Artificial intelligence/machine learning for epilepsy and seizure diagnosis.

Accurate seizure and epilepsy diagnosis remains a challenging task due to the complexity and variability of manifestations, which can lead to delayed or missed diagnosis. Machine learning (ML) and artificial intelligence (AI) is a rapidly developing field, with growing interest in integrating and applying these tools to aid clinicians facing diagnostic uncertainties. ML algorithms, particularly deep neural networks, are increasingly employed in interpreting electroencephalograms (EEG), neuroimaging, wearable data, and seizure videos. This review discusses the development and testing phases of AI/ML tools, emphasizing the importance of generalizability and interpretability in medical applications, and highlights recent publications that demonstrate the current and potential utility of AI to aid clinicians in diagnosing epilepsy. Current barriers of AI integration in patient care include dataset availability and heterogeneity, which limit studies' quality, interpretability, comparability, and generalizability. ML and AI offer substantial promise in improving the accuracy and efficiency of epilepsy diagnosis. The growing availability of diverse datasets, enhanced processing speed, and ongoing efforts to standardize reporting contribute to the evolving landscape of AI applications in clinical care.

Machine learning applications in epilepsy.

Machine learning leverages statistical and computer science principles to develop algorithms capable of improving performance through interpretation of data rather than through explicit instructions. Alongside widespread use in image recognition, language processing, and data mining, machine learning techniques have received increasing attention in medical applications, ranging from automated imaging analysis to disease forecasting. This review examines the parallel progress made in epilepsy, highlighting applications in automated seizure detection from electroencephalography (EEG), video, and kinetic data, automated imaging analysis and pre-surgical planning, prediction of medication response, and prediction of medical and surgical outcomes using a wide variety of data sources. A brief overview of commonly used machine learning approaches, as well as challenges in further application of machine learning techniques in epilepsy, is also presented. With increasing computational capabilities, availability of effective machine learning algorithms, and accumulation of larger datasets, clinicians and researchers will increasingly benefit from familiarity with these techniques and the significant progress already made in their application in epilepsy.

Magnetic resonance imaging connectivity for the prediction of seizure outcome in temporal lobe epilepsy.

OBJECTIVE: Currently, approximately 60-70% of patients with unilateral temporal lobe epilepsy (TLE) remain seizure-free 3 years after surgery. The goal of this work was to develop a presurgical connectivity-based biomarker to identify those patients who will have an unfavorable seizure outcome 1-year postsurgery. METHODS: Resting-state functional and diffusion-weighted 3T magnetic resonance imaging (MRI) was acquired from 22 unilateral (15 right, 7 left) patients with TLE and 35 healthy controls. A seizure propagation network was identified including ipsilateral (to seizure focus) and contralateral hippocampus, thalamus, and insula, with bilateral midcingulate and precuneus. Between each pair of regions, functional connectivity based on correlations of low frequency functional MRI signals, and structural connectivity based on streamline density of diffusion MRI data were computed and transformed to metrics related to healthy controls of the same age. RESULTS: A consistent connectivity pattern representing the network expected in patients with seizure-free outcome was identified using eight patients who were seizure-free at 1-year postsurgery. The hypothesis that increased similarity to the model would be associated with better seizure outcome was tested in 14 other patients (Engel class IA, seizure-free: n = 5; Engel class IB-II, favorable: n = 4; Engel class III-IV, unfavorable: n = 5) using two similarity metrics: Pearson correlation and Euclidean distance. The seizure-free connectivity model successfully separated all the patients with unfavorable outcome from the seizure-free and favorable outcome patients (p = 0.0005, two-tailed Fisher's exact test) through the combination of the two similarity metrics with 100% accuracy. No other clinical and demographic predictors were successful in this regard. SIGNIFICANCE: This work introduces a methodologic framework to assess individual patients, and demonstrates the ability to use network connectivity as a potential clinical tool for epilepsy surgery outcome prediction after more comprehensive validation.

Recurrent secondary generalization in frontal lobe epilepsy: Predictors and a potential link to surgical outcome?

OBJECTIVE: Frontal lobe epilepsy (FLE) frequently leads to secondary generalized tonic-clonic seizures (SGTCS). However, little is known about the clinical, electrophysiologic, and radiologic correlates of SGTCS and whether these could influence diagnosis and treatment. METHODS: A cohort of 48 patients with confirmed FLE was retrospectively identified and dichotomized into a group with and a group without SGTCS defined by history (≥1/year) or video-electroencephalography (vEEG). Demographics, seizure semiology, vEEG, neuroimaging data, and estimated seizure-onset zone were tabulated, and their association with the occurrence of SGTCS was evaluated with use of a chi-square test. Independent predictors of SGTCS were confirmed using a stepwise multivariate analysis. Similarly, these predictors as well as a history of SGTCS were tested as multivariate predictors of the postoperative International League Against Epilepsy (ILAE) score in the surgical subgroup (n = 25). RESULTS: We identified three independent predictors of a history of SGTCS in FLE, including loss of responsiveness at seizure onset (corrected p = 0.04), a semiology involving early elementary motor signs (corrected p = 0.01), and multifocal spikes on EEG (corrected p = 0.02). A seizure-free outcome occurred in 57% of surgical cases and was more likely in the group without SGTCS (100%, p = 0.001). When considering only SGTCS occurring during video-EEG monitoring, the association with semiology and surgical outcome vanished, but the association with preserved awareness and a multifocal EEG persisted. SIGNIFICANCE: A history of SGTCS is related to a specific ictal semiology and interictal EEG, and may have a role in surgical risk stratification.

Factors affecting bilateral temporal lobe hypometabolism on 18F-FDG PET brain scan in unilateral medial temporal lobe epilepsy.

