Decomposing MRI phenotypic heterogeneity in epilepsy: a step towards personalized classification

Shared differential factors underlying individual spontaneous neural activity abnormalities in major depressive disorder

BACKGROUND

In contemporary neuroimaging studies, it has been observed that patients with major depressive disorder (MDD) exhibit aberrant spontaneous neural activity, commonly quantified through the amplitude of low-frequency fluctuations (ALFF). However, the substantial individual heterogeneity among patients poses a challenge to reaching a unified conclusion.

METHODS

To address this variability, our study adopts a novel framework to parse individualized ALFF abnormalities. We hypothesize that individualized ALFF abnormalities can be portrayed as a unique linear combination of shared differential factors. Our study involved two large multi-center datasets, comprising 2424 patients with MDD and 2183 healthy controls. In patients, individualized ALFF abnormalities were derived through normative modeling and further deconstructed into differential factors using non-negative matrix factorization.

RESULTS

Two positive and two negative factors were identified. These factors were closely linked to clinical characteristics and explained group-level ALFF abnormalities in the two datasets. Moreover, these factors exhibited distinct associations with the distribution of neurotransmitter receptors/transporters, transcriptional profiles of inflammation-related genes, and connectome-informed epicenters, underscoring their neurobiological relevance. Additionally, factor compositions facilitated the identification of four distinct depressive subtypes, each characterized by unique abnormal ALFF patterns and clinical features. Importantly, these findings were successfully replicated in another dataset with different acquisition equipment, protocols, preprocessing strategies, and medication statuses, validating their robustness and generalizability.

CONCLUSIONS

This research identifies shared differential factors underlying individual spontaneous neural activity abnormalities in MDD and contributes novel insights into the heterogeneity of spontaneous neural activity abnormalities in MDD.

Association of Normative and Non-Normative Brain Networks With Cognitive Function in Patients With Temporal Lobe Epilepsy

BACKGROUND AND OBJECTIVES

Despite their temporal lobe pathology, a significant subgroup of patients with temporal lobe epilepsy (TLE) is able to maintain normative cognitive functioning. In this study, we identify patients with TLE with intact vs impaired neurocognitive profiles and interrogate for the presence of both normative and highly individual intrinsic connectivity networks (ICNs)-all toward understanding the transition from impaired to intact neurocognitive status.

METHODS

This retrospective cross-sectional study included patients with TLE and matched healthy controls (HCs) from the Thomas Jefferson Comprehensive Epilepsy Center. Functional MRI data were decomposed using independent component analysis to obtain individualized ICNs. In this article, we calculated the degree of match between individualized ICNs and canonical ICNs (e.g., 17 resting-state networks by Yeo et al.) and divided each participant's ICNs into normative or non-normative status based on the degree of match.

RESULTS

100 patients with TLE (mean age 42.0 [SD: 13.7] years, 47 women) and 92 HCs were included in this study. We found that the individualized networks matched to the canonical networks less well in the cognitively impaired (n = 24) compared with the cognitively intact (n = 63) patients with TLE by 2-way mixed-measures analysis of variance (impaired vs intact mean difference [MD] -0.165 [-0.317, -0.013], p = 0.028). The cognitively impaired patients showed significant abnormalities in the profiles of both normative (impaired vs intact MD -0.537 [-0.998, -0.076], p = 0.017, intact vs HC MD -0.221 [-0.536, 0.924], p = 0.220, and impaired vs HC MD -0.759 [-1.200, -0.319], p < 0.001) and non-normative networks (impaired vs intact MD 0.484 [0.030, 0.937], p = 0.033, intact vs HC MD 0.369 [0.059, 0.678], p = 0.014, and impaired vs HC MD 0.853 [0.419, 1.286], p < 0.001) while the intact patients showed abnormalities only in non-normative networks. At the same time, we found that normative networks held a strong, positive association with the neuropsychological measures, with this association negative in non-normative networks.

DISCUSSION

Our data demonstrated that significant cognitive deficits are associated with the status of both canonical and highly individual ICNs, making clear that the transition from intact to impaired cognitive status is not simply the result of disruption to normative brain networks.

