Exploring individual fixel-based white matter abnormalities in epilepsy

Abstract Diffusion MRI has provided insight into the widespread structural connectivity changes that characterize epilepsies. Although syndrome-specific white matter abnormalities have been demonstrated, studies to date have predominantly relied on statistical comparisons between patient and control groups. For diffusion MRI techniques to be of clinical value, they should be able to detect white matter microstructural changes in individual patients. In this study, we apply an individualized approach to a technique known as fixel-based analysis, to examine fibre-tract-specific abnormalities in individuals with epilepsy. We explore the potential clinical value of this individualized fixel-based approach in epilepsy patients with differing syndromic diagnoses. Diffusion MRI data from 90 neurologically healthy control participants and 10 patients with epilepsy (temporal lobe epilepsy, progressive myoclonus epilepsy, and Dravet Syndrome, malformations of cortical development) were included in this study. Measures of fibre density and cross-section were extracted for all participants across brain white matter fixels, and mean values were computed within select tracts-of-interest. Scanner harmonized and normalized data were then used to compute Z-scores for individual patients with epilepsy. White matter abnormalities were observed in distinct patterns in individual patients with epilepsy, both at the tract and fixel level. For patients with specific epilepsy syndromes, the detected white matter abnormalities were in line with expected syndrome-specific clinical phenotypes. In patients with lesional epilepsies (e.g. hippocampal sclerosis, periventricular nodular heterotopia, and bottom-of-sulcus dysplasia), white matter abnormalities were spatially concordant with lesion location. This proof-of-principle study demonstrates the clinical potential of translating advanced diffusion MRI methodology to individual-patient-level use in epilepsy. This technique could be useful both in aiding diagnosis of specific epilepsy syndromes, and in localizing structural abnormalities, and is readily amenable to other neurological disorders. We have included code and data for this study so that individualized white matter changes can be explored robustly in larger cohorts in future work.


Introduction
Epilepsy is widely considered a network disorder, with hyperexcitability and hypersynchrony within brain networks driving disturbance in brain function. 1Although the study of brain network disturbances in epilepsy has predominantly focused on abnormal functional connectivity, there is growing interest in changes to the brain's structural connections, which likely subserve these functional network disruptions.Indeed, both focal and generalized epilepsies have been shown to exhibit widespread abnormalities in white matter (WM) fibre pathways, and these structural network disruptions are now understood to be a key characteristic of the epilepsies. 1,2iffusion MRI or diffusion-weighted imaging (DWI) is currently the only tool available to noninvasively probe the in vivo WM architecture of the brain and has played a crucial role in untangling structural network changes in epilepsy.Diffusion MRI studies have revealed syndrome-specific patterns of WM alterations, both within predefined cohorts, [3][4][5][6] and across large consortia studies. 2 These WM changes appear clinically meaningful, as they are associated with cognitive function 7,8 and postsurgical outcome. 9,10In epilepsy patients with focal brain abnormalities, diffusion MRI measures may also help to localize epileptogenic lesions. 11,12ver the past two decades, major technical advancements in diffusion MRI have brought the corresponding potential for clinical utility in epilepsy.Yet, in clinical practice, the use of this technique is limited to a few well-defined applications; for example, to identify and avoid the optic radiation or corticospinal tract (CST) when planning surgical interventions in focal epilepsy. 13This could be attributed to diffusion MRI studies in epilepsy being predominantly focused on group analyses, comparing patient cohorts with control cohorts.These group-level studies are inherently desensitized to individual patient differences, where clinical value is likely to be derived.For advanced neuroimaging techniques to be valuable in clinical practice, they must provide biologically and clinically meaningful information about individuals.
Unfortunately, there are challenges in translating diffusion MRI methods into individual-patient use.First, DWI metrics often relate to features of the observed diffusion process rather than direct properties of the underlying biology, such that clinical interpretation is often difficult. 14,15econd, the detection of individual-level changes in neuroimaging-derived measures is hampered by the multiple comparison problem when using classical statistical frameworks.][18] In this study, we apply the fixel-based analysis (FBA) framework at the individual-patient level, to examine WM fibre-tract-specific changes in individuals with epilepsy.Unlike the more commonly used voxel-averaged measures derived from techniques like diffusion tensor imaging (DTI), FBA involves modelling of multiple fibre orientations within a voxel, 19,20 such that derived measures are both sensitive and specific to individual fibre orientations within a voxel.This provides more biologically interpretable measures of WM fibre-tract changes than can DTI, particularly in crossingfibre voxels that constitute the majority of brain WM. 21ere, we provide the first (to our knowledge) individualized framework for FBA, and examine fibre-tract-specific changes in individuals with various causes of epilepsy.We provide means to quantify and visualize individual-level tract changes and incorporate data harmonization approaches into the fixel-based framework, including code and data for further exploration.We demonstrate clinically meaningful changes in individual epilepsy patients, which could be useful in the clinical work-up of syndromic epilepsies, and in localizing small epileptogenic lesions.

