Abstract: The accurate detection and classification of schizophrenia are vital for early intervention and treatment. The realm of medical image analysis encounters a hurdle due to the scarcity of publicly accessible data, often compelling researchers to grapple with small and imbalanced datasets. To address this issue, transfer learning techniques come to the rescue by allowing the utilization of general features from smaller target datasets. This research investigates the application of the EfficientNets for the classification of Schizophrenia rs-fmri images obtained from schizconnect. The dataset is split into three sets with 60% for training, 20% for testing and the remaining 20% for evaluation The results indicate promising classification performance for different EfficientNet convnet variants, with high precision, recall, F1-score, and accuracy. The B3 architecture, in particular, demonstrates exceptional performance, achieving 99.7% accuracy. The findings of this research provide valuable insights into the classification of schizophrenia using neuroimaging data and state-of-the-art neural network models.
Keywords: EfficientNet, Schizophrenia, ConvNets. Transfer Learning.
Title: A Deep Learning Approach to Classifying Schizophrenia Based on EfficientNet Convolutional Neural Network Models
Author: Michael Ayeni, Prof. S.O Adewale, Dr. Ibam
International Journal of Computer Science and Information Technology Research
ISSN 2348-1196 (print), ISSN 2348-120X (online)
Vol. 11, Issue 4, October 2023 - December 2023
Page No: 72-79
Research Publish Journals
Website: www.researchpublish.com
Published Date: 15-December-2023
