Curriculum Neuroscience, technology, and society, XXXVII series  Grant sponsor Università degli Studi di Padova Supervisor s Alberto Testolin Co-Supervisor s Alessandro Sperduti Contact sina.shafiezadeh@studenti.unipd.it

Project description

During my Ph.D. I will investigate the use of deep learning techniques to implement multivariate analysis of neuroimaging and neurophysiological data, with the additional goal of obtaining explainable artificial intelligence models that can be used in clinical practice. In particular, I will initially focus on electroencephalogram (EEG) signals by developing efficient techniques for automatic detection and prediction of epileptic seizures. Epilepsy is characterized by abnormal electrical activities of the brain nerve cells resulting in recurrent seizures, unusual behaviour, and often loss of consciousness. It is the fourth most frequent neurological condition, affecting 70 million people worldwide, thus calling for the development of effective diagnostic tools. Recently, deep neural networks have attained promising results on this kind of task. DL-based methods do not require manual feature extraction and selection processes, and they allow to exploit multi-channel EEG recordings and image-based representations of EEG signals. I will apply deep learning techniques to predict seizures based on EEG signals collected on a wide group of epileptic patients, with a particular emphasis on designing explainable models that can inform clinicians about the most relevant features that characterize the abnormal EEG activity in epilepsy.