Neurologists often face challenges in identifying epileptic activities within multichannel EEG recordings, requiring extensive hours of analysis. Computer-aided diagnosis systems have been proposed to reduce manual inspection of EEG signals by neurologists. However, direct analysis of EEG signals is difficult due to their complex and dynamic nature, with variation across multiple patients. Therefore, researchers have proposed the short-time Fourier transform (STFT) to capture dynamic events indicative of seizures through time-varying frequency representation of EEG signals. However, tradeoffs between time and frequency resolution limited the spectrogram's interpretability and affected clinical deployment. Hence, this study proposes extracting high-resolution channels via a novel STFT spectrogram construction algorithm encompassing taper functions for seizure diagnosis. Initially, we extracted seizure and non-seizure segments from each channel of selected patients in the CHB-MIT dataset. Next, we systematically apply taper functions like Hann and Gaussian windows to minimize the edge effect during the construction of spectrogram images. Finally, we employ Dilated Convolutional Squeeze and Excitation Networks (DCSENets) through leave-one-patient-out cross-validation (LOPOCV) to perform patient-independent seizure classification. The proposed DCSENets achieve an average accuracy of 87.20±11.48% and 87.29±10.48% with Hann and Gaussian taper functions, respectively, and 86.85±11.56% without the taper function. Most patients with high performances indicate similarity in train-test sample distribution using the Kolmogorov-Smirnov test at 0.01 0.05. Furthermore, the Grad CAM deep visual explainer integration enhances the interpretability of the deep learning model's decision-making process. Consequently, neurologists are provided not only with enhanced visualized spectrograms but also a transparent model for improved seizure diagnosis.

Abstract Deep Brain Stimulation can improve tremor, bradykinesia, rigidity, and axial symptoms in patients with Parkinson’s disease. Potentially, improving each symptom may require stimulation of different white matter tracts. Here, we study a large cohort of patients (N = 237 from five centers) to identify tracts associated with improvements in each of the four symptom domains. Tremor improvements were associated with stimulation of tracts connected to primary motor cortex and cerebellum. In contrast, axial symptoms are associated with stimulation of tracts connected to the supplementary motor cortex and brainstem. Bradykinesia and rigidity improvements are associated with the stimulation of tracts connected to the supplementary motor and premotor cortices, respectively. We introduce an algorithm that uses these symptom-response tracts to suggest optimal stimulation parameters for DBS based on individual patient’s symptom profiles. Application of the algorithm illustrates that our symptom-tract library may bear potential in personalizing stimulation treatment based on the symptoms that are most burdensome in an individual patient.

Deep learning, particularly convolutional neural networks (CNNs), is increasingly used for epileptic seizure detection, leveraging their ability to directly extract features from EEG signals. Since CNNs excel with image data, various methods have been developed to convert signals into images, including the Gramian Angular Field (GAF) transformation algorithm. The algorithm is subtyped into the Gramian Angular Summation Field (GASF) and the Gramian Angular Difference Field (GADF). Each of these transformation techniques has been employed individually in earlier studies. However, each method focuses on unique properties of the EEG signal during image construction. Therefore, we proposed a feature fusion of the two transformation methods using the CNN model to learn richer representations. The experiment was conducted on publicly available Bonn datasets. Initially, the datasets are segmented into epochs of length 0.5s and step size 0.2s. Images are generated for each of the segments, resulting in 8600 images for each set of Bonn datasets. The generated images from each of the transformations are fed into parallel CNNs, and the extracted features are fused for improved performance. We conducted binary, ternary, and multiclass (5-class) classifications. The results show that feature fusion of GASF and GADF improves the classification performance with an average accuracy increase of 0.255%, 0.295%, and 0.660% for binary, ternary, and multiclass, respectively. This feature fusion of the GAF transformation approach shows promising potential for distinguishing various EEG patterns, underscoring its potential in seizure classification tasks.

