Enhancing COVID-19 Diagnosis: A Multi-Modal Approach Utilizing the CNN Algorithm in Automated Applications

Abstract

Rapidly identifying COVID-19 patients is essential for effective disease control and management. To address this need, we have developed an automated application that utilizes multi-modal data, including Chest X-ray, Electrocardiogram (ECG), and CT scan images, for precise COVID-19 patient identification. This application comprises a two-stage process, starting with a web-based questionnaire and then the submission of medical images for verification. Leveraging various ML and DL techniques, including CNN, KNN, Logistic Regression, Decision Tree, and NaiveBayes, We conducted extensive model training and validation for LSTM, InceptionV3, SVM, Resnet, and MobileNet. The CNN algorithm emerged as the top-performing method across all modalities, demonstrating exceptional accuracy, precision, recall, F-score, and a minimal false prediction rate. Confusion matrices were employed for comprehensive result evaluation. This study highlights the potential of multi-modal data analysis, particularly the CNN algorithm, for efficiently and accurately identifying COVID-19 patients.