Learning to see the invisible: A data-driven approach to finding the underlying patterns of abnormality in visually normal brain magnetic resonance images in patients with temporal lobe epilepsy

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.

Identification of epileptic networks with graph convolutional network incorporating oscillatory activities and evoked synaptic responses

Stereoelectroencephalography (SEEG) offers unique neural data from in-depth brain structures with fine temporal resolutions to better investigate the origin of epileptic brain activities. Although oscillatory patterns from different frequency bands and functional connectivity computed from the SEEG datasets are employed to study the epileptic zones, direct electrical stimulation-evoked electrophysiological recordings of synaptic responses, namely cortical-cortical evoked potentials (CCEPs), from the same SEEG electrodes are not explored for the localization of epileptic zones. Here we proposed a two-stream model with unsupervised learning and graph convolutional network tailored to the SEEG and CCEP datasets in individual patients to perform localization of epileptic zones. We compared our localization results with the clinically marked electrode sites determined for surgical resections. Our model had good classification capability when compared to other state-of-the-art methods. Furthermore, based on our prediction results we performed group-level brain-area mapping analysis for temporal, frontal and parietal epilepsy patients and found that epileptic and non-epileptic brain networks were distinct in patients with different types of focal epilepsy. Our unsupervised data-driven model provides personalized localization analysis for the epileptic zones. The epileptic and non-epileptic brain areas disclosed by the prediction model provide novel insights into the network-level pathological characteristics of epilepsy.

Convolutional Neural Network Algorithm to Determine Lateralization of Seizure Onset in Patients With Epilepsy: A Proof-of-Principle Study

BACKGROUND AND OBJECTIVES

A new frontier in diagnostic radiology is the inclusion of machine-assisted support tools that facilitate the identification of subtle lesions often not visible to the human eye. Structural neuroimaging plays an essential role in the identification of lesions in patients with epilepsy, which often coincide with the seizure focus. In this study, we explored the potential for a convolutional neural network (CNN) to determine lateralization of seizure onset in patients with epilepsy using T1-weighted structural MRI scans as input.

METHODS

Using a dataset of 359 patients with temporal lobe epilepsy (TLE) from 7 surgical centers, we tested whether a CNN based on T1-weighted images could classify seizure laterality concordant with clinical team consensus. This CNN was compared with a randomized model (comparison with chance) and a hippocampal volume logistic regression (comparison with current clinically available measures). Furthermore, we leveraged a CNN feature visualization technique to identify regions used to classify patients.

RESULTS

Across 100 runs, the CNN model was concordant with clinician lateralization on average 78% (SD = 5.1%) of runs with the best-performing model achieving 89% concordance. The CNN outperformed the randomized model (average concordance of 51.7%) on 100% of runs with an average improvement of 26.2% and outperformed the hippocampal volume model (average concordance of 71.7%) on 85% of runs with an average improvement of 6.25%. Feature visualization maps revealed that in addition to the medial temporal lobe, regions in the lateral temporal lobe, cingulate, and precentral gyrus aided in classification.

DISCUSSION

These extratemporal lobe features underscore the importance of whole-brain models to highlight areas worthy of clinician scrutiny during temporal lobe epilepsy lateralization. This proof-of-concept study illustrates that a CNN applied to structural MRI data can visually aid clinician-led localization of epileptogenic zone and identify extrahippocampal regions that may require additional radiologic attention.

CLASSIFICATION OF EVIDENCE

This study provides Class II evidence that in patients with drug-resistant unilateral temporal lobe epilepsy, a convolutional neural network algorithm derived from T1-weighted MRI can correctly classify seizure laterality.

Multi-Scale Deep Learning of Clinically Acquired Multi-Modal MRI Improves the Localization of Seizure Onset Zone in Children With Drug-Resistant Epilepsy

The present study investigates the effectiveness of a deep learning neural network for non-invasively localizing the seizure onset zone (SOZ) using multi-modal MRI data that are clinically acquired from children with drug-resistant epilepsy. A cortical parcellation was applied to localize the SOZ in cortical nodes of the epileptogenic hemisphere. At each node, the laminar surface analysis was followed to sample 1) the relative intensity of gray matter and white matter in multi-modal MRI and 2) the neighboring white matter connectivity using diffusion tractography edge strengths. A cross-validation was employed to train and test all layers of a multi-scale residual neural network (msResNet) that can classify SOZ node in an end-to-end fashion. A prediction probability of a given node belonging to the SOZ class was proposed as a non-invasive MRI marker of seizure onset likelihood. In an independent validation cohort, the proposed MRI marker provided a very large effect size of Cohen's d = 1.21 between SOZ and non-SOZ, and classified SOZ with a balanced accuracy of 0.75 in lesional and 0.67 in non-lesional MRI groups. The subsequent multi-variate logistic regression found the incorporation of the proposed MRI marker into interictal intracranial EEG (iEEG) markers further improves the differentiation between the epileptogenic focus (defined as SOZ resected during surgery) and non-epileptogenic sites (i.e., non-SOZ sites preserved during surgery) up to 15 % in non-lesional MRI group, suggesting that the proposed MRI marker could improve the localization of epileptogenic foci for successful pediatric epilepsy surgery.

