In recent years, artificial intelligence (AI) has become an increasingly prominent focus of medical research, significantly impacting epileptology as well. Studies on deep learning (DL) and machine learning (ML) - the core of AI - have explored their applications in epilepsy imaging, primarily focusing on lesion detection, lateralization and localization of epileptogenic areas, postsurgical outcome prediction and automatic differentiation between people with epilepsy and healthy individuals. Various AI-driven approaches are being investigated across different neuroimaging modalities, with the ultimate goal of integrating these tools into clinical practice to enhance the diagnosis and treatment of epilepsy. As computing power continues to advance, the development, research integration, and clinical implementation of AI applications are expected to accelerate, making them even more effective and accessible. However, ensuring the safety of patient data will require strict regulatory measures. Despite these challenges, AI represents a transformative opportunity for medicine, particularly in epilepsy neuroimaging. Since ML and DL models thrive on large datasets, fostering collaborations and expanding open-access databases will become increasingly pivotal in the future.
Evidence suggests that epilepsy may contribute to both functional and structural brain alterations, with atypical language organization (ALO) being more frequently observed in epilepsy patients compared to healthy controls. This study aimed to investigate structural connectivity differences between left-hemispheric language dominance (LLD) and ALO in patients with temporal lobe epilepsy (TLE) using probabilistic tractography. Twenty-two TLE patients with ALO and 27 age- and sex-matched patients with LLD who underwent pre-surgical language assessment using 3 Tesla functional magnetic resonance imaging (fMRI) were included in this study. Both groups were subdivided further based on the seizure onset hemisphere. SPM12 was used to determine the functional task-based language pattern. Individual structural connectivity network was investigated based on tractography and whole brain segmentation. In patients with ALO, the structural connectivity of the inferior frontal gyrus (IFG) exhibited stronger lateralization to the right hemisphere compared to LLD patients (ANOVA, pFDR = 0.030). FMRI activations in the IFG and superior temporal gyrus (STG) ipsilateral to seizure onset were negatively associated with epilepsy duration (puncorrected < 0.001) in ALO with right-hemispheric TLE. Furthermore, lateralization index of the STG- (Pearson's correlation: ρ = 0.714, puncorrected = 0.020) and IFG-connectivity (ρ = -0.608, puncorrected = 0.036) correlated with epilepsy duration in ALO. Several similarities regarding structural connectivity were detected across all groups. However, increased right-hemispheric connectivity of the IFG was detected in ALO, showing a correlation with epilepsy duration. Network analysis may enhance the understanding reorganization patterns underlying epilepsy and help to improve neurosurgical outcome.
Objective: Anterior temporal lobe resection (ATLR) is an effective treatment for drug‐resistant temporal lobe epilepsy (TLE) but carries a substantial risk of language impairment, particularly in naming. Understanding and predicting the impact of ATLR on language functions remains a major clinical challenge. Methods: In this study, we used functional magnetic resonance imaging (fMRI) to investigate the short‐term effects of ATLR on the organization of the functional language connectome with a focus on the role of the nondominant hemisphere. We studied 44 patients with TLE due to unilateral hippocampal sclerosis (24 left, 20 right) who underwent language fMRI and neuropsychological testing preoperatively and 4 months after ATLR. We examined functional connectivity changes pre‐ and postsurgery and their relationship with neuropsychological performance. Results: ATLR induced widespread alterations in functional connectivity, with distinct ipsilateral disruptions and contralateral compensatory changes. Left ATLR reduced mainly interhemispheric temporal connectivity, whereas right ATLR primarily affected bilateral frontal connections. Postoperatively, left ATLR showed increased intrahemispheric frontotemporal connectivity, and right ATLR exhibited more widespread intra‐ and interhemispheric increases. In left TLE, better preoperative naming was associated with stronger connectivity of the right (nondominant) temporal lobe and between the left inferior frontal cortex and bilateral posterior hippocampi. Postoperatively, reduced right frontotemporal integration was linked to greater naming decline, underscoring a compensatory role of the nondominant hemisphere. Finally, a machine learning model using preoperative functional connectivity fingerprints outperformed demographic and clinical variables in predicting clinically significant naming decline following ATLR. Significance: Our findings highlight the critical role of the right temporal lobe in supporting naming function after left ATLR and suggest that preoperative assessment of its connectivity may improve prediction of postoperative language outcomes.
