brain region, is deeply engaged in the preliminary phases of odor processing. It has a vital function in differentiating an odorant as either new or known, even in cases where the odorant was not recognized previously. This underscores the significant role of the superior temporal gyrus within our olfactory system, enhancing our capacity to understand and navigate our sensory surroundings. This region’s involvement in odor recognition and differentiation further emphasizes the complexity of our sensory systems and the intricate neural networks that facilitate our interactions with the world around us. ⁵²

Our findings indicated an upsurge in brain activity within the COT group in reaction to the rose stimulant, and similarly in the MOT group in response to the vanilla stimulant. This observed increase in brain activity could potentially be attributed to the patients’ recent exposure to these specific odor stimulants during their training program prior to the imaging experiment. ⁴⁴ Our findings lend credence to the hypothesis that olfactory recognition is significantly modulated by olfactory memory. This substantiates the existing body of evidence highlighting the intimate anatomical interconnections between the olfactory system and the neural circuits implicated in memory processing. This complex interplay accentuates the pivotal role memory plays in our sensory perception, especially within the realm of olfactory recognition. It is worth noting that the olfactory system is unique in its direct connection to the hippocampus, a brain region critical for memory and learning. This connection bypasses the thalamus, which is the primary relay station for other sensory systems.^53,54 Moreover, the olfactory bulb, which processes incoming smell information, has direct links to the amygdala, which is involved in emotional learning and memory. This could explain why smells often evoke powerful, emotion-laden memories. Our study thus contributes to the growing understanding of the intricate relationship between sensory perception and memory, particularly in the context of olfaction.

The piriform cortex (PIRC), the most substantial structure and the principal component of the primary olfactory cortex in humans, plays a pivotal role in olfactory processing.^55,56 It serves as a conduit for information transmission from the olfactory bulb to various brain regions including the orbitofrontal cortex (OFC), amygdala, hypothalamus, insular cortex (IC), entorhinal cortex, and hippocampus. ⁵⁷ Our findings revealed an augmentation in the number of neural connections to the seed regions in both the MOT and COT groups, in comparison to the control group. This suggests that training may enhance the connectivity of the olfactory network. In our study, we observed that the caudal part of the OFC, which is associated with odor detection, established a connection with the PIRC in the MOT group. Similarly, the right pre-central gyrus, a region implicated in the motor network, was found to be connected to the PIRC in the COT group. These findings underscore the complex interplay between sensory perception and motor function in the context of olfactory processing. They also highlight the potential for targeted training to modulate these connections, thereby enhancing olfactory recognition and response. ⁵⁸

In their seminal work, Kallendorfer and colleagues documented the reconfiguration of functional connectivity as a consequence of a training regimen in patients suffering from olfactory loss post-infection. ¹⁵ Their study revealed a significant reduction in the functional connectivity between non-olfactory networks and the piriform cortex, which was used as the seed region. Contrastingly, our findings indicate that OT led to an enhancement in the connections with the piriform cortex in the MOT group, as compared to the pre-training state. Specifically, we observed a negative correlation with the bilateral middle frontal gyrus and the right superior frontal gyrus, and a positive correlation with the right lateral occipital cortex. This apparent discrepancy could potentially be attributed to the distinct nature of the underlying damage in the two conditions. In the case of head trauma, the damage is predominantly localized in the olfactory centers and bulbs. On the other hand, degenerative damage in post-infectious olfactory loss primarily occurs in the olfactory epithelium. ⁵⁹ This highlights the importance of considering the specific etiology of olfactory loss when interpreting the effects of olfactory training on brain connectivity.

Moreover, we were unable to identify any connectivity between the PIRC and the entire brain networks in both the COT and control groups, relative to their pre-training state. It is crucial to acknowledge that traumatic injuries inflict damage on distributed brain networks, and these functional alterations are frequently linked with neurocognitive deficits and unfavorable functional outcomes. ⁶⁰ Therefore, our current investigation into PTA suggests that the functional reorganization of the brains of injured patients necessitates an extended period for the re-establishment of olfactory networks. The connections we observed between the motor and visual cortex networks and the piriform cortex provide evidence for a compensatory mechanism involving the recruitment of non-olfactory networks. ⁶¹ This is consistent with a recent fMRI study that reported enhanced functional connectivity in visual, motor, and subcortical networks in patients suffering from traumatic anosmia. These findings underscore the resilience of the brain and its capacity to adapt and reorganize in response to injury.

The results of our research investigating the correlations between functional connectivity and cortical thickness in olfaction-related brain regions in PTA after OT provide intriguing insights. The exploratory correlations between brain connections and brain structure measures revealed five significant correlations. This suggests that changes in brain connectivity associated with OT may be attributed to anatomical measures, providing a potential link between structural and functional changes in the brain. Recent investigations have utilized cortical thickness as a metric to examine brain networks. These studies have unveiled an inherent characteristic of the entire brain network. Additionally, they have revealed a modular architectural attribute of the entire brain.^62,63 These findings provide valuable insights into the potential mechanisms underlying the effects of OT on brain structure and function in patients with PTA. However, further research is needed to confirm these findings and to explore their implications for the optimization of OT methods. It would also be interesting to investigate whether similar correlations exist for other scents and whether these correlations vary depending on the specific characteristics of the scents used in OT. Substantial evidence exists demonstrating brain plasticity in response to pathological damage caused by various sudden neurological deficits. ⁶⁴ Our belief, based on a comparison of results between trained patients and a control group, is that repeated short-term exposure to odors in the training program appears to stimulate the growth of olfactory receptor neurons. ⁶⁵ Although periodically modifying the “training odors” could potentially enhance the treatment’s effectiveness, ¹⁰ the MOT method did not prove superior to the COT method in this study. However, for satisfactory outcomes from olfactory training, periodically changing the odors has been shown to increase patient satisfaction with the treatment.

The current study has a few limitations. Firstly, while fMRI is more expensive and less accessible compared to subjective psychophysical testing, it offers a more detailed understanding of the alterations in brain functional connectivity associated with olfaction. ⁶⁶ Secondly, the limited sample size, especially in the MOT group, restricts the statistical power of our results. As a result, it is crucial for future research to confirm our findings using a larger sample size to determine the efficacy of traditional and modified olfactory training in treating patients with traumatic anosmia. Thirdly, our evaluation was confined to two familiar and pleasant olfactory stimuli, and exploring unpleasant stimuli could yield different outcomes. Additional research is needed to refine the MOT technique.

Conclusion

This comprehensive study of patients with PTA undergoing OT yields detailed insights. The fMRI results suggest that smell training aims to reactivate dormant olfactory areas and reorganize brain structure and functional connectivity. The correlations between brain function and structure highlight the intricate interplay between these domains, emphasizing a holistic approach to comprehending and managing olfactory dysfunction.

Ethical statement

Ethical approval

All study procedures involving human participants were in accordance with the ethical standards of Iran University of Medical Sciences (Ethical code: (IR.IUMS.FMD.REC.1397.065), and were registered in the Iranian Registry of Clinical Trials (IRCT2017102034225N1). Informed consent was obtained from all individual participants included in the study.