The habenula (Hb) is a phylogenetically old structure connecting forebrain and brainstem monoaminergic nuclei that has been implicated in the pathogenesis of depression. Here, to investigate the clinical efficacy and neural mechanisms of stimulating the Hb for alleviating depression symptoms in humans, we bilaterally implanted electrodes in six patients with treatment-resistant depression and delivered high-frequency stimulation. Compared to baseline, we observed a substantial reduction in Hamilton Depression Rating Scale scores: 62.1% at 1-month, 64.0% at 3-month and 66.2% at 6-month follow-up. Local field potential data showed that acute Hb stimulation increased theta-band power, especially in the right side, which was related to the following clinical remission. Moreover, functional magnetic resonance imaging data showed that acute Hb stimulation enhanced blood oxygen level-dependent responses of the medial orbitofrontal cortex, raphe and substantia nigra, which are important components of the dopaminergic and serotonergic systems. Our findings demonstrated that Hb stimulation can alleviate depressive symptoms and modulate the activity of the medial orbitofrontal cortex, raphe and substantia nigra in treatment-resistant depression patients. This trial was registered under the clinical trial numbers NCT03667872 and ChiCTR2100045363.

Epilepsy is a prevalent neurological disorder that poses life-threatening emergencies. Early electroencephalogram (EEG) seizure detection can mitigate the risks and aid in the treatment of patients with epilepsy. EEG based epileptic seizure (ES) detection has significant applications in epilepsy treatment and medical diagnosis. Therefore, this paper presents an innovative framework for efficient ES detection, providing coefficient and distance correlation feature selection algorithms, a Bagged Tree-based classifer (BTBC), and Explainable Artificial Intelligence (XAI). Initially, the Butterworth filter is employed to eliminate various artifacts, and the discrete wavelet transform (DWT) is used to decompose the EEG signals and extract various eigenvalue features of the statistical time domain (STD) as linear and Fractal dimension-based non-linear (FD-NL). The optimal features are then identified through correlation coefficients with P-value and distance correlation analysis.These features are subsequently utilized by the Bagged Tree-based classifer (BTBC). The proposed model provides best performance in mitigating overfitting issues and improves the average accuracy by 2% using (CD, E), (AB, CD, E), and (A, B) experimental types as compared to other machine learning (ML) models using well-known Bonn and UCI-EEG benchmark datasets. Finally, SHapley additive exPlanation (SHAP) was used to interpret and explain the decision-making process of the proposed model. The results highlight the framework’s capability to accurately classify ES, thereby improving the diagnosis process in patients with brain dysfunctions.

More than half of adults with epilepsy undergoing resective epilepsy surgery achieve long-term seizure freedom and might consider withdrawing antiseizure medications. We aimed to identify predictors of seizure recurrence after starting postoperative antiseizure medication withdrawal and develop and validate predictive models. We performed an international multicentre observational cohort study in nine tertiary epilepsy referral centres. We included 850 adults who started antiseizure medication withdrawal following resective epilepsy surgery and were free of seizures other than focal non-motor aware seizures before starting antiseizure medication withdrawal. We developed a model predicting recurrent seizures, other than focal non-motor aware seizures, using Cox proportional hazards regression in a derivation cohort (n = 231). Independent predictors of seizure recurrence, other than focal non-motor aware seizures, following the start of antiseizure medication withdrawal were focal non-motor aware seizures after surgery and before withdrawal [adjusted hazard ratio (aHR) 5.5, 95% confidence interval (CI) 2.7-11.1], history of focal to bilateral tonic-clonic seizures before surgery (aHR 1.6, 95% CI 0.9-2.8), time from surgery to the start of antiseizure medication withdrawal (aHR 0.9, 95% CI 0.8-0.9) and number of antiseizure medications at time of surgery (aHR 1.2, 95% CI 0.9-1.6). Model discrimination showed a concordance statistic of 0.67 (95% CI 0.63-0.71) in the external validation cohorts (n = 500). A secondary model predicting recurrence of any seizures (including focal non-motor aware seizures) was developed and validated in a subgroup that did not have focal non-motor aware seizures before withdrawal (n = 639), showing a concordance statistic of 0.68 (95% CI 0.64-0.72). Calibration plots indicated high agreement of predicted and observed outcomes for both models. We show that simple algorithms, available as graphical nomograms and online tools (predictepilepsy.github.io), can provide probabilities of seizure outcomes after starting postoperative antiseizure medication withdrawal. These multicentre-validated models may assist clinicians when discussing antiseizure medication withdrawal after surgery with their patients.