De novo depression following temporal lobe epilepsy surgery

Surgical removal of the mesial temporal lobe can effectively treat drug-resistant epilepsy but may lead to mood disorders. This fact is of particular interest in patients without a prior psychiatric history. The study investigates the relationship between Temporal Lobe Epilepsy (TLE), mood disorders, and the functional connectivity of the Hippocampus (Hipp) and Nucleus Accumbens (NAcc). In this case control study, twenty-seven TLE patients and 18 control subjects participated, undergoing structural and functional magnetic resonance imaging (MRI) scans before and after surgery. Post-surgery, patients were categorized into those developing de novo depression (DnD) within the first year and those without depression (nD). Functional connectivity maps between NAcc and the whole brain were generated, and connectivity strength between the to-be-resected Hipp area and NAcc was compared. Within the first year post-surgery, 7 out of 27 patients developed DnD. Most patients (88.8 %) exhibited a significant reduction in NAcc-Hipp connectivity compared to controls. The DnD group showed notably lower connectivity values than the nD group, with statistically significant disparities. Receiver Operating Characteristic (ROC) curve analysis identified a potential biomarker threshold (Crawford-T value of -2.08) with a sensitivity of 0.83 and specificity of 0.76. The results suggest that functional connectivity patterns within the reward network could serve as a potential biomarker for predicting de novo mood disorders in TLE patients undergoing surgery. This insight may assist in identifying individuals at a higher risk of developing DnD after surgery, enhancing therapeutic guidance and clinical decision-making.

Region-specific MRI predictors of surgical outcome in temporal lobe epilepsy

OBJECTIVE

In drug-resistant temporal lobe epilepsy (TLE), it is not well-established in how far surgery should target morphological anomalies to achieve seizure freedom. Here, we assessed interactions between structural brain compromise and surgery to identify region-specific predictors of seizure outcome.

METHODS

We obtained pre- and post-operative 3D T1-weighted MRI in 55 TLE patients who underwent selective amygdalo-hippocampectomy (SAH) or anterior temporal lobectomy (ATL) and 40 age and sex-matched healthy subjects. We measured surface-based morphological alterations of the mesiotemporal lobe structures (hippocampus, amygdala, entorhinal and piriform cortices), the neocortex and the thalamus on both pre- and post-operative MRI. Using precise co-registration, in each patient we mapped the surgical cavity onto the MRI acquired before surgery, thereby quantifying the amount of pathological tissue resected; these features, together with the preoperative morphometric data, served as input to a supervised classification algorithm for postsurgical outcome prediction.

RESULTS

On pre-operative MRI, patients who became seizure-free (TLE-SF) presented with severe ipsilateral amygdalar and hippocampal atrophy, while not seizure-free patients (TLE-NSF) displayed amygdalar hypertrophy. Stratifying patients based on the surgical approach, post-operative MRI showed similar patterns of mesiotemporal and thalamic changes, but divergent neocortical thinning affecting the parieto-temporo-occipital regions following ATL and the frontal lobes after SAH. Irrespective of the surgical approach, hippocampal atrophy on pre-operative MRI and its extent of resection were the most predictive features of seizure-freedom in 89% of patients (selected 100% across validations).

SIGNIFICANCE

Our study indicates a critical role of the extent of resection of MRI-derived hippocampal morphological anomalies on seizure outcome. Precise pre-operative quantification of the mesiotemporal lobe provides non-invasive prognostics for individualized surgery.

MRI-Derived Modeling of Disease Progression Patterns in Patients With Temporal Lobe Epilepsy

BACKGROUND AND OBJECTIVES

Temporal lobe epilepsy (TLE) is assumed to follow a steady course that is similar across patients. To date, phenotypic and temporal diversities of TLE evolution remain unknown. In this study, we aimed at simultaneously characterizing these sources of variability based on cross-sectional data.

METHODS

We studied consecutive patients with TLE referred for evaluation by neurologists to the Montreal Neurological Institute epilepsy clinic, who underwent in-patient video EEG monitoring and multimodal imaging at 3 Tesla, comprising 3D T1 and fluid-attenuated inversion recovery and 2D diffusion-weighted MRI. The cohort included patients with drug-resistant epilepsy and patients with drug-responsive epilepsy. The neuropsychological evaluation included Wechsler Adult Intelligence Scale-III and Leonard tapping task. The control group consisted of participants without TLE recruited through advertisement and who underwent the same MRI acquisition as patients. Based on surface-based analysis of key MRI markers of pathology (gray matter morphology and white matter microstructure), the Subtype and Stage Inference algorithm estimated subtypes and stages of brain pathology to which individual patients were assigned. The number of subtypes was determined by running the algorithm 100 times and estimating mean and SD of disease trajectories and the consistency of patients' assignments based on 1,000 bootstrap samples. Effect of normal aging was subtracted from patients. We examined associations with clinical and cognitive parameters and utility for individualized predictions.

