Machine learning-based algorithm of drug-resistant prediction in newly diagnosed patients with temporal lobe epilepsy

• Published in: Clinical neurophysiology Vol. 171; pp. 154 - 163
• Main Authors: Mao, Lingyan, Zheng, Gaoxing, Cai, Yang, Luo, Wenyi, Zhang, Yijun, Wu, Kuidong, Ding, Jing, Wang, Xin
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.03.2025
• Subjects: Adolescent; Adult; Algorithms; Drug Resistant Epilepsy - diagnosis; Drug Resistant Epilepsy - physiopathology; EEG; Electroencephalography - methods; Ensemble learning technique; Epilepsy, Temporal Lobe - diagnosis; Epilepsy, Temporal Lobe - drug therapy; Epilepsy, Temporal Lobe - physiopathology; Female; Frontotemporal network; Functional network; Humans; Machine Learning; Male; Middle Aged; Neurology; Predicted algorithm; Prospective Studies; Young Adult; CC; Frontotemporal network; D; EEG; Predicted algorithm; ASMs; ROC; IFCN; KNN; HS; SVM; PLI; TLE; NDRE; AUC; Q; mTLE; SEEG; TH; Ensemble learning technique; MF; LF; PL; Functional network; phase lag index; leaf fraction; non-drug-resistant patients; area under curve; modularity; diameter; International Federation of Clinical Neurophysiology; stereo-electroencephalograph; tree hierarchy; receiver operating characteristic; temporal lobe epilepsy; support vector machines; clustering coefficient; characteristic path length; hippocampal sclerosis; electroencephalogram; mesial temporal lobe epilepsy; k-nearest neighbors; mean frequency; antiseizure medications
• ISSN: 1388-2457; 1872-8952; 1872-8952
• DOI: 10.1016/j.clinph.2025.01.008

•Alterations of EEG features were detected in 2-year drug-resistant outcomes in newly-diagnosed temporal lobe epilepsy patients.•EEG functional network-based algorithm predicts drug-resistant outcomes with 91.5% accuracy.•EEG features from the frontotemporal network enhance the prediction algorithm’s classification performance. To develop a predicted algorithm for drug-resistant epilepsy (DRE) in newly diagnosed temporal lobe epilepsy (TLE) patients. A total of 139 newly diagnosed TLE patients were prospectively enrolled, and long-term video EEG monitoring was recorded. Clinical evaluations, including seizure frequency and antiseizure medications (ASMs) usage, were collected and prospectively followed up for 24 months. Interictal EEG data were used for feature extraction, identifying 216 EEG network features. Traditional machine learning and ensemble learning techniques were employed to predict DRE outcomes. Over two years, TLE patients with DRE exhibited significant EEG differences, particularly in frontotemporal θ-band networks, characterized by increased connectivity metrics such as phase lag index (P = 0.000), etc. The predictive algorithm based on EEG features achieved accuracies between 59.2 %-84.6 % (AUC: 0.60–0.87). When compared to the whole brain, EEG features of the frontotemporal network showed improved classification performance in Naïve Bayes (P = 0.032), Tree Bagger (P = 0.021), and Subspace Discriminant (P = 0.022) models. The ensemble learning technique (Tree Bagger) delivered the best prediction results, achieving 91.5 % accuracy, 97 % sensitivity, 81 % specificity, and AUC of 0.92. Increased frontotemporal EEG connectivity was observed in TLE patients with 2-year DRE. A predictive model based on routine EEG provides an accessible method for forecasting ASMs efficacy. This study highlights the clinical utility of EEG-based algorithms in identifying DRE early, aiding personalized treatment strategies and improving patient outcomes.