Skip to main navigation menu Skip to main content Skip to site footer

Vol. 10 No. 1 (2024)
Special Issue: Proceedings of CVIS 2024

Articles

Enhancing Parkinson’s Disease Diagnosis through Synthetic Image Augmentation and Deep Learning Model Evaluation

https://doi.org/10.15353/jcvis.v10i1.10000

Published 2025-02-19

  • Mosarrat Rumman
    • ,
  • Heidar Davoudi
    • ,
  • Mehran Ebrahimi

Abstract

Parkinson’s disease (PD) is a progressive neurodegenerative disorder that can be clinically diagnosed through various neuroimaging techniques. Single-photon emission computed tomography (SPECT) has proven to be an effective tool for the early detection of PD. Automatic detection of PD from SPECT images, using machine learning or deep learning models is crucial for providing faster, more accurate diagnoses, and facilitating early intervention. While large datasets of SPECT scans for PD are available, they are often highly imbalanced, which can significantly hinder the performance of deep learning models. In this paper, we explore how synthetic image generation can address the dataset imbalance problem and improve the accuracy of deep learning models. We evaluated the performance of several state-of-the-art pre-trained deep learning models, including Vision Transformer (ViT), VGG-16, EfficientNet, and a newly proposed hybrid model, Inception-VGG16. Experimental results demonstrate that augmenting the dataset with synthetic images significantly improves the performance of all models, with ViT achieving the highest test accuracy of 98\%. The proposed Inception-VGG16 model performed second best, achieving a test accuracy of 95\%. These results suggest that synthetic augmentation can enhance the performance of pre-trained models in detecting Parkinson's disease, presenting a promising approach for enhancing automatic diagnostic tools. The implementation of this work is available at https://github.com/mosarrat28/Parkinsons_disease_detection_using_SPECT_image.
PDF