Bilateral temporal lobe hypometabolism (BTH) on (18)F-FDG PET brain scan is frequently seen in unilateral medial temporal lobe epilepsy (mTLE). This study aimed to identify the factors that influence BTH in patients with mTLE in order to minimize the significant factor(s) prior to performing a FDG-PET brain scan. Forty patients with unilateral mTLE who underwent (18)F-FDG PET scan for presurgical epilepsy workup were included. Bilateral temporal lobe hypometabolism of the anterior and medial parts of the temporal lobe was identified by a semiquantitative visual scale. Lateralization of TLE was identified by either intracranial EEG (22/40 cases) and/or improvement of seizure 2 years after temporal lobectomy (37/40 cases). The factors analyzed included basic demographic characteristics (age, sex, occupation, years of education, and handedness), history related to seizure (age at epilepsy onset and epilepsy duration, history of febrile seizure and head injury, frequency of seizure with impaired cognition in the last 3 months, presence of secondarily generalized tonic-clonic seizure, automatism side, presence of postictal confusion, and side of MRI temporal abnormality), information during video-EEG monitoring (clinical lateralization, interictal scalp EEG lateralization (interictal epileptiform discharge), and ictal scalp EEG lateralization), and information during the FDG-PET study (duration from the last seizure (≤2 days or >2 days), last seizure type, and the presence of slow waves or sharp waves during the FDG uptake period). Significant factors related to BTH were analyzed using multivariate analysis. Only the ≤2-day duration from the last seizure to the PET scan shows a significant effect (p=0.021) on BTH finding with 15 times greater incidence compared to a duration >2 days. Bilateral temporal lobe hypometabolism, which causes conflict in lateralizing the epileptogenic zone in temporal lobe epilepsy, can be avoided by performing PET scan more than 2 days after the last seizure.

Impact of reconstruction parameters on lesion detection and localization in joint ictal/inter-ictal SPECT reconstruction.

PURPOSE: Previously, a joint ictal/inter-ictal SPECT reconstruction was proposed to reconstruct a differential image representing the change of brain SPECT image from an inter-ictal to an ictal study. The so-called joint method yielded better performance for epileptic foci localization than the conventional subtraction method. In this study, we evaluated the performance of different reconstruction settings of the joint reconstruction of ictal/inter-ictal SPECT data, which creates a differential image showing the difference between ictal and inter-ictal images, in lesion detection and localization in epilepsy imaging. METHODS: Differential images reconstructed from phantom data using the joint and the subtraction methods were compared based on lesion detection performance (channelized Hotelling observer signal-to-noise ratio (SNRCHO) averaged across four lesion-to-background contrast levels) at the optimal iteration. The joint-initial method which was the joint method that was initialized by the subtraction method at optimal iteration was also used to reconstruct differential images. These three methods with respective optimal iteration and the subtraction method with four iterations were applied to epileptic patient datasets. A human observer lesion localization study was performed based on localization receiver operating characteristic (LROC) analysis. RESULTS: From the phantom study, at their respective optimal iteration, the joint method yielded an improvement in lesion detection performance over the subtraction method of 26%, which increased to 145% when using the joint-initial method. From the patient study, the joint-initial method yielded the highest area under the LROC curve as compared with those of the joint and the subtraction methods with optimal iteration and with 4 iterations (0.44 vs 0.41, 0.39 and 0.36, respectively). CONCLUSIONS: In lesion detection and localization, the joint method at optimal iteration outperformed the subtraction method at optimal iteration and at iteration typically used in clinical practice. Furthermore, initialization by the subtraction method improved the performance of the joint method.

Surgical disconnection of epilepsy network correlates with improved outcomes.

PURPOSE: A novel software algorithm combining non-invasive EEG and resting state functional MRI data to map networks of cortex correlated to epileptogenic tissue was used to map an epilepsy network non-invasively. The relationship between epilepsy network connectivity and outcomes after surgery was investigated using this non-invasive and non-concurrent modeling algorithm. METHOD: Scalp EEG and resting state functional MRI were acquired for nineteen patients with temporal lobe epilepsy. The hypothetical irritative zone was mapped, and resting state functional MRI data was used to model regions functionally correlated with the irritative zone. Epilepsy network connectivity was measured in patient with temporal lobe epilepsy (n = 19) both pre- and post-operatively. Temporal networks were also mapped in healthy control participants (n = 6). RESULTS: Thirteen of nineteen patients (68 %) were seizure free after 20.3 ± 4.8 months. Epilepsy network connectivity within the temporal lobe was significantly higher among patients with temporal lobe epilepsy compared to the healthy control patients (p < 0.05). Disconnection of the epilepsy network was significantly higher in patients who were seizure free. Using spearman rho analyses, neuropsychological function after surgery was found to be relatively better in patients with higher degree of epilepsy network disconnection. CONCLUSIONS: The magnitude of network disconnection after surgery was strongly associated with increased rates of seizure freedom and relatively better neuropsychological measures of memory and naming function. It was shown that seizure-free outcomes and relatively improved neuropsychological function correlated with surgical disconnection of a highly synchronous epilepsy network.

Hybrid MR-PET in Neuroimaging.

Hybrid magnetic resonance (MR)-positron emission tomography (MR-PET) is a novel technology with advantages over sequential MR and PET imaging, allowing maintain full individual diagnostic performance with negligible mutual interference between the two hardware settings. Obvious synergies between MR and PET in acquisition of anatomical, functional, and molecular information for neurological diseases into one single image pave the way for establishing clear clinical indications for hybrid MR-PET as well as addressing unmet neuroimaging needs in future clinics and research. Further developments in attenuation correction, quantification, workflow, and effective MR-PET data management might unfold the full potential of integrated multimodality imaging.