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

Connectome architecture for gray matter atrophy and surgical outcomes in temporal lobe epilepsy

OBJECTIVE

Temporal lobe epilepsy (TLE) has been recognized as a network disorder with widespread gray matter atrophy. However, the role of connectome architecture in shaping morphological alterations and identifying atrophy epicenters remains unclear. Furthermore, individualized modeling of atrophy epicenters and their potential clinical applications have not been well established. This study aims to explore how gray matter atrophy correlates with normal connectome architecture, identify potential atrophy epicenters, and employ individualized modeling approach to evaluate the impact of different epicenter patterns on surgical outcomes in patients with TLE.

METHODS

This study utilized anatomic MRI data from 126 refractory TLE patients who underwent anterior temporal lobectomy and 60 healthy controls (HCs), along with normative functional and structural connectome data, to investigate the relationship between gray matter volume (GMV) changes and functional or structural connectivity. Two models were employed to identify atrophy epicenters: a data-driven approach evaluating nodal and neighbor atrophy rankings, and a network diffusion model (NDM) simulating the spread of pathology from different seed regions. K-means clustering was applied in patient-tailored modeling to uncover distinct epicenter subtypes.

RESULTS

Our findings indicate that the pattern of gray matter atrophy in TLE is constrained primarily by structural connectivity rather than by functional connectivity. Using the structural connectome, we pinpointed the hippocampus and adjacent temporo-limbic regions as key atrophy epicenters. The patient-tailored modeling revealed significant variability in epicenter distribution, allowing us to categorize them into two distinct subtypes. Notably, patients in subtype 2, with epicenters localized to the ipsilateral temporal pole and medial temporal lobe, exhibited significantly higher seizure-free rates compared to patients in subtype 1, whose epicenters situated in frontocentral regions.

SIGNIFICANCE

These findings highlight the central role of structural connectivity in shaping TLE-related morphological changes. Individualized epicenter modeling may enhance surgical decisions and improve prognostic stratification in TLE management.

Normative structural connectome constrains spreading transient brain activity in generalized epilepsy

BACKGROUND

Genetic generalized epilepsy is characterized by transient episodes of spontaneous abnormal neural activity in anatomically distributed brain regions that ultimately propagate to wider areas. However, the connectome-based mechanisms shaping these abnormalities remain largely unknown. We aimed to investigate how the normative structural connectome constrains abnormal brain activity spread in genetic generalized epilepsy with generalized tonic-clonic seizure (GGE-GTCS).

METHODS

Abnormal transient activity patterns between individuals with GGE-GTCS (n = 97) and healthy controls (n = 141) were estimated from the amplitude of low-frequency fluctuations measured by resting-state functional MRI. The normative structural connectome was derived from diffusion-weighted images acquired in an independent cohort of healthy adults (n = 326). Structural neighborhood analysis was applied to assess the degree of constraints between activity vulnerability and structural connectome. Dominance analysis was used to determine the potential molecular underpinnings of these constraints. Furthermore, a network-based diffusion model was utilized to simulate the spread of pathology and identify potential disease epicenters.

RESULTS

Brain activity abnormalities among patients with GGE-GTCS were primarily located in the temporal, cingulate, prefrontal, and parietal cortices. The collective abnormality of structurally connected neighbors significantly predicted regional activity abnormality, indicating that white matter network architecture constrains aberrant activity patterns. Molecular fingerprints, particularly laminar differentiation and neurotransmitter receptor profiles, constituted key predictors of these connectome-constrained activity abnormalities. Network-based diffusion modeling effectively replicated transient pathological activity spreading patterns, identifying the limbic-temporal, dorsolateral prefrontal, and occipital cortices as putative disease epicenters. These results were robust across different clinical factors and individual patients.

CONCLUSIONS

Our findings suggest that the structural connectome shapes the spatial patterning of brain activity abnormalities, advancing our understanding of the network-level mechanisms underlying vulnerability to abnormal brain activity onset and propagation in GGE-GTCS.

