Nine AI-equipped portable X-ray units let Kyrgyzstan's National TB Programme screen over 10,125 people in rural, hard-to-reach areas in under a year, finding 51 new tuberculosis cases and linking them to treatment.
7900
Estimated new TB cases in Kyrgyzstan (2023)
~40 %
TB cases never diagnosed or notified (2023)
9 units
Portable AI X-ray units deployed
~300 screenings/day
Screening capacity per device
10125 people
People screened in the pilot (under one year)
51 people
People identified with TB and connected to treatment
Details
Maturity
Scaling
Promoter
National Tuberculosis Programme, Ministry of Health of the Kyrgyz Republic, with UNDP and the Global Fund
Period
2023-2025
Keywords
public health, healthcare delivery, disease surveillance
Context
Of an estimated 7,900 people who contracted tuberculosis in Kyrgyzstan in 2023, around 40% were never diagnosed or notified — a gap driven partly by a scarcity of specialist radiologists able to reach remote rural communities.
Objectives
Kyrgyzstan's National TB Programme, working with UNDP and financing from the Global Fund, set out to close the case-detection gap by deploying AI-assisted portable chest X-ray screening that does not depend on an on-site specialist radiologist.
Activities
The programme deployed nine InferAir ultra-portable X-ray machines fitted with Infervision's AI-based chest X-ray reading software (InferRead) in regional TB control centres and the Bishkek City TB Hospital. Because the AI system flags likely TB, pneumonia, cardiac and chronic-lung findings automatically, the lightweight units can be carried into remote rural communities, and each device can process up to roughly 300 screenings a day.
Results
In under one year the pilot screened more than 10,125 people, identifying 51 people with tuberculosis who were then connected to treatment. Following the pilot's results, Kyrgyzstan's National TB Programme has moved to procure additional Infervision units, with further onsite deployment and staff training continuing through 2024 and 2025.
Conclusions
The screening and case-detection counts are independently reported by UNDP as well as by the equipment vendor, but no published third-party clinical validation of the AI's diagnostic accuracy in the Kyrgyz deployment (e.g. sensitivity/specificity against confirmed diagnoses) is yet available — a real gap given how central the AI reading step is to the whole approach.
Implementation
Indicative cost
Medium (€50k–€500k) — Financed by the Global Fund with UNDP support; no unit-cost or total-budget figures are published for the nine InferAir/InferRead deployments.
Time to results
Medium (1–3 years) — The pilot screened over 10,125 people and identified 51 TB cases in under one year; further unit procurement, deployment and staff training are continuing through 2024 and 2025.
Staffing & skills
National TB Programme (Ministry of Health of the Kyrgyz Republic) implementation, with UNDP support and Global Fund financing, Staff training for onsite deployment continuing through 2024 and 2025, Portable units usable without an on-site specialist radiologist due to automated AI reading
Conditions for success
AI-based automated reading (InferRead) removes reliance on scarce specialist radiologists in remote areas
Ultra-portable device design (InferAir) allows carrying into remote rural communities
Multi-partner financing and implementation structure (national programme, UNDP and Global Fund)
Common failure modes
No published third-party clinical validation of the AI's diagnostic accuracy (sensitivity/specificity) in the Kyrgyz deployment
Screening and case-detection counts are reported by UNDP and the equipment vendor, not by an independent evaluator
Where it fits
Governance type
national health programme with international donor/UN support
Scale
national (Kyrgyzstan), rural and regional TB centres plus Bishkek City TB Hospital
Income level
lower-middle income
Commonly funded by
National / regional programmes
Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.
Do you run this practice?
Claim it —
verified implementers get a public contact pathway and can propose corrections.
Data sources
Where this practice's information was retrieved from, and when.