# Twelve millimetres, or three centimetres.

A breast cancer found by imaging is typically about 12 mm across. One found by a hand — a woman's own, or a clinician's — is typically about 21 mm. In India, where almost no one is screened, four out of five recorded tumours are already 3 cm or larger when they are found.

Canonical: https://brightaifuture.com/discoveries/medcognetics-mammography-workflow
Format: discovery
Source publication: 2025-12-11
Bright publication: 2026-09-09
Substantive update: 2026-09-12
Evidence and review: Emerging; confidence: unassessed; approved; ai-assisted. Re-reported from primary sources in September 2026. The regulatory status comes from FDA clearance documents, the Indian epidemiology from the ICMR national registry and NFHS-5, and the detection-size contrast from a published hospital series. The deployment account is published by the participating companies and is not independent verification. Bright first encountered this work through material curated by NVIDIA; the 'lives saved' figure in that material appears in no source and is not repeated here.

## The human problem

India has no organised breast screening programme. In NFHS-5, 0.9% of women aged 30 to 49 had ever had a clinical breast examination. The cancer is found late because nobody is looking early.

## The prior constraint

Mammography needs a machine, a building and a radiologist. Rural districts have none of the three within reach, and a screening programme creates more images than the available radiologists can read.

## AI’s actual role

MedCognetics' cleared software does one narrow thing: it flags an exam so that suspicious studies rise in the radiologist's reading queue. FDA cleared it for triage and worklist prioritization only. It does not mark lesions, does not remove a study from the queue, and must not be relied on to make or confirm a diagnosis. A radiologist still reads everything.

## The documented result

MedCognetics holds two FDA 510(k) clearances for mammography triage: K220080, cleared 29 September 2022, and K252482, cleared 11 December 2025. NVIDIA reported in October 2025 that the software runs on mobile screening vans operated by Health Within Reach in India, and that more than 3,500 women had been screened; that account is the companies' own.

## Why it may matter

Access, not cleverness, is the whole of this story. A van that reaches a district where 0.9% of women have ever been examined changes what stage the disease is found at — and stage is most of what survival depends on.

## Limitations

The screening numbers and any account of lives saved come from the participating companies. No independent evaluation, audit or named patient exists, and early detection is not the same as a life saved.

The cleared indication is triage only. It is not a diagnostic device, and its own labelling notes reduced performance for dense breasts and small lesions — common in the younger population it is deployed among.

The strongest trial evidence for AI in mammography, MASAI in Sweden, did not measure mortality and ran in a mature organised double-reading programme, a setting almost unlike rural India.

Screening only helps if treatment follows. In a documented Indian camp programme in Varanasi district, only 220 of 732 people who screened positive — 30.1% — completed follow-up, despite free transport and diagnostics.

## Unresolved questions

What share of women flagged by these vans reach diagnosis and treatment?

Does an earlier stage at detection here translate into survival, and who will measure it?

## Verified in the source

FDA cleared this software for mammography triage and worklist prioritization only, in 2022 and again in 2025; it is explicitly non-diagnostic. India's national hospital registry shows that among women with a recorded tumour size, 80.9% were already 3 cm or larger, and the last national survey found 0.9% of women aged 30 to 49 had ever had a clinical breast examination.

## Claimed / not established here

The van deployment, the number of women screened and any account of lives saved are published by the participating companies. No independent evaluation, audit or named patient exists. The supplied figure of 24 lives saved appears in no source Bright could find and is not carried.

## Bright editorial interpretation

Access, not cleverness, is the story: what changes is the stage at which the disease is found, and only if treatment follows.

## Independent support

FDA clearance documents define what the software is permitted to do. The ICMR national registry, NFHS-5 and a BMJ Open study of screening follow-up in Varanasi district supply the Indian context and the 30.1% follow-up completion rate independently of any participant.

## Provenance and history

{
  "dates": {
    "eventDate": null,
    "publicationDate": "2025-12-11",
    "captureDate": "2026-09-09",
    "lastReviewedDate": "2026-09-12"
  },
  "provenance": {
    "origin": "editorial",
    "externalId": "medcognetics-mammography-workflow"
  },
  "revisions": [
    {
      "id": "revision:d0731f2401c365d5e9ec",
      "recordedAt": "2026-09-09",
      "summary": "Added to Bright as a bounded source claim with visible evidence distinctions.",
      "sourceIds": [
        "source:medcognetics-mammography-workflow"
      ]
    },
    {
      "id": "revision:medcog-2026-09-12",
      "recordedAt": "2026-09-12",
      "summary": "Re-reported around access and tumour size at detection, with the FDA-cleared indication, the Indian registry evidence and the follow-up drop-off made explicit. The supplied 'saved 24 lives' figure was traced to no source and is not carried.",
      "sourceIds": [
        "source:fda-k252482",
        "source:ncdir-hbcr-2026",
        "source:guth-2008",
        "source:nfhs5-breast-exam",
        "source:varanasi-followup-2023"
      ]
    }
  ],
  "corrections": []
}

## Original sources

- [FDA 510(k) clearance K252482 — CogNet AI-MT+ (MedCognetics)](https://www.accessdata.fda.gov/cdrh_docs/pdf25/K252482.pdf)
- [Metastatic presentation of breast cancer in India: evidence from the national cancer registry (2009–2020) · Lancet Regional Health SE Asia](https://pmc.ncbi.nlm.nih.gov/articles/PMC13092702/)
- [Tumor size and detection in breast cancer: self-examination and clinical breast examination are at their limit · Cancer Detection and Prevention](https://doi.org/10.1016/j.cdp.2008.04.002)
- [Female cancer screening in India: uptake and inequality, NFHS-5](https://pmc.ncbi.nlm.nih.gov/articles/PMC11610822/)
- [Community-based cancer screening at primary care level in Varanasi district · BMJ Open](https://pmc.ncbi.nlm.nih.gov/articles/PMC10729271/)
- [AI-supported screen reading versus standard double reading (MASAI) · The Lancet](https://www.thelancet.com/journals/landig/article/PIIS2589-7500(24)00267-X/fulltext)
- [MedCognetics: mammography workflow](https://www.medcognetics.com/)

## Continue exploring

- [Other records Bright met the same way](https://brightaifuture.com/collections/nvidia)
- [Health & biology](https://brightaifuture.com/worlds/health)
- [Work & learning](https://brightaifuture.com/worlds/work)
- [People & culture](https://brightaifuture.com/worlds/culture)
- [What if the body could answer intention again?](https://brightaifuture.com/threads/agency)
- [What becomes possible when we can see and shape life’s machinery?](https://brightaifuture.com/threads/biology)
- [When does a technical result become a public capability?](https://brightaifuture.com/threads/community)