Do Seizures Damage the Brain?-Cumulative Effects of Seizures and Epilepsy: A 2025 Perspective

In 1885, William Gowers proposed that epilepsy is a progressive disease, based on clinical evidence before any effective treatments were available. His long-standing hypothesis has been summarized with the statement "seizures beget seizures." Whether this is the case and related questions about seizure-induced modification and damage of brain circuits are of fundamental importance for neurobiological understanding of epilepsy, development of effective treatment strategies, clinical management, and prognostication. Consensus about progression and seizure-induced damage has remained controversial. Here, we critically review these long-standing questions, incorporating perspectives about perceived inconsistencies in past studies, potential implications of recent longitudinal imaging and cognitive studies, and emphasize experimental and clinical gaps that have proved challenging. Answers to these questions are important for development of management strategies to achieve prompt effective acute control of seizures and prevention of their potential recurrence and long-term comorbidities.

Machine learning reveals connections between preclinical type 2 diabetes subtypes and brain health

Previous research has established type 2 diabetes mellitus as a significant risk factor for various disorders, adversely impacting human health. While evidence increasingly links type 2 diabetes to cognitive impairment and brain disorders, understanding the causal effects of its preclinical stage on brain health is yet to be fully known. This knowledge gap hinders advancements in screening and preventing neurological and psychiatric diseases. To address this gap, we employed a robust machine learning algorithm (Subtype and Stage Inference, SuStaIn) with cross-sectional clinical data from the UK Biobank (20 277 preclinical type 2 diabetes participants and 20 277 controls) to identify underlying subtypes and stages for preclinical type 2 diabetes. Our analysis revealed one subtype distinguished by elevated circulating leptin levels and decreased leptin receptor levels, coupled with increased body mass index, diminished lipid metabolism, and heightened susceptibility to psychiatric conditions such as anxiety disorder, depression disorder, and bipolar disorder. Conversely, individuals in the second subtype manifested typical abnormalities in glucose metabolism, including rising glucose and haemoglobin A1c levels, with observed correlations with neurodegenerative disorders. A >10-year follow-up of these individuals revealed differential declines in brain health and significant clinical outcome disparities between subtypes. The first subtype exhibited faster progression and higher risk for psychiatric conditions, while the second subtype was associated with more severe progression of Alzheimer's disease and Parkinson's disease and faster progression to type 2 diabetes. Our findings highlight that monitoring and addressing the brain health needs of individuals in the preclinical stage of type 2 diabetes is imperative.

A generalized epilepsy network derived from brain abnormalities and deep brain stimulation

Idiopathic generalized epilepsy (IGE) is a brain network disease, but the location of this network and its relevance for treatment remain unclear. We combine the locations of brain abnormalities in IGE (131 coordinates from 21 studies) with the human connectome to identify an IGE network. We validate this network by showing alignment with structural brain abnormalities previously identified in IGE and brain areas activated by generalized epileptiform discharges in simultaneous electroencephalogram-functional magnetic resonance imaging. The topography of the IGE network aligns with brain networks involved in motor control and loss of consciousness consistent with generalized seizure semiology. To investigate therapeutic relevance, we analyze data from 21 patients with IGE treated with deep brain stimulation (DBS) for generalized seizures. Seizure frequency reduced a median 90% after DBS and stimulation sites intersect an IGE network peak in the centromedian nucleus of the thalamus. Together, this study helps unify prior findings in IGE and identify a brain network target that can be tested in clinical trials of brain stimulation to control generalized seizures.

Temporal Lobe Epilepsy Perturbs the Brain-Wide Excitation-Inhibition Balance: Associations with Microcircuit Organization, Clinical Parameters, and Cognitive Dysfunction

Excitation-inhibition (E/I) imbalance is theorized as a key mechanism in the pathophysiology of epilepsy, with ample research focusing on elucidating its cellular manifestations. However, few studies investigate E/I imbalance at the macroscale, whole-brain level, and its microcircuit-level mechanisms and clinical significance remain incompletely understood. Here, the Hurst exponent, an index of the E/I ratio, is computed from resting-state fMRI time series, and microcircuit parameters are simulated using biophysical models. A broad decrease in the Hurst exponent is observed in pharmaco-resistant temporal lobe epilepsy (TLE), suggesting more excitable network dynamics. Connectome decoders point to temporolimbic and frontocentral cortices as plausible network epicenters of E/I imbalance. Furthermore, computational simulations reveal that enhancing cortical excitability in TLE reflects atypical increases in recurrent connection strength of local neuronal ensembles. Mixed cross-sectional and longitudinal analyses show stronger E/I ratio elevation in patients with longer disease duration, more frequent electroclinical seizures as well as interictal epileptic spikes, and worse cognitive functioning. Hurst exponent-informed classifiers discriminate patients from healthy controls with high accuracy (72.4% [57.5%-82.5%]). Replicated in an independent dataset, this work provides in vivo evidence of a macroscale shift in E/I balance in TLE patients and points to progressive functional imbalances that relate to cognitive decline.