Automated lesion segmentation can support rapid and reproducible quantitative assessment and long-term monitoring of multiple sclerosis. Ultra-high-field 7T MRI provides increased spatial resolution and lesion visibility but introduces challenges for automated analysis, including field inhomogeneity, susceptibility-related artifacts, and spatially varying image contrast. Although deep learning methods perform well at conventional field strengths, their application to 7T MRI remains underexplored. We evaluated nnUNet using FLAIR alone, MP2RAGE alone, and combined MP2RAGE+FLAIR. The dataset included 115 longitudinal 7T MRI scans from 57 patients with MS, with 20 scans from 11 independent patients reserved for testing. DeepMedic was included as a comparative architecture. On the test set, nnUNet achieved mean DSC values of 0.72, 0.71, and 0.66 for combined, MP2RAGE-only, and FLAIR-only inputs, respectively. No significant DSC differences were observed between nnUNet and DeepMedic. For both architectures, adding FLAIR to MP2RAGE did not show a statistically significant additional benefit in the present test cohort, whereas MP2RAGE-containing inputs significantly outperformed FLAIR-only input. The findings of the present study provide preliminary evidence suggesting that the observed voxel-level segmentation performance was primarily associated with the inclusion of MP2RAGE. However, lesion-level performance depended on lesion size, indicating further investigation of contrast-specific effects on small-lesion detection.
Background: Language processing is organized in brain networks, generally lateralized to the left hemisphere. In clinical routine, task-based functional MRI (fMRI) is the gold standard for non-invasive evaluation of language lateralization. However, standard fMRI does not account for the individual heterogeneity of language networks. fMRI-based functional connectivity analysis (FCA) has addressed this shortcoming by identifying correlates of dysfunctional language networks in epilepsy patients. In order to individualize FCA, translation of group-based network analysis to individual epilepsy patients is necessary. Here, we introduce the functional connectivity laterality index (LI FC ) as a potential clinical parameter. Methods: This retrospective study included adults with unilateral temporal lobe epilepsy (TLE) who underwent task-based fMRI and a control group. To account for individual variations, fMRI was performed at a 3T scanner and processed entirely in native space. Laterality indices were computed from both activation maps and seed-based functional connectivity across frontal, temporal, and combined language regions. Group comparisons and associations with neuropsychological language performance were analysed. Results: Forty patients with unilateral TLE and 25 healthy controls were included. Groups were comparable in terms of sex, handedness, imaging findings, and epilepsy duration, although the controls were younger. Classical LI showed predominantly left-lateralization, while LI FC showed mostly bilateral language lateralization. Neuropsychological performance did not differ between TLE-left and TLE-right patients. While a difference in semantic fluency was initially observed, it did not survive correction for multiple comparisons. However, analyses based on functional network organization revealed a critical pattern: Patients with language networks lateralized to the seizure side—as quantified by LI FC —showed significantly poorer semantic fluency and lower composite language scores. This potential clinically relevant result was exclusive to FCA and was not captured by activation data alone. Conclusion: Our findings suggest an individual vulnerability of language networks to seizure burden, although this uncorrected, exploratory finding requires replication in larger, independent samples. Incorporating functional connectivity measures into presurgical clinical fMRI analysis may enhance our ability to predict and protect language function in patients with TLE by prioritizing further therapies or surgical interventions for those at higher risk of language decline.
Resective neurosurgery is a cornerstone treatment for many neurological conditions. Although traditionally viewed as localised procedure, increasing evidence from advanced magnetic resonance imaging (MRI) shows that also non-resected anatomy can degenerate following surgery. The relationship between local tissue removal and these postoperative changes remains thus far speculative. Here, we investigate the hypothesis that degenerative changes to surgically preserved grey and white matter are mediated by transneuronal degeneration , a deterioration of intact neuronal populations due to lost axonal input. Using a robust structural and diffusion MRI framework, we first identify widespread postoperative atrophy: pronounced cortical thickness decreases near the resection, and extensive white matter impairments across the ipsilateral hemisphere. Importantly, we then link these alterations to surgical white matter disruption, revealing a sequential network atrophy following neurosurgery. Beyond degenerative effects, we also demonstrate often reported structural network reorganisations as an artefact of image processing, indicating limited capacity for macroscale plasticity post-resection. Key Points: Resective neurosurgery can lead to postoperative atrophy of non-resected grey and white matter. Utilising a robust longitudinal neuroimaging framework, we evaluate these changes as consequences of transneuronal degeneration , demonstrating a sequential atrophy of brain anatomy following the loss of surgically disrupted white matter. The majority of severed connections projected to atrophic cortices near the resection, indicating them to be short association fibers. Consecutive white matter atrophy however affected extensive ipsilateral pathways, highlighting a cascading network effect of resective brain surgery. We did not find evidence to support often reported structural reorganisations of large-scale brain networks post-resection. Yet, we demonstrate that evidence typically interpreted as such can be replicated when using non-robust reconstruction methods.