RESULTS

We studied 82 patients with TLE (52 female, mean age 35 ± 10 years; 11 drug-responsive) and 41 control participants (23 male, mean age 32 ± 8 years). Among 57 operated, 43/37/20 had Engel-I outcome/hippocampal sclerosis/hippocampal isolated gliosis, respectively. We identified 3 trajectory subtypes: S1 (n = 35), led by ipsilateral hippocampal atrophy and gliosis, followed by white-matter damage; S2 (n = 27), characterized by bilateral neocortical atrophy, followed by ipsilateral hippocampal atrophy and gliosis; and S3 (n = 20), typified by bilateral limbic white-matter damage, followed by bilateral hippocampal gliosis. Patients showed high assignability to their subtypes and stages (>90% bootstrap agreement). S1 had the highest proportions of patients with early disease onset (effect size d = 0.27 vs S2, d = 0.73 vs S3), febrile convulsions (χ2 = 3.70), drug resistance (χ2 = 2.94), a positive MRI (χ2 = 8.42), hippocampal sclerosis (χ2 = 7.57), and Engel-I outcome (χ2 = 1.51), pFDR < 0.05 across all comparisons. S2 and S3 exhibited the intermediate and lowest proportions, respectively. Verbal IQ and digit span were lower in S1 (d = 0.65 and d = 0.50, pFDR < 0.05) and S2 (d = 0.76 and d = 1.09, pFDR < 0.05), compared with S3. We observed progressive decline in sequential motor tapping in S1 and S3 (T = -3.38 and T = -4.94, pFDR = 0.027), compared with S2 (T = 2.14, pFDR = 0.035). S3 showed progressive decline in digit span (T = -5.83, p = 0.021). Supervised classifiers trained on subtype and stage outperformed subtype-only and stage-only models predicting drug response in 73% ± 1.0% (vs 70% ± 1.4% and 63% ± 1.3%) and 76% ± 1.6% for Engel-I outcome (vs 71% ± 0.8% and 72% ± 1.1%), pFDR < 0.05 across all comparisons.

DISCUSSION

Cross-sectional MRI-derived models provide reliable prognostic markers of TLE disease evolution, which follows distinct trajectories, each associated with divergent patterns of hippocampal and whole-brain structural alterations, as well as cognitive and clinical profiles.

Decomposed FDG PET-based phenotypic heterogeneity predicting clinical prognosis and decision-making in temporal lobe epilepsy patients

OBJECTIVE

This study utilized a data-driven Bayesian model to automatically identify distinct latent disease factors represented by overlapping glucose metabolism patterns from 18F-Fluorodeoxyglucose PET (18F-FDG PET) to analyze heterogeneity among patients with TLE.

METHODS

We employed unsupervised machine learning to estimate latent disease factors from 18F-FDG PET scans, representing whole-brain glucose metabolism patterns in seventy patients with TLE. We estimated the extent to which multiple distinct factors were expressed within each participant and analyzed their relevance to epilepsy burden, including seizure onset, duration, and frequency. Additionally, we established a predictive model for clinical prognosis and decision-making.

RESULTS

We identified three latent disease factors: hypometabolism in the unilateral temporal lobe and hippocampus (factor 1), hypometabolism in bilateral prefrontal lobes (factor 2), and hypometabolism in bilateral temporal lobes (factor 3), variably co-expressed within each patient. Factor 3 demonstrated the strongest negative correlation with the age of onset and duration (r =  - 0.33, - 0.38 respectively, P < 0.05). The supervised classifier, trained on latent disease factors for predicting patient-specific antiepileptic drug (AED) responses, achieved an area under the curve (AUC) of 0.655. For post-surgical seizure outcomes, the AUC was 0.857, and for clinical decision-making, it was 0.965.

CONCLUSIONS

Decomposing 18F-FDG PET-based phenotypic heterogeneity facilitates individual-level predictions relevant to disease monitoring and personalized therapeutic strategies.

The network is more important than the node: stereo-EEG evidence of neurocognitive networks in epilepsy

Introduction

Neuropsychological assessment forms an integral part of the presurgical evaluation for patients with medically refractory focal epilepsy. Our understanding of cognitive impairment in epilepsy is based on seminal lesional studies that have demonstrated important structure-function relationships within the brain. However, a growing body of literature demonstrating heterogeneity in the cognitive profiles of patients with focal epilepsy (e.g., temporal lobe epilepsy; TLE) has led researchers to speculate that cognition may be impacted by regions outside the seizure onset zone, such as those involved in the interictal or "irritative" network.