Longitudinal alterations in brain networks and thalamocortical connectivity in paediatric focal epilepsy: a structural connectomics pilot study

Epilepsy is an archetypal brain network disorder characterized by recurrent seizures and associated psychological, cognitive and behavioural sequelae. Progressive brain network dysfunction may contribute to poorer outcomes following treatment, but this has never been tested in humans. In this structural connectomics pilot study, we assess whether there is progressive brain network dysfunction in a cohort of 23 children undergoing repeated multi-shell diffusion tensor imaging as part of their pre-surgical evaluation of focal epilepsy prior to epilepsy surgery. We analyse global and nodal graph metrics and thalamocortical connectivity, comparing the longitudinal changes to a cross-sectional cohort of 57 healthy controls. We identify no robust longitudinal changes in global or nodal network properties over a median of 1.15 years between scans. We also do not identify robust longitudinal changes in thalamic connectivity between scans. On sensitivity analyses, we identify increases in weighted degree at higher scales of brain parcellation and a decrease in the proportion of nodes with a low participation coefficient, suggesting progressive increases in intermodular connections. These findings of no or subtle structural longitudinal brain network changes over a relatively short timeframe indicate that either there are no progressive structural brain network changes over time in epilepsy or the changes appear over longer timescales. Larger studies with longer timeframes between scans may help clarify these findings.

Preclinical Testing Strategies for Epilepsy Therapy Development

The development of antiepileptogenic and disease-modifying treatments for epilepsy is a key goal of epilepsy research. Technological and scientific advances over the past two decades have seen the development of numerous therapeutic approaches, many of which show great promise in animal models. To facilitate and de-risk the translation of these promising approaches, however, rigorous preclinical testing is needed. For the present review, we discuss challenges and approaches to conduct preclinical testing of antiepileptogenic and disease-modifying treatments in animal models.

Imaging blood-brain barrier dysfunction in drug-resistant epilepsy: A multi-center feasibility study

OBJECTIVE

Blood-brain barrier dysfunction (BBBD) has been linked to various neurological disorders, including epilepsy. This study aims to utilize dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) to identify and compare brain regions with BBBD in patients with epilepsy (PWE) and healthy individuals.

METHODS

We scanned 50 drug-resistant epilepsy (DRE) patients and 58 control participants from four global specialized epilepsy centers using DCE-MRI. The presence and extent of BBBD were analyzed and compared between PWE and healthy controls.

RESULTS

Both greater brain volume and higher number of brain regions with BBBD were significantly present in PWE compared to healthy controls (p < 10-7). No differences in total brain volume with BBBD were observed in patients diagnosed with either focal seizures or generalized epilepsy, despite variations in the affected regions. Overall brain volume with BBBD did not differ in PWE with MRI-visible lesions compared with non-lesional cases. BBBD was observed in brain regions suspected to be related to the onset of seizures in 82% of patients (n = 39) and was typically identified in, adjacent to, and/or in the same hemisphere as the suspected epileptogenic lesion (n = 10).

SIGNIFICANCE

These findings are consistent with pre-clinical studies that highlight the role of BBBD in the development of DRE and identify microvascular stabilization as a potential therapeutic strategy.

Brain Networks for Cortical Atrophy and Responsive Neurostimulation in Temporal Lobe Epilepsy

Importance

Drug-resistant temporal lobe epilepsy (TLE) has been associated with hippocampal pathology. Most surgical treatment strategies, including resection and responsive neurostimulation (RNS), focus on this disease epicenter; however, imaging alterations distant from the hippocampus, as well as emerging data from responsive neurostimulation trials, suggest conceptualizing TLE as a network disorder.

