Schizophrenia Detection using Electroencephalography Signals Using Deep Learning

The objective of the project is to see whether deep learning can be used to identify schizophrenia in teenagers by analysing their EEG data. The work takes advantage of both conventional and novel preprocessing techniques, such as the Short-Time Fourier Transform, to extract characteristic signals associated with schizophrenia from EEG data. The study highlights encouraging results with models including VGG-16, Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN)[24], and Pyramidal Spatial-based Feature Attention Network (PSFAN). Notably, PSFAN presents a novel diagnostic methodology that extracts multiscale features from 4-second EEG recordings and converts them into 2-dimensional images using dilated convolutions. On two datasets, custom CNN shows statistical superiority over eleven state-of-the-art algorithms through subject-dependent, subject-independent, and cross-dataset experiments [16]. This paper highlights the usefulness of deep learning in medical settings and highlights its potential.