AIIMS Tuberculosis Detection
A cross-platform clinical interface that predicts the likelihood of Crohn's disease versus intestinal tuberculosis.
.png)
Crohn's disease and intestinal tuberculosis present so similarly that distinguishing them is a genuine diagnostic problem — and the treatments diverge sharply. Getting it wrong costs patients months.
This was the applied arm of the research I did at TavLab under Dr. Tavpritesh Sethi, alongside the work that became the ECCO'25 publication. It leverages machine learning over patient inputs to predict which of the two conditions is more likely, delivered through a cross-platform interface clinicians could actually use rather than a notebook.
-
Model over structured patient inputs
Analyzed user-supplied clinical inputs with machine learning algorithms to produce a likelihood across the two conditions.
-
Cross-platform delivery
Built the interface in Flutter so the same tool ran on Android and iOS, putting the model in front of clinicians instead of leaving it in a research environment.
-
Part of a year-long interdisciplinary study
Fed into the wider TavLab work on AI-driven modeling of inflammatory bowel disease progression, co-authored as a peer-reviewed publication in the Journal of Crohn's and Colitis (ECCO'25).