Cure8 research brief
Distinguishing Crohn’s disease from intestinal tuberculosis is a frequent problem in TB-endemic areas; a validated mobile ML tool could help clinicians make more accurate, faster diagnostic decisions and avoid inappropriate therapies.
Clinicians and gastroenterologists working in TB-endemic regions, researchers in AI and diagnostic tools, and adult patients being evaluated for suspected Crohn’s disease vs intestinal tuberculosis.
Researchers trained gradient-boosted decision-tree models (CatBoost among others) on 30 clinical, endoscopic and radiological variables from a retrospective cohort (n=1,066) and integrated the lead model into a mobile app.
The model was prospectively validated in 121 patients at the same centre and reported high diagnostic accuracy (AUROC 0.921) with balanced sensitivity and specificity (~86% and ~85% at the chosen threshold).
The study used explainable-AI (SHAP) methods to identify the most influential features (for example, symptom duration, transverse ulcers, granuloma, rectosigmoid involvement, and perianal disease). The tool is intended as point-of-care decision support to reduce misdiagnosis between ITB and CD in resource-limited, TB-endemic settings.
The report is an abstract-level summary from a peer-reviewed journal; the brief is grounded in that abstract rather than a full independent appraisal.
This classification and brief are based on the article abstract supplied by the journal. The validation cohort was from the same tertiary centre as the training data; additional external validation in other settings would strengthen generalisability. As an abstract-level source, the brief does not replace full-text critical appraisal.
Review the original publication for the complete reporting, methods, and context.
Funding disclosed by the source: Indian Council of Medical Research - 55/4/11/CARE-ID/2018-NCD-III
This Cure8 brief is based on source text from the linked article. Cure8 is informational only and is not a substitute for professional medical advice, diagnosis, or treatment.
A validated AI biomarker that predicts which IBD treatments are likely to work could reduce trial-and-error in treatment choices, potentially shortening time to effective therapy for people with Crohn's disease or...
This project aims to improve prediction of autoimmune disease onset and immunotherapy-related autoimmunity by combining genetics and plasma proteins.
Genetic ancestry can influence IBD risk and disease course, and most prior studies focused on European populations.