Methods

Neuropsychological data from 48 patients who underwent stereoelectroencephalography (SEEG) monitoring between 2012 and 2023 were reviewed. Patients were categorized based on the site of seizure onset, as well as their irritative network, to determine the impact of wider network activity on cognition. Neuropsychological data were compared with normative standards (i.e., z = 0), and between groups.

Results

There were very few distinguishing cognitive features between patients when categorized based purely on the seizure onset zone (i.e., frontal lobe vs. temporal lobe epilepsy). In contrast, patients with localized irritative networks (i.e., frontal or temporal interictal epileptiform discharges [IEDs]) demonstrated more circumscribed profiles of impairment compared with those demonstrating wider irritative networks (i.e., frontotemporal IEDs). Furthermore, the directionality of propagation within the irritative network was found to influence the manifestations of cognitive impairment.

Discussion

The findings suggest that neuropsychological assessment is sensitive to network activity beyond the site of seizure onset. As such, an overly focal interpretation may not accurately reflect the distribution of the underlying pathology. This has important implications for presurgical work-up in epilepsy, as well as subsequent surgical outcomes.

Connectome reorganization associated with temporal lobe pathology and its surgical resection

Network neuroscience offers a unique framework to understand the organizational principles of the human brain. Despite recent progress, our understanding of how the brain is modulated by focal lesions remains incomplete. Resection of the temporal lobe is the most effective treatment to control seizures in pharmaco-resistant temporal lobe epilepsy (TLE), making this syndrome a powerful model to study lesional effects on network organization in young and middle-aged adults. Here, we assessed the downstream consequences of a focal lesion and its surgical resection on the brain's structural connectome, and explored how this reorganization relates to clinical variables at the individual patient level. We included adults with pharmaco-resistant TLE (n = 37) who underwent anterior temporal lobectomy between two imaging time points, as well as age- and sex-matched healthy controls who underwent comparable imaging (n = 31). Core to our analysis was the projection of high-dimensional structural connectome data-derived from diffusion MRI tractography from each subject-into lower-dimensional gradients. We then compared connectome gradients in patients relative to controls before surgery, tracked surgically-induced connectome reconfiguration from pre- to postoperative time points, and examined associations to patient-specific clinical and imaging phenotypes. Before surgery, individuals with TLE presented with marked connectome changes in bilateral temporo-parietal regions, reflecting an increased segregation of the ipsilateral anterior temporal lobe from the rest of the brain. Surgery-induced connectome reorganization was localized to this temporo-parietal subnetwork, but primarily involved postoperative integration of contralateral regions with the rest of the brain. Using a partial least-squares analysis, we uncovered a latent clinical imaging signature underlying this pre- to postoperative connectome reorganization, showing that patients who displayed postoperative integration in bilateral fronto-occipital cortices also had greater preoperative ipsilateral hippocampal atrophy, lower seizure frequency and secondarily generalized seizures. Our results bridge the effects of focal brain lesions and their surgical resections with large-scale network reorganization and interindividual clinical variability, thus offering new avenues to examine the fundamental malleability of the human brain.

Artificial intelligence in epilepsy - applications and pathways to the clinic

Artificial intelligence (AI) is rapidly transforming health care, and its applications in epilepsy have increased exponentially over the past decade. Integration of AI into epilepsy management promises to revolutionize the diagnosis and treatment of this complex disorder. However, translation of AI into neurology clinical practice has not yet been successful, emphasizing the need to consider progress to date and assess challenges and limitations of AI. In this Review, we provide an overview of AI applications that have been developed in epilepsy using a variety of data modalities: neuroimaging, electroencephalography, electronic health records, medical devices and multimodal data integration. For each, we consider potential applications, including seizure detection and prediction, seizure lateralization, localization of the seizure-onset zone and assessment for surgical or neurostimulation interventions, and review the performance of AI tools developed to date. We also discuss methodological considerations and challenges that must be addressed to successfully integrate AI into clinical practice. Our goal is to provide an overview of the current state of the field and provide guidance for leveraging AI in future to improve management of epilepsy.

Cortical Thickness Patterns of Cognitive Impairment Phenotypes in Drug-Resistant Temporal Lobe Epilepsy

OBJECTIVE

In temporal lobe epilepsy (TLE), a taxonomy classifying patients into 3 cognitive phenotypes has been adopted: minimally, focally, or multidomain cognitively impaired (CI). We examined gray matter (GM) thickness patterns of cognitive phenotypes in drug-resistant TLE and assessed potential use for predicting postsurgical cognitive outcomes.

