Since previous machine-learning methods for the detection of schizophrenia often have a high degree of complexity, require a lot of data, channels and are computationally intensive, we developed a novel, fast and reliable classifier for schizophrenia. For this, we only use one-minute recording time and a fine granular division of resting-state EEG spectra. Based on that, we identify the five most predictive electro-encephalography (EEG) channels (F3, T5, T6, Fz, and T4), reducing the amount of non-discriminant features. Using Random Forest and majority voting, we detect schizophrenia based on those five channels. By maximizing the information value of each input feature, our classification approach achieves a balanced accuracy of 97.71% in schizophrenia classification. Our classifier is cost- and time-efficient, and easy to implement into daily medical life. Furthermore, new insights about the rele-vance of the right temporal lobe were gained.
Baumgartl, Hermann; Scholtz, Stefanie; Sauter, Daniel; and Buettner, Ricardo, "Detection of Schizophrenia Using Machine Learning on the Five Most Predictive EEG-Channels" (2021). PACIS 2021 Proceedings. 38.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.