COVID-19 Infection Detection in X-Ray Images Using Enhanced Image Segmentation and Machine Learning Models

The COVID-19 pandemic has shown a significant challenge to healthcare associations worldwide demanding the progress of reliable diagnostic techniques. Chest X-ray imaging is an easily available and cost-effective tool for evaluating lung defects combined with COVID-19. This study aims to introduce a novel COVID-19 infection detection and recognition method in X-ray images through image segmentation. By employing machine learning (ML) methods, particularly CNN-based U-Net architectures, the lung sections are accurately segmented, and infection sections are classified based on specific radiological features. The model was trained on a large dataset of annotated X-ray images assisting it to distinguish between healthy lungs, COVID-19-affected lungs, and other lung pathologies. The proposed method apprehended a segmentation accuracy of 96.6% with a Dice coefficient of 94.36% offering a potential solution for mechanical detection and 98.4% with KNN analysis of COVID-19 infection in chest X-rays in that way assisting radiologists in making faster and more accurate diagnoses.