Participants
Ten individuals with epilepsy were included in this study: two patients with temporal lobe epilepsy (TLE) and hippocampal sclerosis (HS); two patients with MRI-negative TLE; two patients with focal cortical dysplasia (bottom of the sulcus dysplasia); two patients with periventricular nodular heterotopia (PVNH) [one who had additional polymicrogyria (PMG)]; one patient with Dravet's syndrome; and one patient with progressive myoclonic epilepsy.People with epilepsy were recruited through the Austin Health Comprehensive Epilepsy Program, the first seizure clinic and/or neurology department at Austin Health and had a confirmed diagnosis of epilepsy.Patients were selected from previous imaging studies where we performed group comparisons between patients with specific epilepsy syndromes or causes and control participants.For this proof-of-principle study, we selected patients who exhibited clinical signs that were representative of the particular syndrome or cause of epilepsy, rather than patients who showed complex aetiology.Clinical data for each patient included in this study is available in Table 1.
Ninety control participants were included in the normative sample.Control participants were adults [mean age: 36.9 years (±13.3),range: 18-64 years] who had no history of seizures, psychiatric illness, or significant head injury.Age histograms for the healthy control cohort are provided in the Supplementary Material (Supplementary Fig. 1).
Written informed consent was provided by all participants or their parents/guardians.The study was approved by the Austin Health Human Research Ethics Committee.

MRI data
All participants underwent an MRI scan at the Florey Institute of Neuroscience and Mental Health.MRI data were acquired at 3 T either on a 3 T Siemens Skyra or 3 T Siemens Tim Trio between 2011 and 2019.
DWI data were acquired on the Skyra with a 20-channel head coil receiver, with the following parameters: 60 axial slices, TR/TE (repetition time/time to echo) = 8400/110 ms, 2.5 mm isotropic voxels, in-plane parallel acceleration factor 2, 64 diffusion-weighted images (b = 3000 s/mm 2 ) and at least 1 b = 0 image.Equivalent DWI was acquired on a Siemens Tim Trio with a 12-channel head coil receiver with the following parameters: 60 axial slices, TR/TE = 8300/110 ms, 2.5 mm isotropic voxels, in-plane parallel acceleration factor 2, 60 diffusion-weighted images (b = 3000 s/mm 2 ) and 8 b = 0 images.A reverse phase-encoded b = 0 image was acquired in all cases (on both scanners) to correct for B0 field inhomogeneities.

Diffusion-weighted image processing
All DWI data were pre-processed and analyzed primarily using MRtrix3, 23 with several specific tasks deferred to other software as described below.DWI data from all participants were pre-processed before analysis.Data were first denoised, 24 after which Gibbs ringing artefacts were removed. 257][28] Bias field correction was then performed based on the mean b = 0 image using ANTs (Advanced Normalization Tools) N4, 29 and DWI images were upsampled to a voxel size of 1.3 mm 2 using cubic b-spline interpolation.
Following these pre-processing steps, fibre orientation distribution functions (ODFs) were computed using single-shell 3-tissue constrained spherical deconvolution (SS3T-CSD), 30 using group-averaged response functions for WM, grey matter (GM) and CSF. 31,32Joint bias field correction and intensity normalization were then performed. 33,34 'healthy' population template image was generated using WM ODF images from a subset of 20 healthy control participants selected randomly from the control cohort, using an iterative registration and averaging approach. 35Spatial correspondence was achieved by registering WM ODF images from all participants (controls and patients) to this template using ODF-guided non-linear registration. 35,36 whole-brain tractogram was generated using probabilistic tractography on the population template image.Twenty million streamlines were generated, after which the SIFT (Spherical-deconvolution Informed Filtering of Tractograms) algorithm was applied to filter the tractogram to 2 million streamlines to reduce reconstruction biases. 37e then applied the FBA 20 framework, where the term 'fixel' refers to a specific fibre population within a single image voxel.Different voxels within an image may contain a different number of fixels.These are computed based on segmentation of the WM ODFs 37 with correspondence established between subject-specific fixels and those of the WM ODF template. 20A measure of fibre density and crosssection (FDC) was obtained at each WM fixel in the population template space for all participants.This FDC metric combines microstructural and morphological information, such that it is sensitive to changes in the density of fibres passing in a particular direction within a given voxel, as well as the changes in the cross-section of fibre bundles that traverse multiple image voxels. 20Connectivity-based smoothing was performed on fixel-based measures for all participants. 38

