A granular functional network classifier for brain diseases analysis

ABSTRACT The development of interpretable and readable diagnosis support models in the medical field is becoming an active research area. Neuroimaging technology has been widely used in the study of various brain diseases, supported by several kind of machine learning algorithms. Such algorithms, in spite of their accuracy, have often a lack of interpretability. In order to address these issues, in this paper, a new classifier based on functional networks, revised from a granular perspective is proposed. Granular architectures allow high accuracy, without losing interpretability. Two classes of neurodevelopmental disorders, that is Attention Deficit Hyperactivity Disorder and Autism Spectrum Disorder, will be considered to perform numerical experiments on publicly available data. The numerical results against state-of-the-art methods show the good performance of the proposed scheme, by confirming some theoretical achievements.