The Temporal Lobe as a Symptomatogenic Zone in Medial Parietal Lobe Epilepsy

Some surgical failures after temporal lobe epilepsy surgery may be due to the presence of an extratemporal epileptogenic zone. Of particular interest is the medial parietal lobe due to its robust connectivity with mesial temporal structures. Seizures in that area may be clinically silent before propagating to the symptomatogenic temporal lobe. In this paper, we present an overview of the anatomical connectivity, semiology, radiology, electroencephalography, neuropsychology, and outcomes in medial parietal lobe epilepsy. We also present two illustrative cases of seizures originating from the precuneus and the posterior cingulate cortex. We conclude that the medial parietal lobe should be strongly considered for sampling by intracranial electrodes in individuals with nonlesional temporal lobe epilepsy, especially if scrutinizing the presurgical data produces discordant findings.

Common functional connectivity alterations in focal epilepsies identified by machine learning

OBJECTIVE

This study was undertaken to identify shared functional network characteristics among focal epilepsies of different etiologies, to distinguish epilepsy patients from controls, and to lateralize seizure focus using functional connectivity (FC) measures derived from resting state functional magnetic resonance imaging (MRI).

METHODS

Data were taken from 103 adult and 65 pediatric focal epilepsy patients (with or without lesion on MRI) and 109 controls across four epilepsy centers. We used three whole-brain FC measures: parcelwise connectivity matrix, mean FC, and degree of FC. We trained support vector machine models with fivefold cross-validation (1) to distinguish patients from controls and (2) to lateralize the hemisphere of seizure onset in patients. We reported the regions and connections with the highest importance from each model as the common FC differences between the compared groups.

RESULTS

FC measures related to the default mode and limbic networks had higher importance relative to other networks for distinguishing epilepsy patients from controls. In lateralization models, regions related to somatosensory, visual, default mode, and basal ganglia showed higher importance. The epilepsy versus control classification model trained using a 400-parcel connectivity matrix achieved a median testing accuracy of 75.6% (median area under the curve [AUC] = .83) in repeated independent testing. Lateralization accuracy using the 400-parcel connectivity matrix reached a median accuracy of 64.0% (median AUC = .69).

SIGNIFICANCE

Machine learning models revealed common FC alterations in a heterogeneous group of patients with focal epilepsies. The distribution of the most altered regions supports the hypothesis that shared functional alteration exists beyond the seizure onset zone and its epileptic network. We showed that FC measures can distinguish patients from controls, and further lateralize focal epilepsies. Future studies are needed to confirm these findings by using larger numbers of epilepsy patients.

Machine learning applications on neuroimaging for diagnosis and prognosis of epilepsy: A review

Machine learning is playing an increasingly important role in medical image analysis, spawning new advances in the clinical application of neuroimaging. There have been some reviews on machine learning and epilepsy before, and they mainly focused on electrophysiological signals such as electroencephalography (EEG) and stereo electroencephalography (SEEG), while neglecting the potential of neuroimaging in epilepsy research. Neuroimaging has its important advantages in confirming the range of the epileptic region, which is essential in presurgical evaluation and assessment after surgery. However, it is difficult for EEG to locate the accurate epilepsy lesion region in the brain. In this review, we emphasize the interaction between neuroimaging and machine learning in the context of epilepsy diagnosis and prognosis. We start with an overview of epilepsy and typical neuroimaging modalities used in epilepsy clinics, MRI, DWI, fMRI, and PET. Then, we elaborate two approaches in applying machine learning methods to neuroimaging data: (i) the conventional machine learning approach combining manual feature engineering and classifiers, (ii) the deep learning approach, such as the convolutional neural networks and autoencoders. Subsequently, the application of machine learning on epilepsy neuroimaging, such as segmentation, localization, and lateralization tasks, as well as tasks directly related to diagnosis and prognosis are looked into in detail. Finally, we discuss the current achievements, challenges, and potential future directions in this field, hoping to pave the way for computer-aided diagnosis and prognosis of epilepsy.

Accurate lateralization and classification of MRI-negative 18F-FDG-PET-positive temporal lobe epilepsy using double inversion recovery and machine-learning

OBJECTIVE

The main objective of this study was to determine the ability of double inversion recovery (DIR) data coupled with machine-learning algorithms to distinguish normal individuals from epileptic subjects and to identify the laterality of the focus side in MRI-negative, PET-positive temporal lobe epilepsy (TLE) patients.

