India's Central TB Division has deployed the AI tool DeepCXR to flag tuberculosis on chest X-rays across 8 states during the 2024-25 TB Mukt Bharat campaign, though a policy-watch source flags it lacks published validation studies.
54,000 chest X-ray images
Training dataset size
<1 minute per X-ray
Result turnaround time
8 states/UTs
States/Union Territories deployed
347 districts
100-day campaign priority districts covered (campaign-wide, not DeepCXR-specific) (100-day drive from 7 Dec 2024)
~130 million people (12.97 crore)
100-day campaign people screened (campaign-wide, not DeepCXR-specific) (100-day drive)
719,000 cases (7.19 lakh)
100-day campaign new TB cases detected (campaign-wide, not DeepCXR-specific) (100-day drive)
Details
Maturity
Pilot
Promoter
Central TB Division / National Tuberculosis Elimination Programme (NTEP), Ministry of Health and Family Welfare, Government of India; tool developed by the Institute for Plasma Research, Gandhinagar
Period
Deployed under NTEP from 2024; used during the 100-day TB Mukt Bharat Abhiyan campaign, 7 December 2024 - March 2025
Keywords
AI-assisted health screening, tuberculosis diagnosis, computer-aided detection, public health
Context
India's Central TB Division has formally recommended DeepCXR, an AI chest X-ray interpretation tool developed by the Institute for Plasma Research (Gandhinagar), for use under the National Tuberculosis Elimination Programme. Trained on roughly 54,000 X-ray images, the tool flags TB-related lung abnormalities in under a minute and is intended to ease reliance on scarce radiologists in resource-limited public facilities.
Objectives
The tool aims to speed up and extend TB screening capacity in government facilities lacking radiologists, supporting the wider 100-day 'TB Mukt Bharat Abhiyan' intensified case-finding campaign.
Activities
DeepCXR was deployed free of charge across eight states and Union Territories to analyse X-rays of presumptive TB cases, used alongside AI-enabled handheld X-ray units and a separate AI tool automating Line Probe Assay drug-resistance readouts, during the 100-day campaign (from 7 December 2024) that covered 347 priority districts.
Results
The 100-day campaign as a whole screened approximately 12.97 crore (130 million) people and detected 7.19 lakh (719,000) new TB cases plus 2.85 lakh asymptomatic cases nationally; these are campaign-wide totals across all screening methods, not figures isolated to DeepCXR specifically.
Conclusions
DeepCXR is a genuine, government-endorsed, multi-state AI deployment, but its own diagnostic performance remains unpublished, unlike a competing commercial tool (Qure.ai's qXR), and rollout communication to all states was incomplete as of the source review.
Implementation
Indicative cost
Low (< €50k) — No public price or budget figures are disclosed for DeepCXR itself; it is provided free to government facilities, so cost is estimated as low, though the encompassing 100-day national campaign involved much larger, unquantified public health spending.
Time to results
Short (< 1 year) — Deployed within the 100-day intensified 'TB Mukt Bharat Abhiyan' campaign starting 7 December 2024; parliamentary reporting followed in March 2025.
Staffing & skills
Institute for Plasma Research, Gandhinagar developed the tool, Central TB Division / NTEP, Ministry of Health and Family Welfare formally recommends and rolls it out, Deployed alongside AI-enabled handheld X-ray units at facility level
Conditions for success
Provided free of charge to government facilities, removing a licensing-cost barrier to adoption
Bundled with handheld X-ray units and a companion drug-resistance AI tool as part of the same intensified campaign
Common failure modes
No published peer-reviewed validation study of sensitivity/specificity, unlike a comparable commercial tool
Central TB Division had not formally communicated the rollout to all states as of the review
Campaign-wide screening/detection totals are not separable into DeepCXR's specific contribution
Where it fits
Governance type
national government health programme
Scale
multi-state (8 states/UTs) within a national campaign
Income level
lower-middle-income
Data sources
Where this practice's information was retrieved from, and when.