The onset of Lennox-Gastaut syndrome (LGS), a severe epilepsy syndrome, is typically before 8 years of age. Late-onset LGS (with onset in adolescence and adulthood) is relatively rare clinically and has some differences from classical LGS. Herein, we describe the case of a patient with late-onset LGS and provide a literature review of such cases. The patient had focal epilepsy onset at 8 years of age. After a 9-year evolution, he suffered seizures of different types and had a diagnosis of late-onset LGS. Drug treatment was ineffective. Nothing was found on stereotactic electroencephalography (SEEG) and magnetic resonance imaging (MRI) during the course of the disease. After the second presurgical evaluation, we found a suspicious focus on high-resolution structural MRI which was verified by SEEG at last. After SEEG-guided radiofrequency thermocoagulation (RFTC), his seizures were controlled, and his cognitive function and quality of living clearly improved. However, his seizures recurred 2 years later, and he underwent left occipital resection. Thereafter, his seizures have been controlled until now. This case emphasizes the importance of high-resolution structural MRI in the treatment of LGS. Furthermore, it suggests that late-onset LGS may be caused by focal lesions and evolve from focal epilepsy. Thus, characterizing the clinical symptoms and performing individualized electroencephalographic follow-up are both very important. Additionally, the clinical outcome in this case implies the value and limitations of RFTC in patients with epilepsy and a clear focal lesion. Moreover, this case further supports differences between late-onset and classical LGS in terms of clinical manifestation, cognitive changes, prognosis, and treatment.

Objective: Stereoelectroencephalogram (SEEG) has been widely adapted to detect the electrical activity of patients with epilepsy. Due to the low-quality, large-amount, high-dimensionality characteristics of SEEG data, it is still challenging to comprehensively employ the SEEG signals to automatically and precisely determine the seizure onset zone (SOZ). This is because there is lack of an effective criterion for clinicians to select the target electrodes, which is of great importance for SOZ localization. Methods: We propose a SOZ localization method via analyzing the long-term SEEG monitoring for preoperative planning of epilepsy surgery. Considering that high frequency oscillations can reflect physiological brain activity of epileptic patients, we first extract the high-frequency features of the SEEG signals and utilize the convolutional neural network (CNN) to detect the interictal and seizure segments. Then we propose a novel criterion, namely adaptive high frequency epileptogenicity index (AHFEI), to determine the target electrodes. Results: We compare our SEEG-determined target electrodes with three preoperative planning of successful focal epilepsy resective surgery cases, finding that most localization results of our method are in consistent with clinical successful decision making, while the performance of our method outperforms than the state-of-the-art method for SOZ localization. Conclusion: Our SEEG-determined SOZ localization method can assist clinicians in preliminarily selecting the potential target electrodes according to long-term SEEG data automatically and effectively. Clinical Impact: The proposed automatic SOZ localization method has achieved satisfactory performance in the preliminary study, which has the great potential to be integrated into function-structure fused clinical decision-making system.

Treatment-resistant depression (TRD) is a chronic and severe psychiatric illness associated with limited therapeutic options. Deep brain stimulation (DBS) is a promising therapy for TRD patients. However, its safety and efficacy are still unclear. Here we reported the safety and efficacy of lateral habenula (LHb) DBS for a TRD patient who had failed medical, psychological, electroconvulsive, and ketamine therapy. The DBS system is compatible with 3T magnetic resonance imaging along with local field potential (LFP) streaming. Two DBS electrodes were implanted at the bilateral LHb without any complication. The patient showed acute stimulation effects and achieved long-term improvements in his depression, anxiety, and sleep with left LHb 160 Hz frequency stimulation, accompanying the change of LFPs. These results provided clinical evidence toward the safety and efficacy and electrophysiological basis of LHb DBS for TRD.

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.