Bengaluru : In a new study conducted by researchers at the Indian Institute of Science (IISc), infectious bacteria that can spread diseases can be identified by using artificial intelligence (AI) with Raman spectroscopy, a chemical analysis technique that provides information about chemical structures, molecular interactions, crystallinity, phase and polymorphy.

In Raman spectroscopy, a laser beam is flashed on a clinical sample through an optical microscope. The light is scattered by the bacterium, giving it a unique spectrum. A deep learning model called the Residual Network (ResNet) is deployed to identify and differentiate between the spectra of eight bacterial species that can spread through hospital-acquired infections. The researchers were able to train the model through a process called transfer learning which saw researchers using limited clinical data to teach the model to identify different different bacterial species. The model has been 99.99% accurate when identifying bacterial species.

Studies like this show how AI is useful in reducing time needed for diagnosis in patients and that treatment and management of bacterial infections can be done more quickly. Current identification tests are either time consuming or only work with certain types of clinical samples.

The teams working on this study were led by Siva Umapathy (Director of the IISc) and Deepak Kumar Saini (an associate professor of IISc), who have been working for many years to combine Raman Spectroscopy with AI for applications in healthcare and diagnostics. They have shown that the combination can be used to recognize COVID-19 biomarkers in blood samples, which won them the Challenger Award at the NASSOM AI Game Changer Awards Programme in 2021.

Click Here for More Science & Technology