METHODS

TLE patients undergoing presurgical evaluation were categorized into cognitive phenotypes. Network edge weights and distances were calculated using type III analysis of variance F-statistics from comparisons of GM regions within each TLE cognitive phenotype and age- and sex-matched healthy participants. In resected patients, logistic regression models (LRMs) based on network analysis results were used for prediction of postsurgical cognitive outcome.

RESULTS

A total of 124 patients (63 females, mean age ± standard deviation [SD] = 36.0 ± 12.0 years) and 117 healthy controls (63 females, mean age ± SD = 36.1 ± 12.0 years) were analyzed. In the multidomain CI group (n = 66, 53.2%), 28 GM regions were significantly thinner compared to healthy controls. Focally impaired patients (n = 37, 29.8%) showed 13 regions, whereas minimally impaired patients (n = 21, 16.9%) had 2 significantly thinner GM regions. Regions affected in both multidomain and focally impaired patients included the anterior cingulate cortex, medial prefrontal cortex, medial temporal, and lateral temporal regions. In 69 (35 females, mean age ± SD = 33.6 ± 18.0 years) patients who underwent surgery, LRMs based on network-identified GM regions predicted postsurgical verbal memory worsening with a receiver operating curve area under the curve of 0.70 ± 0.15.

INTERPRETATION

A differential pattern of GM thickness can be found across different cognitive phenotypes in TLE. Including magnetic resonance imaging with clinical measures associated with cognitive profiles has potential in predicting postsurgical cognitive outcomes in drug-resistant TLE. ANN NEUROL 2024;95:984-997.

The use of automated and AI-driven algorithms for the detection of hippocampal sclerosis and focal cortical dysplasia

In drug-resistant epilepsy, magnetic resonance imaging (MRI) plays a central role in detecting lesions as it offers unmatched spatial resolution and whole-brain coverage. In addition, the last decade has witnessed continued developments in MRI-based computer-aided machine-learning techniques for improved diagnosis and prognosis. In this review, we focus on automated algorithms for the detection of hippocampal sclerosis and focal cortical dysplasia, particularly in cases deemed as MRI negative, with an emphasis on studies with histologically validated data. In addition, we discuss imaging-derived prognostic markers, including response to anti-seizure medication, post-surgical seizure outcome, and cognitive reserves. We also highlight the advantages and limitations of these approaches and discuss future directions toward person-centered care.

Machine learning applied to epilepsy: bibliometric and visual analysis from 2004 to 2023

Background

Epilepsy is one of the most common serious chronic neurological disorders, which can have a serious negative impact on individuals, families and society, and even death. With the increasing application of machine learning techniques in medicine in recent years, the integration of machine learning with epilepsy has received close attention, and machine learning has the potential to provide reliable and optimal performance for clinical diagnosis, prediction, and precision medicine in epilepsy through the use of various types of mathematical algorithms, and promises to make better parallel advances. However, no bibliometric assessment has been conducted to evaluate the scientific progress in this area. Therefore, this study aims to visually analyze the trend of the current state of research related to the application of machine learning in epilepsy through bibliometrics and visualization.

Methods

Relevant articles and reviews were searched for 2004-2023 using Web of Science Core Collection database, and bibliometric analyses and visualizations were performed in VOSviewer, CiteSpace, and Bibliometrix (R-Tool of R-Studio).

Results

A total of 1,284 papers related to machine learning in epilepsy were retrieved from the Wo SCC database. The number of papers shows an increasing trend year by year. These papers were mainly from 1,957 organizations in 87 countries/regions, with the majority from the United States and China. The journal with the highest number of published papers is EPILEPSIA. Acharya, U. Rajendra (Ngee Ann Polytechnic, Singapore) is the authoritative author in the field and his paper "Deep Convolutional Neural Networks for Automated Detection and Diagnosis of Epileptic Seizures Using EEG Signals" was the most cited. Literature and keyword analysis shows that seizure prediction, epilepsy management and epilepsy neuroimaging are current research hotspots and developments.

Conclusions

This study is the first to use bibliometric methods to visualize and analyze research in areas related to the application of machine learning in epilepsy, revealing research trends and frontiers in the field. This information will provide a useful reference for epilepsy researchers focusing on machine learning.

Mapping hippocampal and thalamic atrophy in epilepsy: A 7-T magnetic resonance imaging study

OBJECTIVE

Epilepsy patients are often grouped together by clinical variables. Quantitative neuroimaging metrics can provide a data-driven alternative for grouping of patients. In this work, we leverage ultra-high-field 7-T structural magnetic resonance imaging (MRI) to characterize volumetric atrophy patterns across hippocampal subfields and thalamic nuclei in drug-resistant focal epilepsy.