Fibre-tract extraction
Major fibre tracts of interest were delineated on the WM ODF template using TractSeg. 39Tracts of interest were selected for analysis based on having previously exhibited abnormality in group studies comparing epilepsy patients to control cohorts were selected for analysis.These tracts were: the arcuate fasciculus (AF), corpus callosum (CC), cingulum bundle (CG), CST, fornix (FX), inferior fronto-occipital fasciculus (IFOF), inferior longitudinal fasciculus (ILF), superior cerebral peduncle (SCP), three segments of the superior longitudinal fasciculus (SLF I, II and III) and uncinate fasciculus (UF).With the exception of the CC, all tracts were bilateral (one left hemisphere tract and one right hemisphere tract).
Delineated fibre tracts were then converted into fixel masks using the 'tck2fixel' command in MRtrix3, 23 and subsequently thresholding this fixel image to create a binary tract fixel mask.The mean FDC value across all fixels within each tract-of-interest was computed for each participant.Given that DWI data from patients included in this study were acquired from two different scanners, we first performed tract data harmonization using ComBat. 18Following this, assumptions of normality were tested for each tract across the healthy control cohort, after which rank-based inverse normal transformations (INTs) were applied.

Tract-level analysis
The mean and standard deviation (SD) of tract FDC were computed across the healthy control cohort for each WM tract.Following this, Z-scores were calculated for individual patients with epilepsy when compared to the healthy control cohort.Spider plots were used to display tract Z-scores for individual patients across all tracts.The pipeline for individual tract-based analysis is shown in Fig. 1.
A summary laterality index (LI) was computed using the respective sums of all Z-scores across all tracts in each hemisphere: LI = L sum -R sum /L sum + R sum .
All tract-level analysis was performed in R statistical software (version 3.6.3).Packages utilized include the 'neuroCombat' package for tract data harmonization, 'RNOmni' package 40 to perform rank-based INTs, and 'fmsb' package 41 for generation of spider plots.The code to perform analysis and generate plots is available in R, along with tract data from this study (https://github.com/remikamito/spidey).

Fixel-level analysis
Subject-specific deviations from the healthy control population were also explored at the level of each individual WM fixel.Here, a similar pipeline to the tract-based approach was implemented, whereby the healthy population FDC mean and SD were computed at each WM fixel.Z-scores were then calculated at each template fixel for each epilepsy patient to create a Z-map of fixel-level deviations compared to the control cohort.

Results
The results of the individualized fixel analysis framework are illustrated with 10 case studies.Table 2 summarizes the key findings for each patient, including a summary LI for tractbased results, and WM regions that exhibited decreased FDC in the whole-brain fixel-level maps.As expected, individual patients with epilepsy exhibited highly variable patterns of WM abnormality compared to the healthy control cohort, both at the tract-level and fixel-level.Figure 2 is a demonstration of the spider plot display that is used to present patient tract-level results in Figs 3-5 (data from Patient 1).In patients with TLE and HS (Patients 1 and 2), tract-level spider plots demonstrate predominantly ipsilateral abnormality of tracts related to the temporal lobe (Table 2), with the greatest FDC reductions in the UF, AF, ILF and IFOF.Fixel-level    In all patients, the tract LI indicated greater tract abnormality on the side ipsilateral to the respective lesion(s).Tract-level findings also appear to reflect the clinical or cognitive profiles in these individuals: for example, in Patient 8, left hemisphere tracts appear normal while there is notable right hemisphere WM abnormality, in keeping with the cognitive profile for this individual, who had normal intellect but some non-verbal memory deficits.