Objective

To assess whether brain networks connected to areas of atrophy in the hippocampus align with the topography of distant neuroimaging alterations and RNS response.

Design, Setting, and Participants

This retrospective case-control study was conducted between July 2009 and June 2022. Data collection for this multicenter, population-based study took place across 4 tertiary referral centers in Montréal, Canada; Querétaro, México; Nanjing, China; and Salt Lake City, Utah. Eligible patients were diagnosed with TLE according to International League Against Epilepsy criteria and received either neuroimaging or neuroimaging and RNS to the hippocampus. Patients with encephalitis, traumatic brain injury, or bilateral TLE were excluded.

Main Outcomes and Measures

Spatial alignment between brain network topographies.

Results

Of the 110 eligible patients, 94 individuals diagnosed with TLE were analyzed (51 [54%] female; mean [SD] age, 31.3 [10.9] years). Hippocampal thickness maps in TLE were compared to 120 healthy control individuals (66 [55%] female; mean [SD] age, 29.8 [9.5] years), and areas of atrophy were identified. Using an atlas of normative connectivity (n = 1000), 2 brain networks were identified that were functionally connected to areas of hippocampal atrophy. The first network was defined by positive correlations to temporolimbic, medial prefrontal, and parietal regions, whereas the second network by negative correlations to frontoparietal regions. White matter changes colocalized to the positive network (t93 = -3.82; P = 2.44 × 10-4). In contrast, cortical atrophy localized to the negative network (t93 = 3.54; P = 6.29 × 10-3). In an additional 38 patients (20 [53%] female; mean [SD] age, 35.8 [11.3] years) treated with RNS, connectivity between the stimulation site and atrophied regions within the negative network was associated with seizure reduction (t212 = -2.74; P = .007).

Conclusions and Relevance

The findings in this study indicate that distributed pathology in TLE may occur in brain networks connected to the hippocampal epicenter. Connectivity to these same networks was associated with improvement following RNS. A network approach to TLE may reveal therapeutic targets outside the traditional target in the hippocampus.

Estimated Disease Progression Trajectory of White Matter Disruption in Unilateral Temporal Lobe Epilepsy: A Data-Driven Machine Learning Approach

BACKGROUND/OBJECTIVES

Although the involvement of progressive brain alterations in epilepsy was recently suggested, individual patients' trajectories of white matter (WM) disruption are not known.

METHODS

We investigated the disease progression patterns of WM damage and its associations with clinical metrics. We examined the cross-sectional diffusion tensor imaging (DTI) data of 155 patients with unilateral temporal lobe epilepsy (TLE) and 270 age/gender-matched healthy controls, and we then calculated the average fractional anisotropy (FA) values within 20 WM tracts of the whole brain. We used the Subtype and Stage Inference (SuStaIn) program to detect the progression trajectory of FA changes and investigated its association with clinical parameters including onset age, disease duration, drug-responsiveness, and the number of anti-seizure medications (ASMs).

RESULTS

The SuStaIn algorithm identified a single subtype model in which the initial damage occurs in the ipsilateral uncinate fasciculus (UF), followed by damage in the forceps, superior longitudinal fasciculus (SLF), and anterior thalamic radiation (ATR). This pattern was replicated when analyzing TLE with hippocampal sclerosis (n = 50) and TLE with no lesions (n = 105) separately. Further-progressed stages were associated with longer disease duration (p < 0.001) and a greater number of ASMs (p = 0.001).

CONCLUSIONS

the disease progression model based on WM tracts may be useful as a novel individual-level biomarker.

Chronicles of Change: The Shrinking Brain in Epilepsy

Identification of Different MRI Atrophy Progression Trajectories in Epilepsy by Subtype and Stage Inference Xiao F, Caciagli L, Wandschneider B, Sone D, Young AL, Vos SB, Winston GP, Zhang Y, Liu W, An D, Kanber B, Zhou D, Sander JW, Thom M, Duncan JS, Alexander DC, Galovic M, Koepp MJ. Brain . 2023;146(11):4702-4716. doi:10.1093/brain/awad284 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.

