Gandhinagar: Researchers at the Indian Institute of Technology Gandhinagar (IITGN) have developed an Artificial Intelligence (AI)-based deep learning tool for detection of COVID-19 from Chest X-ray images. This online tool indicates the probability if a person is infected with COVID-19, which can be used for quick preliminary diagnosis before the medical test.
The tool is developed by Kushpal Singh Yadav (PI), an M.Tech student of Computer Science & Engineering at IITGN, under the guidance of Prof Krishna Prasad Miyapuram, Associate Professor of Cognitive Science and Computer Science at IITGN.
The system is now live as a web-interface on http://covidxray.iitgn.ac.in/. Any user can upload a digital copy of the chest x-ray or CT-scan in standard image formats such as jpeg, png etc. and check whether that person is COVID-19 positive or not. The format of any test image of a new person will be automatically transformed and diagnosed using the AI tool. It will also validate the input images before giving a result. The result is available within a few seconds.
Sharing the motivation and idea behind developing the tool Prof Krishna Prasad Miyapuram and Kushpal Singh Yadav said, “Given the limited testing facilities for COVID19, there is a rush to develop AI tools for quick analysis using x-rays. Developing a reliable tool requires the combination of right algorithms and data. This is where our tool would prove useful, that can be trained for diagnostic purposes and made available for wider use.”
How the AI-based tool analyses X-rays:
AI tools are widely used for medical image diagnosis. The researchers pooled the data of X-ray images of COVID-19 infected patients as well as healthy persons from different sources available on the internet. They trained a machine learning architecture using deep learning algorithms with these images. The deep learning model that IITGN researchers used has 12 layers of neural network, which is similar to the neurons in the human brain. The deep learning method has the advantage that it learns the disease diagnosing features from the x-ray images in an automatic way.
The tool developed by the IITGN team also uses images from other lung infections such as tuberculosis, pneumonia to ensure the specificity of detection of COVID from other lung diseases. This tool out-performs other such high-tech tools available globally. Moreover, it uses simple machine learning architecture, which makes it stand out over others.
IITGN researchers are collaborating with Dr Deepak Saxena from the Indian Institute of Public Health (IIPH), Gandhinagar for further testing of the AI-based tool.
In these difficult times, when the diagnosis of COVID-19 is a challenging task, and there are limited authorised testing facilities, this tool can cater to a large population. It can essentially reduce the burden on medical infrastructure by aiding testing and diagnosis with preliminary results. The tool can be used by medical practitioners while they are waiting for the test results. Public users should be aware that the AI tool is only indicative and clinical consultation is essential to confirm the diagnosis.