METHODS

Forty-two drug-resistant epilepsy patients and 13 controls with 7-T structural neuroimaging were included in this study. We measured hippocampal subfield and thalamic nuclei volumetry, and applied an unsupervised machine learning algorithm, Latent Dirichlet Allocation (LDA), to estimate atrophy patterns across the hippocampal subfields and thalamic nuclei of patients. We studied the association between predefined clinical groups and the estimated atrophy patterns. Additionally, we used hierarchical clustering on the LDA factors to group patients in a data-driven approach.

RESULTS

In patients with mesial temporal sclerosis (MTS), we found a significant decrease in volume across all ipsilateral hippocampal subfields (false discovery rate-corrected p [pFDR] < .01) as well as in some ipsilateral (pFDR < .05) and contralateral (pFDR < .01) thalamic nuclei. In left temporal lobe epilepsy (L-TLE) we saw ipsilateral hippocampal and some bilateral thalamic atrophy (pFDR < .05), whereas in right temporal lobe epilepsy (R-TLE) extensive bilateral hippocampal and thalamic atrophy was observed (pFDR < .05). Atrophy factors demonstrated that our MTS cohort had two atrophy phenotypes: one that affected the ipsilateral hippocampus and one that affected the ipsilateral hippocampus and bilateral anterior thalamus. Atrophy factors demonstrated posterior thalamic atrophy in R-TLE, whereas an anterior thalamic atrophy pattern was more common in L-TLE. Finally, hierarchical clustering of atrophy patterns recapitulated clusters with homogeneous clinical properties.

SIGNIFICANCE

Leveraging 7-T MRI, we demonstrate widespread hippocampal and thalamic atrophy in epilepsy. Through unsupervised machine learning, we demonstrate patterns of volumetric atrophy that vary depending on disease subtype. Incorporating these atrophy patterns into clinical practice could help better stratify patients to surgical treatments and specific device implantation strategies.

Exploring individual fixel-based white matter abnormalities in epilepsy

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.

Machine Learning and Artificial Intelligence Applications to Epilepsy: a Review for the Practicing Epileptologist

PURPOSE OF REVIEW

Machine Learning (ML) and Artificial Intelligence (AI) are data-driven techniques to translate raw data into applicable and interpretable insights that can assist in clinical decision making. Some of these tools have extremely promising initial results, earning both great excitement and creating hype. This non-technical article reviews recent developments in ML/AI in epilepsy to assist the current practicing epileptologist in understanding both the benefits and limitations of integrating ML/AI tools into their clinical practice.

RECENT FINDINGS

ML/AI tools have been developed to assist clinicians in almost every clinical decision including (1) predicting future epilepsy in people at risk, (2) detecting and monitoring for seizures, (3) differentiating epilepsy from mimics, (4) using data to improve neuroanatomic localization and lateralization, and (5) tracking and predicting response to medical and surgical treatments. We also discuss practical, ethical, and equity considerations in the development and application of ML/AI tools including chatbots based on Large Language Models (e.g., ChatGPT). ML/AI tools will change how clinical medicine is practiced, but, with rare exceptions, the transferability to other centers, effectiveness, and safety of these approaches have not yet been established rigorously. In the future, ML/AI will not replace epileptologists, but epileptologists with ML/AI will replace epileptologists without ML/AI.

Identification of different MRI atrophy progression trajectories in epilepsy by subtype and stage inference

Artificial intelligence (AI)-based tools are widely employed, but their use for diagnosis and prognosis of neurological disorders is still evolving. Here we analyse a cross-sectional multicentre structural MRI dataset of 696 people with epilepsy and 118 control subjects. We use an innovative machine-learning algorithm, Subtype and Stage Inference, to develop a novel data-driven disease taxonomy, whereby epilepsy subtypes correspond to distinct patterns of spatiotemporal progression of brain atrophy.In a discovery cohort of 814 individuals, we identify two subtypes common to focal and idiopathic generalized epilepsies, characterized by progression of grey matter atrophy driven by the cortex or the basal ganglia. A third subtype, only detected in focal epilepsies, was characterized by hippocampal atrophy. We corroborate external validity via an independent cohort of 254 people and confirm that the basal ganglia subtype is associated with the most severe epilepsy.Our findings suggest fundamental processes underlying the progression of epilepsy-related brain atrophy. We deliver a novel MRI- and AI-guided epilepsy taxonomy, which could be used for individualized prognostics and targeted therapeutics.