Patients 5 and 6: progressive myoclonus epilepsy and dravet syndrome
In most cases, either the most affected tract or the pattern of tract-level changes provided some indication of the spatial location of the structural epileptogenic abnormality; however, tract-level changes were of insufficient granularity to identify the focal lesion, particularly in the two cases with BOSD.Importantly, fixel-level Z-maps provided finer grain detail, exhibiting patterns of WM abnormality that were concordant with the epileptogenic lesions.In the two patients with highly focal BOSD, although Patient 9 exhibited a striking fixel-level finding that was spatially concordant with the epileptogenic lesion (right precentral sulcus BOSD), Patient 10 exhibited only subtle WM changes, with small foci of decreased FDC and only a small cluster of potentially concordant change in the right central region near the lesion (right post-central BOSD).

Discussion
We present an advanced diffusion MRI approach quantifying WM abnormalities in individual epilepsy patients, by adapting the FBA framework to individualized assessment.This approach provides proof-of-principle evidence of the potential clinical use of this advanced diffusion MRI technique in epilepsy.In this case series, we show individual-level findings of tract-and fixel-specific change in patients with varying types and causes of epilepsy (TLE, PME, Dravet syndrome and MCDs).We demonstrate that patients with epilepsy exhibit variable patterns of WM abnormalities and that these WM changes are concordant with the clinical profile of these patients.

Individualized white matter abnormalities reflect syndrome-specific patterns
In this study, we assessed fixel-based metrics in 10 patients with varying epilepsy causes and syndromes.This enabled the identification of individualized differences within white matter fibre tracts when compared to a control cohort.The patterns of WM abnormality in these patients appeared concordant with the specific epilepsy syndrome or cause in each individual.However, patient-level variability was also observed, highlighting potentially heterogeneous WM profiles across individuals, particularly in those with spatially distributed focal lesions.
For the patients with TLE, variable patterns of WM abnormality were observed, in line with the distinct TLE subtypes of these patients.Although WM abnormalities appeared particularly marked within the temporal lobe ipsilateral to the lesion in the two patients with HS, they also extended beyond the affected temporal lobes.The tract-specific decreases in FDC were more marked in the two patients with HS (Patients 1 and 2), than in the two LN TLE patients, who exhibited relatively minimal WM abnormality.This is consistent with group-level findings, which have demonstrated more extensive patterns of abnormality in patients with HS than in MRI-negative TLE, both within WM 2,4,5 and in the cortex. 42,43However, these differences have been suggested to be influenced by clinical characteristics such as disease duration. 2Of note, both MRI-negative TLE patients in our study had a later age of onset and shorter disease duration than the two TLE-HS patients, and the difference in extent of WM abnormalities may reflect these clinical characteristics.In contrast to the patients with TLE, the two patients with specific epilepsy syndromes associated with generalized seizure types exhibited much more bilateral patterns of WM abnormality.The patient with PME exhibited a highly cerebellar signature in their pattern of WM abnormality.Patients with PME exhibit progressive neurological dysfunction, including cerebellar ataxia, and the causes of PME are often associated with cerebellar pathology that is absent in the cerebrum. 44he pattern of tract-specific WM abnormality in the PME patient included in this study (Patient 5) suggested a marked cerebellar focus on this individual's pathology, which is concordant with the clinical signature of PME.Although this individual did not have the more common form of Unverricht-Lundborg Disease (ULD), a group study in ULD using DTI has similarly demonstrated cerebellar, subcortical and thalamocortical WM changes. 45ravet Syndrome is a severe drug-resistant epilepsy syndrome characterized by generalized tonic-clonic seizures, often with intellectual disability.The patient with Dravet syndrome (Patient 6) exhibited widespread abnormality in WM tract FDC, with virtually all tracts of interest exhibiting substantial decreases in FDC, and much of the brain's WM being implicated at the fixel-level.This finding is in line with the marked WM changes demonstrated at a group level using FBA, 46,47 which may result from abnormal axonal growth in Dravet syndrome, as shown in animal models. 48he individuals with malformations of cortical development exhibited highly patient-specific changes.In three of these four individuals (Patients 7-9), the pattern of WM abnormality was spatially concordant with the identified seizure-causing structural abnormality.Patients 7 and 8, who had unilateral PVNH (with additional PMG in Patient 8), exhibited WM abnormalities in a highly lateralized manner, with the most notable abnormalities in the WM proximal to the heterotopic lesions.Patient 9 exhibited focal abnormality in the right precentral region, spatially concordant with the epileptogenic BOSD in this individual, while Patient 10, who had somewhat more subtle BOSD in a spatially proximal area (right post-central region), did not show any notable WM changes.
These individual findings were somewhat in contrast to previous group-level analyses.0][51][52][53][54] In the case of BOSD patients, we recently showed that despite spatial heterogeneity in the location of the epileptogenic lesions, patients exhibited a typical pattern of bilateral structural network abnormality. 6These bilaterally distributed WM changes have been suggested to reflect the secondary effect of epilepsy on the brain, 6,52 rather than primary pathology in these patients, given the heterogeneous distribution of these lesions.In group-level analyses, we are likely sensitive to such seizure-affected WM changes, which may be common to patients within an otherwise quite heterogeneous group, while being largely desensitized to subject-specific abnormalities.On the other hand, individual-level abnormalities are likely to be more marked in the area immediately proximal to these epileptogenic lesions and may be more clinically meaningful than group-level findings.