MATERIALS AND METHODS

We used whole-brain DIR data as the input features with which to train a linear support-vector machine model in 63 participants who underwent high-resolution structural MRI and DIR scans. The subjects included 20 left TLE patients, 19 right TLE patients, and 24 healthy controls (HCs).

RESULTS

Using the DIR data, we achieved a robust accuracy of 87.30% for discriminating among the left TLE, right TLE, and HC groups as well as 84.61%, 97.72%, and 93.02% prediction accuracies for distinguishing left TLE from right TLE, HC from right TLE, and HC from left TLE, respectively.

INTERPRETATION

Our experimental results suggest that DIR data coupled with machine-learning algorithms provide a promising approach to identifying MRI-negative TLE patients, especially when fluorodeoxyglucose-PET is not available.

Making the Invisible Visible: Advanced Neuroimaging Techniques in Focal Epilepsy

It has been a clinically important, long-standing challenge to accurately localize epileptogenic focus in drug-resistant focal epilepsy because more intensive intervention to the detected focus, including resection neurosurgery, can provide significant seizure reduction. In addition to neurophysiological examinations, neuroimaging plays a crucial role in the detection of focus by providing morphological and neuroanatomical information. On the other hand, epileptogenic lesions in the brain may sometimes show only subtle or even invisible abnormalities on conventional MRI sequences, and thus, efforts have been made for better visualization and improved detection of the focus lesions. Recent advance in neuroimaging has been attracting attention because of the potentials to better visualize the epileptogenic lesions as well as provide novel information about the pathophysiology of epilepsy. While the progress of newer neuroimaging techniques, including the non-Gaussian diffusion model and arterial spin labeling, could non-invasively detect decreased neurite parameters or hypoperfusion within the focus lesions, advances in analytic technology may also provide usefulness for both focus detection and understanding of epilepsy. There has been an increasing number of clinical and experimental applications of machine learning and network analysis in the field of epilepsy. This review article will shed light on recent advances in neuroimaging for focal epilepsy, including both technical progress of images and newer analytical methodologies and discuss about the potential usefulness in clinical practice.

Evaluation of Brain Network Properties in Patients with MRI-Negative Temporal Lobe Epilepsy: An MEG Study

Abnormal functional brain networks of temporal lobe epilepsy (TLE) patients with structural abnormalities may partially reflect structural lesions rather than either TLE per se or functional compensatory processes. In this study, we sought to investigate the brain-network properties of intractable TLE patients apart from the effects of structural abnormalities. The brain network properties of 20 left and 23 right MRI-negative TLE patients and 22 healthy controls were evaluated using magnetoencephalographic recordings in six main frequency bands. A slowing of oscillatory brain activity was observed for the left or right TLE group vs. healthy controls. The TLE groups presented significantly increased functional connectivity in the delta, theta, lower alpha and beta bands, and significantly greater values in the normalized clustering coefficient and path length, and significantly smaller values in the weighted small-world measure in the theta band when compared to healthy controls. Alterations in global and regional band powers can be attributed to spectral slowing in TLE patients. The brain networks of TLE patients displayed abnormally high synchronization in multi-frequency bands and shifted toward a more regular architecture with worse network efficiency in the theta band. Without the contamination of structural lesions, these significant findings can be helpful for better understanding of the pathophysiological mechanism of TLE. The theta band can be considered as a preferred frequency band for investigating the brain-network dysfunction of MRI-negative intractable TLE patients.

Clinical Application of Machine Learning Models for Brain Imaging in Epilepsy: A Review

Epilepsy is a common neurological disorder characterized by recurrent and disabling seizures. An increasing number of clinical and experimental applications of machine learning (ML) methods for epilepsy and other neurological and psychiatric disorders are available. ML methods have the potential to provide a reliable and optimal performance for clinical diagnoses, prediction, and personalized medicine by using mathematical algorithms and computational approaches. There are now several applications of ML for epilepsy, including neuroimaging analyses. For precise and reliable clinical applications in epilepsy and neuroimaging, the diverse ML methodologies should be examined and validated. We review the clinical applications of ML models for brain imaging in epilepsy obtained from a PubMed database search in February 2021. We first present an overview of typical neuroimaging modalities and ML models used in the epilepsy studies and then focus on the existing applications of ML models for brain imaging in epilepsy based on the following clinical aspects: (i) distinguishing individuals with epilepsy from healthy controls, (ii) lateralization of the temporal lobe epilepsy focus, (iii) the identification of epileptogenic foci, (iv) the prediction of clinical outcomes, and (v) brain-age prediction. We address the practical problems and challenges described in the literature and suggest some future research directions.