Parsing altered gray matter morphology of depression using a framework integrating the normative model and non-negative matrix factorization

The high inter-individual heterogeneity in individuals with depression limits neuroimaging studies with case-control approaches to identify promising biomarkers for individualized clinical decision-making. We put forward a framework integrating the normative model and non-negative matrix factorization (NMF) to quantitatively assess altered gray matter morphology in depression from a dimensional perspective. The proposed framework parses altered gray matter morphology into overlapping latent disease factors, and assigns patients distinct factor compositions, thus preserving inter-individual variability. We identified four robust disease factors with distinct clinical symptoms and cognitive processes in depression. In addition, we showed the quantitative relationship between the group-level gray matter morphological differences and disease factors. Furthermore, this framework significantly predicted factor compositions of patients in an independent dataset. The framework provides an approach to resolve neuroanatomical heterogeneity in depression.

Mapping the neuroanatomical heterogeneity of OCD using a framework integrating normative model and non-negative matrix factorization

Obsessive-compulsive disorder (OCD) is a spectrum disorder with high interindividual heterogeneity. We propose a comprehensible framework integrating normative model and non-negative matrix factorization (NMF) to quantitatively estimate the neuroanatomical heterogeneity of OCD from a dimensional perspective. T1-weighted magnetic resonance images of 98 first-episode untreated patients with OCD and matched healthy controls (HCs, n = 130) were acquired. We derived individualized differences in gray matter morphometry using normative model and parsed them into latent disease factors using NMF. Four robust disease factors were identified. Each patient expressed multiple factors and exhibited a unique factor composition. Factor compositions of patients were significantly correlated with severity of symptom, age of onset, illness duration, and exhibited sex differences, capturing sources of clinical heterogeneity. In addition, the group-level morphological differences obtained with two-sample t test could be quantitatively derived from the identified disease factors, reconciling the group-level and subject-level findings in neuroimaging studies. Finally, we uncovered two distinct subtypes with opposite morphological differences compared with HCs from factor compositions. Our findings suggest that morphological differences of individuals with OCD are the unique combination of distinct neuroanatomical patterns. The proposed framework quantitatively estimating neuroanatomical heterogeneity paves the way for precision medicine in OCD.

The Role of Molecular Imaging in Personalized Medicine

The concept of personalized medicine refers to the tailoring of medical treatment to each patient's unique characteristics. Scientific advancements have led to a better understanding of how a person's unique molecular and genetic profile makes them susceptible to certain diseases. It provides individualized medical treatments that will be safe and effective for each patient. Molecular imaging modalities play an essential role in this aspect. They are used widely in screening, detection and diagnosis, treatment, assessing disease heterogeneity and progression planning, molecular characteristics, and long-term follow-up. In contrast to conventional imaging approaches, molecular imaging techniques approach images as the knowledge that can be processed, allowing for the collection of relevant knowledge in addition to the evaluation of enormous patient groups. This review presents the fundamental role of molecular imaging modalities in personalized medicine.

The relationship between the interictal epileptiform discharge source connectivity and cortical structural couplings in temporal lobe epilepsy

Objective

The objective of this study was to explore the relation between interictal epileptiform discharge (IED) source connectivity and cortical structural couplings (SCs) in temporal lobe epilepsy (TLE).

Methods

High-resolution 3D-MRI and 32-sensor EEG data from 59 patients with TLE were collected. Principal component analysis was performed on the morphological data on MRI to obtain the cortical SCs. IEDs were labeled from EEG data and averaged. The standard low-resolution electromagnetic tomography analysis was performed to locate the source of the average IEDs. Phase-locked value was used to evaluate the IED source connectivity. Finally, correlation analysis was used to compare the IED source connectivity and the cortical SCs.

Results

The features of the cortical morphology in left and right TLE were similar across four cortical SCs, which could be mainly described as the default mode network, limbic regions, connections bilateral medial temporal, and connections through the ipsilateral insula. The IED source connectivity at the regions of interest was negatively correlated with the corresponding cortical SCs.

Significance

The cortical SCs were confirmed to be negatively related to IED source connectivity in patients with TLE as detected with MRI and EEG coregistered data. These findings suggest the important role of intervening IEDs in treating TLE.

