Image & Signal Processing

Processing the chest X-ray scan to screen for COVID-19 using deep convolutional neural networks

  • From September 2020

  • Project Status: On-Going

Instead of 'DNA probes' or 'Immunocomplexes,' a tell-tale of X ray-images could save you both time & money in detecting COVID-19.

The first step towards any treatment is figuring out the disease and the causative agent. It is easier said than done, as many diseases have overlapping signs and symptoms. The visible effects (Signs) specific to a particular disease, may contradict what the patient says he/she is experiencing (Symptoms). A 'right' diagnosis can differentiate a stress-related migraine from an allergy-triggered one or identify the underlying cause for a respiratory discomfort - be it pollen or virus. The diagnosis comes at a cost; however, in case of a pandemic, especially a novel strain as COVID-19, the money shelled out alone will trigger a migraine attack and respiratory block, all-in-one shot!

RT-PCR (detection based presence of unique gene) and Rapid Kit (detection based on the antigen-antibody complex) are the only diagnostic strategies for COVID-19, currently in the market. Both these kits differ in their steps, time-duration, sensitivity, and cost. An increase in the active cases and reduced human resources in manufacturing & transport are limiting the availability of these kits to the public. Deep-learning experts at Somaiya Vidyavihar University are teaming up to provide a rapid and cheaper diagnosis using chest radiographs. "Using three-layered deep convolutional neural networks, we can teach the machine to positively identify the COVID-19 patients from other respiratory disorders," says Dr. Ninad Mehendale, the lead investigator of this project.

The team at Somaiya Vidyavihar University has developed an indigenous algorithm that could identify the positive COVID case with 99% accuracy, solely on their radiographs. The next phase of the project is more learning-intensive and aims to (1) distinguish radiographs of COVID-19 from other pulmonary diseases, and (2) categorize chest radiographs based on the extent of COVID-19 infection.

Resource-constraints and the discrepancy in cost per kit charges between public and private labs are adding financial and mental burden on patients and their families. X-ray machines are integral to any hospital set-up, making it easy and accessible for tests. Detecting COVID-19 active cases through X-ray imaging would be a breather for all stakeholders in healthcare management.

Coordinator, Research Promotion-SIRAC

Somaiya Vidyavihar University

Published-19th June 2020

Principal Investigator

Co-Investigators

  • Ankita Shelke

    Python Programmer

  • Talha Chafekar

    Assistant Python Programmer

  • Amanshu Tiwari

    Assistant Python Programmer

  • Aafiya Hussain

    Technical Writer

  • Vruddhi Shah

    Project Manager

What to read next

Nutrition, Heath and Medicine
Nutrition, Heath and Medicine
Nutrition, Heath and Medicine
Nutrition, Heath and Medicine