Detection of covert lesions in focal epilepsy using computational analysis of multimodal magnetic resonance imaging data

Objective

To compare the location of suspect lesions detected by computational analysis of multimodal magnetic resonance imaging data with areas of seizure onset, early propagation, and interictal epileptiform discharges (IEDs) identified with stereoelectroencephalography (SEEG) in a cohort of patients with medically refractory focal epilepsy and radiologically normal magnetic resonance imaging (MRI) scans.

Methods

We developed a method of lesion detection using computational analysis of multimodal MRI data in a cohort of 62 control subjects, and 42 patients with focal epilepsy and MRI‐visible lesions. We then applied it to detect covert lesions in 27 focal epilepsy patients with radiologically normal MRI scans, comparing our findings with the areas of seizure onset, early propagation, and IEDs identified at SEEG.

Results

Seizure‐onset zones (SoZs) were identified at SEEG in 18 of the 27 patients (67%) with radiologically normal MRI scans. In 11 of these 18 cases (61%), concordant abnormalities were detected by our method. In the remaining seven cases, either early seizure propagation or IEDs were observed within the abnormalities detected, or there were additional areas of imaging abnormalities found by our method that were not sampled at SEEG. In one of the nine patients (11%) in whom SEEG was inconclusive, an abnormality, which may have been involved in seizures, was identified by our method and was not sampled at SEEG.

Significance

Computational analysis of multimodal MRI data revealed covert abnormalities in the majority of patients with refractory focal epilepsy and radiologically normal MRI that co‐located with SEEG defined zones of seizure onset. The method could help identify areas that should be targeted with SEEG when considering epilepsy surgery.

FLAIR-Wise Machine-Learning Classification and Lateralization of MRI-Negative 18F-FDG PET-Positive Temporal Lobe Epilepsy

Objective: In this study, we investigated the ability of fluid-attenuated inversion recovery (FLAIR) data coupled with machine-leaning algorithms to differentiate normal and epileptic brains and identify the laterality of focus side in temporal lobe epilepsy (TLE) patients with visually negative MRI.

Materials and Methods: The MRI data were acquired on a 3-T MR system (Philips Medical Systems). After pre-proceeding stage, the FLAIR signal intensities were extracted from specific regions of interest, such as the amygdala, cerebral white matter, inferior temporal gyrus, middle temporal gyrus, parahippocampal gyrus, superior temporal gyrus, and temporal pole, and fed into a classification framework followed by a support vector machine as classifier. The proposed lateralization framework was assessed in a group of MRI-negative unilateral TLE patients (N = 42; 23 left TLE and 19 right TLE) and 34 healthy controls (HCs) based on a leave-one-out cross-validation strategy.

Results: Using the FLAIR data, we obtained a 75% accuracy for discriminating the three groups, as well as 87.71, 83.01, and 76.19% accuracies for HC/right TLE, HC/left TLE, and left TLE/right TLE tasks, respectively.

Interpretation: The experimental results show that FLAIR data can potentially be considered an informative biomarker for improving the pre-surgical diagnostic confidence in patients with MRI-negative TLE.

Learning to see the invisible: A data-driven approach to finding the underlying patterns of abnormality in visually normal brain magnetic resonance images in patients with temporal lobe epilepsy

Objective

To find the covert patterns of abnormality in patients with unilateral temporal lobe epilepsy (TLE) and visually normal brain magnetic resonance images (MRI‐negative), comparing them to those with visible abnormalities (MRI‐positive).

Methods

We used multimodal brain MRI from patients with unilateral TLE and employed contemporary machine learning methods to predict the known laterality of seizure onset in 104 subjects (82 MRI‐positive, 22 MRI‐negative). A visualization approach entitled "Importance Maps" was developed to highlight image features predictive of seizure laterality in both the MRI‐positive and MRI‐negative cases.

Results

Seizure laterality could be predicted with an area under the receiver operating characteristic curve of 0.981 (95% confidence interval [CI] =0.974‐0.989) in MRI‐positive and 0.842 (95% CI = 0.736‐0.949) in MRI‐negative cases. The known image features arising from the hippocampus were the leading predictors of seizure laterality in the MRI‐positive cases, whereas widespread temporal lobe abnormalities were revealed in the MRI‐negative cases.

Significance

Covert abnormalities not discerned on visual reading were detected in MRI‐negative TLE, with a spatial pattern involving the whole temporal lobe, rather than just the hippocampus. This suggests that MRI‐negative TLE may be associated with subtle but widespread temporal lobe abnormalities. These abnormalities merit close inspection and postacquisition processing if there is no overt lesion.