Clinical significance of individual tract-specific analysis
An important consideration in MRI studies in epilepsy, particularly in patient cohorts with spatially varied pathology, is the heterogeneity within patient groups.Group-based studies inherently compare group means, which enables identification of co-localized effects between subject groups, but disregards individual-level differences outside these co-localized regions as noise.Neuroimaging studies over the past decade, including large collaborative studies, 2,42 have predominantly focused on group-level analyses that demonstrate common brain changes in patients with varying types and causes of epilepsy.
There is growing interest in moving from group-level studies to individualized neuroimaging analysis and prediction in epilepsy. 55This is particularly pertinent for precision medicine in epilepsy, where neuroimaging plays a key role in diagnosis and clinical management.To meet this need, neuroimaging techniques must offer clinically valuable insights at an individual-patient level.This study examined whether WM abnormalities in individual patients with varying types and causes of epilepsy offer clinically insightful information, and translate an advanced diffusion MRI tool into the individual-patient level.
In light of our findings, we believe the clinical significance of this individual-level FBA is 2-fold.First, tract-specific WM changes may provide insight into the underlying epilepsy type or syndrome, and we may be able to detect patterns of WM disruption that are diagnostically useful.Second, individual-level Z-score maps of fibre-tract-specific abnormality may help to localize focal epileptogenic abnormalities, particularly in cases where these abnormalities are highly subtle (e.g. in BOSD).
We note that our study is not the first to investigate WM abnormalities in individual patients with diffusion MRI.As with group-level analyses, studies examining WM abnormalities in individual patients with epilepsy have typically adopted the diffusion tensor model, demonstrating subject-specific microstructural abnormalities. 49,56However, DTI-based results are inherently voxel-averaged or voxel-aggregated, rendering them potentially sensitive, though not specific to underlying fibre structures. 15,57Although advanced diffusion MRI methods have similarly been applied to probe microstructural changes, 11,58 these studies have also depended on measures that are quantified on a per-voxel basis.
Importantly, by implementing an FBA approach, this work moves beyond voxel-averaged measures to a fibre-tract-specific model. 20,59This approach enables more comprehensive insight into WM changes within specific fibre populations, as has been demonstrated extensively in many group-level analyses. 60eported changes to the derived metrics (namely, FDC) can therefore be ascribed to specific fibre tracts.In this work, we demonstrate that this advanced diffusion MRI approach can be extended to applicability at an individual-patient level, and in doing so, can offer important clinical insight into the specific fibre tracts that may be affected in individuals.