Ontology-based feature engineering in machine learning workflows for heterogeneous epilepsy patient records

Biomedical ontologies are widely used to harmonize heterogeneous data and integrate large volumes of clinical data from multiple sources. This study analyzed the utility of ontologies beyond their traditional roles, that is, in addressing a challenging and currently underserved field of feature engineering in machine learning workflows. Machine learning workflows are being increasingly used to analyze medical records with heterogeneous phenotypic, genotypic, and related medical terms to improve patient care. We performed a retrospective study using neuropathology reports from the German Neuropathology Reference Center for Epilepsy Surgery at Erlangen, Germany. This cohort included 312 patients who underwent epilepsy surgery and were labeled with one or more diagnoses, including dual pathology, hippocampal sclerosis, malformation of cortical dysplasia, tumor, encephalitis, and gliosis. We modeled the diagnosis terms together with their microscopy, immunohistochemistry, anatomy, etiologies, and imaging findings using the description logic-based Web Ontology Language (OWL) in the Epilepsy and Seizure Ontology (EpSO). Three tree-based machine learning models were used to classify the neuropathology reports into one or more diagnosis classes with and without ontology-based feature engineering. We used five-fold cross validation to avoid overfitting with a fixed number of repetitions while leaving out one subset of data for testing, and we used recall, balanced accuracy, and hamming loss as performance metrics for the multi-label classification task. The epilepsy ontology-based feature engineering approach improved the performance of all the three learning models with an improvement of 35.7%, 54.5%, and 33.3% in logistics regression, random forest, and gradient tree boosting models respectively. The run time performance of all three models improved significantly with ontology-based feature engineering with gradient tree boosting model showing a 93.8% reduction in the time required for training and testing of the model. Although, all three models showed an overall improved performance across the three-performance metrics using ontology-based feature engineering, the rate of improvement was not consistent across all input features. To analyze this variation in performance, we computed feature importance scores and found that microscopy had the highest importance score across the three models, followed by imaging, immunohistochemistry, and anatomy in a decreasing order of importance scores. This study showed that ontologies have an important role in feature engineering to make heterogeneous clinical data accessible to machine learning models and also improve the performance of machine learning models in multilabel multiclass classification tasks.

Interpretable deep learning-based hippocampal sclerosis classification

OBJECTIVE

To evaluate the performance of a deep learning model for hippocampal sclerosis classification on the clinical dataset and suggest plausible visual interpretation for the model prediction.

METHODS

T2-weighted oblique coronal images of the brain MRI epilepsy protocol performed on patients were used. The training set included 320 participants with 160 no, 100 left and 60 right hippocampal sclerosis, and cross-validation was implemented. The test set consisted of 302 participants with 252 no, 25 left and 25 right hippocampal sclerosis. As the test set was imbalanced, we took an average of the accuracy achieved within each group to measure a balanced accuracy for multiclass and binary classifications. The dataset was composed to include not only healthy participants but also participants with abnormalities besides hippocampal sclerosis in the control group. We visualized the reasons for the model prediction using the layer-wise relevance propagation method.

RESULTS

When evaluated on the validation of the training set, we achieved multiclass and binary classification accuracy of 87.5% and 88.8% from the voting ensemble of six models. Evaluated on the test sets, we achieved multiclass and binary classification accuracy of 91.5% and 89.76%. The distinctly sparse visual interpretations were provided for each individual participant and group to suggest the contribution of each input voxel to the prediction on the MRI.

SIGNIFICANCE

The current interpretable deep learning-based model is promising for adapting effectively to clinical settings by utilizing commonly used data, such as MRI, with realistic abnormalities faced by neurologists to support the diagnosis of hippocampal sclerosis with plausible visual interpretation.

Volumetric and structural connectivity abnormalities co-localise in TLE

Patients with temporal lobe epilepsy (TLE) exhibit both volumetric and structural connectivity abnormalities relative to healthy controls. How these abnormalities inter-relate and their mechanisms are unclear. We computed grey matter volumetric changes and white matter structural connectivity abnormalities in 144 patients with unilateral TLE and 96 healthy controls. Regional volumes were calculated using T1-weighted MRI, while structural connectivity was derived using white matter fibre tractography from diffusion-weighted MRI. For each regional volume and each connection strength, we calculated the effect size between patient and control groups in a group-level analysis. We then applied hierarchical regression to investigate the relationship between volumetric and structural connectivity abnormalities in individuals. Additionally, we quantified whether abnormalities co-localised within individual patients by computing Dice similarity scores. In TLE, white matter connectivity abnormalities were greater when joining two grey matter regions with abnormal volumes. Similarly, grey matter volumetric abnormalities were greater when joined by abnormal white matter connections. The extent of volumetric and connectivity abnormalities related to epilepsy duration, but co-localisation did not. Co-localisation was primarily driven by neighbouring abnormalities in the ipsilateral hemisphere. Overall, volumetric and structural connectivity abnormalities were related in TLE. Our results suggest that shared mechanisms may underlie changes in both volume and connectivity alterations in patients with TLE.