Non-Invasive Neuroimaging based Automated Epilepsy and Major Depressive Disorder Detection using N-cylinder and Riemannian Geometry based Feature Engineering and Machine Learning Techniques

Abstract

Major Depressive Disorder (MDD) and Epilepsy are distinct neurological disorders that significantly impact individuals’ well-being and cognitive functioning. MDD is characterized by persistent sadness, hopelessness, and a loss of interest in daily activi ties, while epilepsy is marked by recurrent seizures resulting from abnormal electrical activity in the brain. Detecting and classifying these disorders accurately is crucial for effective diagnosis and treatment planning. This study employs advanced computa tional techniques to extract informative features and classify MDD and healthy subjects using the Riemannian Geometry method using Magnetoencephalogram (MEG) signal biomarkers. The collected MEG data undergoes a comprehensive pipeline consisting of feature extraction and classification stages. Epilepsy classification using Electroen cephalogram (EEG) signal biomarkers is also addressed in this research, aiming to en hance the performance of the classification task. Different feature combinations, such as N-cylinder, FFT, DWT, Kurtosis, LBP, and CSP, are explored to improve the ac curacy and reliability of identifying epilepsy and healthy subjects. Various classifiers, including Random Forest, XGBoost, Logistic Regression, SVM, and KNN, are evalu ated using 5-fold cross-validation and parameter tuning. The results demonstrate promising outcomes, with the Random Forest classifier achieving the highest accuracy, the area under the curve (AUC), and sensitivity per centages among the tested classifiers. The combination of DWT, FFT, and N-cylinder features yields the best performance for epilepsy classification. The research findings contribute to developing practical diagnostic tools for MDD and epilepsy, providing valuable insights for healthcare professionals and researchers. In conclusion, this study showcases the application of computational methods in classifying MDD and epilepsy. By leveraging advanced feature extraction techniques and evaluating various feature combinations, accurate identification and differentiation between MDD and healthy subjects, as well as epilepsy patients, can be achieved. The proposed approach could improve the diagnosis and management of these neurological disorders. Keywords: Major Depressive Disorder, Epilepsy, Riemannian Geometry, Feature Extraction, EEG, MEG, N-cylinder.