Limitations and future directions
Despite the promise of this individualized approach, there are limitations to this work.This study should be considered proof-of-principle, underscoring the importance of developing a methodological approach for detecting individual tractspecific changes.As such, we included a set of epilepsy patients (n = 10) with variable causes of epilepsy to gauge the types of patients for whom these individualized analyses may be helpful in future.We also selected patients from retrospective cohorts who exhibited well-characterized clinical signs typical of their epilepsy diagnosis, which may contribute to some selectivity bias.It is important to note that, unlike group-level analyses, improvements in statistical power cannot be achieved in this approach by including a greater number of patients.However, future studies with larger cohorts will provide a clearer picture of individual changes in WM in epilepsies, both within and across diagnoses, and this work may be considered a pilot study for such larger cohort studies.To this end, the code and data used to generate individualized tract-based changes are included in this work.Future work should systematically test individual-level tract changes in a large cohort of epilepsy patients, including those with complex or atypical aetiology, to determine the clinical utility of this technique.
In this work, we focused on the spatial localization of any WM changes observed, rather than on quantifying the significance of any effect, and therefore, did not perform stringent statistical corrections for multiple comparisons.It is essential to note that clinical tools at the individual level should focus on quantifying effect size over statistical significance.Setting somewhat arbitrary thresholds for statistical significance (e.g. by performing false discovery rate or familywise error corrections as has been the case in other individualized approaches 61 ) runs the risk of disregarding potentially important clinical information; however, the converse of excess false positives may complicate interpretation for clinicians.To this end, approaches that capture variation across multiple clinical and demographic parameters within large control cohorts may be valuable, as they will likely help to identify key deviations that are of biological or clinical importance.
One approach to circumvent the challenges of classical statistical inference methods in individual-patient analysis is to adopt machine learning approaches.In particular, normative modelling approaches that incorporate machine learning models have gained substantial momentum in the last few years, as a means to probe single subject-level deviations against a large normative sample. 62,63Recent work in other neurological conditions and disorders, including psychiatric disorders 63,64 and dementia, 65 has demonstrated that individual-level brain structural abnormalities can be identified using such normative approaches.To a limited extent, microstructural abnormalities have been demonstrated in patients with focal cortical dysplasia 66 using normative diffusion MRI-based models, highlighting the approach's potential in epilepsy.Normative modelling approaches can be applied both independently to select measures (e.g.independent Z-scores per tract), or to extract multivariate measures of deviation (e.g.spatial Z-maps across brain measures), and by implementing machine learning algorithms that are trained on normative features, these approaches are likely to be more sensitive to than classical univariate frameworks. 66However, given that these machine learning approaches rely on computationally expensive training utilizing big datasets, we believe that proof-of-principal or pilot studies such as this will play an important role in firstly validating the potential clinical impact of a given approach prior to generating large-scale models, in order to minimize wasted resources. 67

Conclusion
For neuroimaging studies to be implemented in clinical use, it is crucial to quantify brain abnormalities in individual patients, rather than detect group-level differences between patient and control populations.Our study provides proof-of-principle evidence that by using advanced diffusion MRI analysis, we can identify tract-specific WM abnormalities in individual patients with various types of epilepsy.Our findings suggest that individual WM abnormalities could provide clinically useful information, as the observations appear either concordant with the epilepsy syndromes or spatially concordant with structural abnormalities.We believe these findings are a step towards moving imaging techniques from group-level studies to clinical relevance, supporting precision medicine for the individual patient in future.We have released codes for creating individualized plots for Z-statistics to further encourage other research groups to assess the clinical potential of this approach.

Figure 3
Figure3shows the tract-and fixel-level results for the four patients with TLE (Patients 1-4).Patient 1 is an individual with

Figure 1
Figure 1 Methodology for individual tract-based FDC analysis.1.A measure of FDC is quantified at each WM fixel in template space, for all healthy control participants and individuals with epilepsy.2. Mean FDC is computed within select tracts-of-interest.3. ComBat data harmonization is performed on the tract data to adjust for any scanner effects.4. Assumptions of normality are tested across the control cohort.5.The mean and SD are computed across the healthy control cohort for each tract.6. Tractwise Z-scores are computed for each patient.INT, inverse normal transformation.

