Genetic algorithms for feature selection when classifying severe chronic disorders of consciousness

• Published in: PLOS ONE Vol. 14; no. 7; p. e0219683
• Main Authors: Wutzl, Betty, Leibnitz, Kenji, Rattay, Frank, Kronbichler, Martin, Murata, Masayuki, Golaszewski, Stefan Martin
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science (PLoS) 11.07.2019; Public Library of Science
• Subjects: Adult; Aged; Algorithms; Artificial intelligence; Biology and Life Sciences; Brain; Brain - diagnostic imaging; Brain - physiopathology; Brain damage; Classification; Coma; Communications technology; Computer and Information Sciences; Consciousness; Consciousness Disorders; Consciousness Disorders - diagnostic imaging; Consciousness Disorders - physiopathology; Diagnosis; Diagnostic errors; Diagnostic imaging; Disorders; Engineering and Technology; Female; Functional magnetic resonance imaging; Genetic algorithms; Genetic aspects; Health aspects; Humans; Information science; International conferences; Machine Learning; Magnetic resonance; Magnetic Resonance Imaging; Male; Medical diagnosis; Medical imaging; Medical prognosis; Medicine; Medicine and Health Sciences; Middle Aged; Neural networks; Neurology; Neurosciences; NMR; Nuclear magnetic resonance; Patients; Physical Sciences; Principal components analysis; Prognosis; Q; R; Research and Analysis Methods; Research Article; Science; Sleep; Support Vector Machine; Support vector machines; Systematic review; Unconsciousness; Austria; United States > US; Japan
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0219683

The diagnosis and prognosis of patients with severe chronic disorders of consciousness are still challenging issues and a high rate of misdiagnosis is evident. Hence, new tools are needed for an accurate diagnosis, which will also have an impact on the prognosis. In recent years, functional Magnetic Resonance Imaging (fMRI) has been gaining more and more importance when diagnosing this patient group. Especially resting state scans, i.e., an examination when the patient does not perform any task in particular, seems to be promising for these patient groups. After preprocessing the resting state fMRI data with a standard pipeline, we extracted the correlation matrices of 132 regions of interest. The aim was to find the regions of interest which contributed most to the distinction between the different patient groups and healthy controls. We performed feature selection using a genetic algorithm and a support vector machine. Moreover, we show by using only those regions of interest for classification that are most often selected by our algorithm, we get a much better performance of the classifier.