Identification of four biotypes in temporal lobe epilepsy via machine learning on brain images

Artificial intelligence provides an opportunity to try to redefine disease subtypes based on similar pathobiology. Using a machine-learning algorithm (Subtype and Stage Inference) with cross-sectional MRI from 296 individuals with focal epilepsy originating from the temporal lobe (TLE) and 91 healthy controls, we show phenotypic heterogeneity in the pathophysiological progression of TLE. This study was registered in the Chinese Clinical Trials Registry (number: ChiCTR2200062562). We identify two hippocampus-predominant phenotypes, characterized by atrophy beginning in the left or right hippocampus; a third cortex-predominant phenotype, characterized by hippocampus atrophy after the neocortex; and a fourth phenotype without atrophy but amygdala enlargement. These four subtypes are replicated in the independent validation cohort (109 individuals). These subtypes show differences in neuroanatomical signature, disease progression and epilepsy characteristics. Five-year follow-up observations of these individuals reveal differential seizure outcomes among subtypes, indicating that specific subtypes may benefit from temporal surgery or pharmacological treatment. These findings suggest a diverse pathobiological basis underlying focal epilepsy that potentially yields to stratification and prognostication - a necessary step for precise medicine.

Disease Progression Patterns of Brain Morphology in Schizophrenia: More Progressed Stages in Treatment Resistance

BACKGROUND AND HYPOTHESIS

Given the heterogeneity and possible disease progression in schizophrenia, identifying the neurobiological subtypes and progression patterns in each patient may lead to novel biomarkers. Here, we adopted data-driven machine-learning techniques to identify the progression patterns of brain morphological changes in schizophrenia and investigate the association with treatment resistance.

STUDY DESIGN

In this cross-sectional multicenter study, we included 177 patients with schizophrenia, characterized by treatment response or resistance, with 3D T1-weighted magnetic resonance imaging. Cortical thickness and subcortical volumes calculated by FreeSurfer were converted into z scores using 73 healthy controls data. The Subtype and Stage Inference (SuStaIn) algorithm was used for unsupervised machine-learning analysis.

STUDY RESULTS

SuStaIn identified 3 different subtypes: (1) subcortical volume reduction (SC) type (73 patients), in which volume reduction of subcortical structures occurs first and moderate cortical thinning follows, (2) globus pallidus hypertrophy and cortical thinning (GP-CX) type (42 patients), in which globus pallidus hypertrophy initially occurs followed by progressive cortical thinning, and (3) cortical thinning (pure CX) type (39 patients), in which thinning of the insular and lateral temporal lobe cortices primarily happens. The remaining 23 patients were assigned to baseline stage of progression (no change). SuStaIn also found 84 stages of progression, and treatment-resistant schizophrenia showed significantly more progressed stages than treatment-responsive cases (P = .001). The GP-CX type presented earlier stages than the pure CX type (P = .009).

CONCLUSIONS

The brain morphological progressions in schizophrenia can be classified into 3 subtypes, and treatment resistance was associated with more progressed stages, which may suggest a novel biomarker.

Data-driven modelling of neurodegenerative disease progression: thinking outside the black box

Data-driven disease progression models are an emerging set of computational tools that reconstruct disease timelines for long-term chronic diseases, providing unique insights into disease processes and their underlying mechanisms. Such methods combine a priori human knowledge and assumptions with large-scale data processing and parameter estimation to infer long-term disease trajectories from short-term data. In contrast to 'black box' machine learning tools, data-driven disease progression models typically require fewer data and are inherently interpretable, thereby aiding disease understanding in addition to enabling classification, prediction and stratification. In this Review, we place the current landscape of data-driven disease progression models in a general framework and discuss their enhanced utility for constructing a disease timeline compared with wider machine learning tools that construct static disease profiles. We review the insights they have enabled across multiple neurodegenerative diseases, notably Alzheimer disease, for applications such as determining temporal trajectories of disease biomarkers, testing hypotheses about disease mechanisms and uncovering disease subtypes. We outline key areas for technological development and translation to a broader range of neuroscience and non-neuroscience applications. Finally, we discuss potential pathways and barriers to integrating disease progression models into clinical practice and trial settings.

Out of one, how many? Subtyping in epilepsy

This scientific commentary refers to ‘Identification of different MRI atrophy progression trajectories in epilepsy by subtype and stage inference’ by Xiao et al. (https://doi.org/10.1093/brain/awad284).