Predicting AMR in Tuberculosis Using Deep Learning

About the Technology

Antimicrobial Resistance (AMR) is a growing threat in treating Tuberculosis (TB) in developing nations. The traditional drug-susceptibility testing process is time-consuming, delaying critical treatments.

This research introduces novel sequence encoding techniques to predict antimicrobial resistance in Mycobacterium tuberculosis. Using whole genome sequencing data, a custom deep learning model evaluates susceptibility patterns rapidly. To bring this research directly to clinical settings, the AI model has been integrated onto a portable Jetson Nano hardware device, offering clinical utility right at the point-of-care.

Novel Features

Unlike cloud-dependent systems, this solution is a completely portable edge AI application running locally on a Jetson Nano. The model integrates Explainable AI (XAI) frameworks, providing doctors with logical explanations behind AMR predictions (showing which genetic markers led to a resistance classification).

National & Global Impact

Tuberculosis imposes a massive economic and public health burden on India, costing an estimated $24 billion annually in terms of lost productivity and clinical expenses. A portable, rapid AMR predictor allows clinicians to select effective first-line or second-line drugs instantly, curbing the spread of drug-resistant strains and improving patient survival rates.