Pneumonia Detection using Ensemble of Modified ViT-YOLO Models

Pneumonia, an acute respiratory disease that targets the lungs, can occur in people of all age groups. Its severity can range from mild to life-threatening, hence encouraging research for new methods of detection and cure. Since decades, the detection of Pneumonia has been manual, and involves a set of medical professionals to analyze Chest X-Rays (CXR) and carry out laboratory tests along with a detailed clinical study of patient’s history before the treatment. Pneumonia is also seen to occur as a complication of COVID-19 infection. Hence, in the recent years, Pneumonia cases have significantly risen making it time-consuming and tedious job for clinicians to analyze and detect it. Therefore, this study presents a unique and robust end-to-end pipeline that can automate the process of detection of Pneumonia in the chest radiographs. This computer vision-based pipeline not only speeds up the process, but also reports results with an mAP@0.5 of 0.7, thus outperforming the current state-of-the-art methods of Pneumonia detection and classification.