Challenges Of Ai Driven Diagnosis Of Chest X-rays Transmitted Through Smart Phones: A Case Study In Covid-19

This paper examines the challenges associated with AI-driven diagnostic tools for Chest X-rays (CXRs) transmitted through Smart Phones. It reveals that existing Deep Learning models suffer from prediction instability and explainability issues due to significant compression of high-resolution CXR images. Multi-Task Learning (MTL) is proposed as a solution to these problems, with improved robustness and explainability.