Table 2
Key tract-level and fixel-level findings for each Relatively extensive WM abnormalities in fixels within temporal lobe (L > R) and association fibre pathways connecting into L inferior frontal lobe, along with thalamocortical pathways bilaterally 2 −0.136Right UF −2.73Some diffuse WM abnormalities bilaterally, but with greatest cluster of abnormality within right temporal lobe and connecting into thalamic structures 3 −0.258Left UF −1.00 Small clusters of diffuse abnormality, including small areas within right temporal lobe extending into inferior frontal lobe, along with left inferior frontal cluster 4 0.063 Left cingulum −1.22Some diffuse WM abnormalities in temporal lobes bilaterally (L > R), along with bilateral thalami and splenium of CC 5 WM abnormalities in left CG, splenium of CC and bilateral fornices and surrounding posterior horn of lateral ventricles (L > R), and left medial temporal lobe 8 −1.594Right AF −2.04 Right hemisphere WM abnormality within temporal lobe, WMWM surrounding posterior horn of lateral ventricle and extending into parietal WM superiorly 9 −0.163Right CST −2.19 Major focus of WM abnormality in right inferior parietal/posterior frontal WM surrounding post-central sulcus 10 0.121 Left SCP −0.83No sizeable clusters of WM abnormality observed, although some diffuse subtle abnormalities observed in right central sulcus and right superior cerebellar WM a Positive values indicate L > R abnormality; Negative indicates R > L abnormality.Tract laterality is computed across all included tracts in each hemisphere.

Figure 2
Figure 2 Exemplar spider plot display for tract z-scores in an individual patient (patient 1).Z-scores are plotted for an individual patient (n = 1) in a spider plot display.Each data point represents the Z-score for a given tract in the individual patient, with Z-scores for left hemisphere tracts shown on the lower layer (in blue) and Z-scores for right hemisphere tracts shown on the upper layer (in red).The inner dotted line (and corresponding data points in black) reflects a Z-score of −1.96, and outer solid line reflects the healthy control mean.The solid grey lines correspond to Z-scores of −1, −2, −3 and −4, with greater magnitude (more negative) scores toward the centre of the plot.The CC is included at the top centre, and the Z-score for this tract is included in both left and right hemisphere spider plots.Tracts with a Z-score of <−1.96 in either hemisphere are labelled in bold.AF, arcuate fasciculus; CC, corpus callosum; CG, cingulum; CST, corticospinal tract; FX, fornix; IFOF, inferior fronto-occipital fasciculus; SCP, superior cerebellar peduncles; SLF, superior longitudinal fasciculus; UF, uncinate fasciculus.

Figure 4 Figure 5
Figure 4 shows tract and fixel-level results for two examples of patients with specific epilepsy syndromes: progressive myoclonus epilepsy (PME) and Dravet syndrome.Patient 5 is an individual with PME with reasonable cognitive function.PME is commonly associated with cerebellar abnormalities, and the tract-level analysis revealed striking bilateral abnormality within the SCP, with minimal

Figure 3
Figure 3 Tract-and fixel-level results for four patients with TLE.(A) Two patients with HS: Patient 1 is an individual who has left TLE with HS (L TLE-HS); Patient 2 is an individual with right TLE with HS (R TLE-HS).(B) Two patients with LN TLE: Patient 3 is an individual with MRI-negative or LN-RTLE; and Patient 4 is an individual with LN-LTLE.In all cases, spider plots on the left show Z-scores for FDC measures in each tract-of-interest (see Fig. 2 for interpretation).Fixel-level Z-score maps are shown on the right, with fixels exhibiting a Z-score of −1.96 or greater.Axial slices are placed at 10 mm increments.AF, arcuate fasciculus; CC, corpus callosum; CG, cingulum; CST, corticospinal tract; FX, fornix; IFOF, inferior fronto-occipital fasciculus; SCP, superior cerebellar peduncles; SLF, superior longitudinal fasciculus; UF, uncinate fasciculus.

Figure 5
Figure 5 Results in lesional epilepsies.(A) Pane shows individual-level results for two patients with PVNH, one with additional PMG.In both cases, lesions were unilateral, with Patient 7 exhibiting left-sided PVNH, and Patient 8 exhibiting right-sided PVNH and PMG.(B) Panel shows individual-level results for two patients with BOSD: Patient 9 had a small BOSD in the right precentral sulcus; Patient 10 had a BOSD in the right post-central sulcus.See Fig. 2 for spider plot interpretation for Panels A and B. (C) Panel shows the lesion locations for all patients (Patients 7-10) on a single axial slice from a T1-weighted image in each subject's own space. 22