The Problem
Pneumonia is the leading cause of death among children. However, it receives less than 2% of the global development fund for health. It is also more prevalent in countries where there is insufficient medical expertise and lack of access to radiology diagnostics. Chest x-rays are the most common tool used to diagnose pneumonia. However, understanding chest X-rays require domain expertise and professional radiologists. We applied machine learning so that a computer can be used to detect signs of pneumonia given a chest x-ray, increasing the ease of access to resources for pneumonia detection.
What we did
We successfully implemented and compared three machine learning models: YOLOv3, RetinaNet and Mask RCNN. We tested and compared their effectiveness in recognising pneumonia from chest x-rays. Our key finding was that an ensemble of RetinaNet models supporting different input shapes outperformed other models, achieving the highest mean Average Precision (mAP) of 48.1%.
We presented this work at the 13th National University of Singapore (NUS) School of Computing Term Project Showcase (STEPs).
