Point screening at the neighbourhoods most likely to be missed
AI-ACF turns the data a tuberculosis programme already collects into a ranked list of places to screen. Teams walk fewer streets to find the people a broad sweep is most likely to miss.
A visit list, not a heat map
Routine notification data from Nikshay, together with population, environment, and building information, is scored for every small area of a district. The result is an ordered list of neighbourhoods a team can walk, with a name, the local language, imagery, and directions.
Four steps from district data to a walkable plan
A supervisor sees the district. A worker sees the next door.
- 01
Score each small area
Notifications, population, environment, and building footprints produce a risk score for every small area, including settlements that block-level plans often skip.
- 02
Rank by expected yield
Areas are ordered by how many cases a visit is expected to find, so a fixed amount of screening goes to the places most likely to repay it.
- 03
Group into a field plan
The highest-priority areas are gathered into clusters a supervisor can review before anyone leaves the office.
- 04
Issue a named visit list
The confirmed plan goes to a worker's device as places to visit, in order, with imagery and turn-by-turn directions.
The same plan, read two ways
Before the campaign
District supervisors
One view of the district, ranked by expected yield, ready to assign.
See which neighbourhoods rise to the top before teams go out.
Review the plan, then send it to the worker who will walk it.
On the round
Frontline workers
The day's work is a list of real places, not a shaded map.
Each stop has a place name in the local language.
Satellite imagery, photographs, and directions sit with that stop.
Pune Pilot Study Results
AI-NETRA is a targeting strategy - not more screening effort. In a government-partnered evaluation across four TB Units in Pune, AI-guided active case finding concentrated teams where transmission risk is highest, finding confirmed cases at nearly 3× the standard rate while screening about one-third as many people.
Notified cases found from under one-third of the screened population
Confirmed pulmonary TB per person screened vs standard ACF
6,610 people screened per confirmed case - down from 19,702
Down from Rs 27.6 lakh - roughly two-thirds lower
What does a team receive in the field?
An ordered list of places to visit. Each place has a name, guidance in the local language, imagery, and directions.
What data does AI-ACF use?
Routine Nikshay notification records, population and environment layers, building information, and feedback from frontline workers. Records used by the platform are deidentified.
Does it replace a screening campaign?
No. It decides where a campaign should go, so the same field effort is spent in higher-yield neighbourhoods.
Where can I read the evidence?
The Pune pilot study page holds the methods, the comparison with standard active case finding, and the limits of that evidence.