Hybrid deep learning-based model for covid-19 prediction and interpretation using multiple data modalities
Date
2024-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Federal University of Technology, Owerri
Abstract
This research addresses the critical need for accurate and timely COVID-19 diagnosis and prognosis by developing a hybrid deep learning model that integrates multiple data modalities, including chest X-rays, Computed Tomography (CT) scans, blood smears, and clinical data. The model employs specialized architectures such as Residual Network with 50 Layers (ResNet50) for Chest X-ray, InceptionV3 for CT scans, Convolutional Neural Network (CNN) for blood smears, and a Random Forest classifier for clinical data analysis. The results demonstrate high accuracy rates: 96.7% for ResNet50, 97.58% for InceptionV3, 96.12% for CNN, and 98.30% for the Random Forest classifier. Grad-CAM enhances transparency by visualizing critical regions in the images, aiding healthcare professionals in understanding the model's decisions. This hybrid model offers improved accuracy and reliability for COVID-19 diagnosis and prognosis, making it a valuable tool for clinical settings and resource allocation. The research underscores the potential of multi-modal data integration in medical AI and suggests further exploration and refinement of such models for broader healthcare applications.
Description
The dissertation has tables and figures
Keywords
Deep learning, ML, explainable AI, classification, covid-19, grad-cam, Department of Information Management Technology
Citation
Dokun, O. (2024). Hybrid deep learning-based model for covid-19 prediction and interpretation using multiple data modalities (Unpublished Master's Thesis). Federal University of Technology, Owerri, Nigeria