IIT Jodhpur researchers develop COVID-19 diagnosis technique using chest X-ray images

The researchers from IIT Jodhpur have developed an automated AI solution for COVID-19 screening. The experiment was performed with more than 2500 chest X-ray images and achieved about 96.80 per cent sensitivity

With the increasing number of COVID-19 cases across various waves around the globe, countries have faced challenges with the limited availability of testing kits and processing centres in remote areas. This has been the key motivation for researchers to find out the alternate methods of testing which are reliable, easily accessible, and faster. Since the symptoms of COVID-19 are visible on Chest X-Ray, it has become one of the modalities that have gained acceptance as a screening technique. To cater to this problem, IIT Jodhpur researchers have developed an explainable Artificial Intelligence solution for predicting COVID-19 from chest x-rays.

The researchers at IIT Jodhpur have proposed a deep learning-based algorithm called COMiT-Net, which learns the abnormalities present in the chest X-Ray images to differentiate between a COVID-19 affected lung and a non-COVID affected lung. The developed AI algorithm not only predicts whether the CXR has COVID-19 pneumonia or not, but it is also capable of identifying the infected regions in the lungs, thus making them explainable.

While there have been numerous research studies in COVID-19 detection using X-ray or CT scans in the past year, most of them fail to provide an explainable solution. The uniqueness of this research and its salient features are as follows:

• The proposed study can visually showcase the region which is infected.
• The technique interprets only from the lung region.
• The Artificial Intelligence solution used in this research is explainable from both algorithmic and medical points of view.

The team that contributed to this research are Aakarsh Malhotra, Visiting Research Scholar at IIT Jodhpur, Surbhi Mittal, PhD Scholar, Computer Science, IIT Jodhpur, Puspita Majumdar, Visiting Research Scholar at IIT Jodhpur, Saheb Chhabra, Visiting Research Scholar at IIT Jodhpur, Kartik Thakral, PhD Scholar, Computer Science, IIT Jodhpur, Mayank Vatsa, Professor, Computer Science, IIT Jodhpur, Richa Singh, Professor and Head of Department, Computer Science, IIT Jodhpur, Santanu Chaudhury, Professor and Director, IIT Jodhpur, Ashwin Pudrod, Consultant Pulmonologist, Ashwini Hospital and Ramakant Heart Care Centre, India, Anjali Agrawal, Consultant Radiologist, TeleRadiology Solutions, India.

A research paper on this project has been published in the journal “Pattern Recognition (Volume 122)”. The research is part of the RAKSHAK project under NM-CPS DST and iHuB Drishti at IIT Jodhpur. The researchers aim at developing a full-scale prototype through the knowledge learnt in this project.