A family travels to a Government Medical College for a CT or MRI. The scan is done that morning. Then: come back in two weeks for the report. They travel home and travel back again.
VitalView AI drafts the report in under two minutes. The patient learns what the scan shows before leaving. The radiologist's signed report still follows.
The scan report becomes like an X-ray. You get it before you leave.
We read the scan and write a draft. The doctor checks it and signs it.
The CBCT research at Government Dental College, Thiruvananthapuram is the proof: the same system reads a dental scan and a chest scan. Radiology, dental, orthopaedics, neurology, any department that produces images can use it.
For the public first, then for the hospital, then for Kerala's own scans.
For a daily-wage earner from a hill district, the second trip can cost more than the treatment. A government scanner that answers the same day is a different hospital to the public, and nothing was rebuilt.
The waiting list is not a shortage of machines. It is a shortage of people to read them. India has roughly one radiologist per 100,000 people, and 70 to 80% work in the big metros. Kerala's government tertiary hospitals compete for what is left.
What each number turns into. A tertiary department at 35 to 82% of strength cannot hire its way out this year, but the same doctors clear far more of the queue starting from a finished draft. The 5,476 posts stay vacant and the work still gets done. The accident cases are the sharpest version: a bleed on the brain cannot wait until morning, and at 2 a.m. a draft is the only reading available for hours.
Medical colleges get less load, not more. A district hospital that cannot report a scan sends the patient up to the medical college. Give it a same-day draft and most of those referrals never travel. The medical college keeps only the cases that need one.
Wayanad shows the pattern. The district got its medical college in September 2025, and a dialysis centre the same month. ₹23.75 crore sanctioned in 2021-22 for a critical care block at Kalpetta Government General Hospital is still unbuilt, and in October 2025 the constituency MP, Priyanka Gandhi Vadra, warned publicly that accident victims die on the road to Kozhikode Medical College. Facilities keep arriving in Wayanad. The specialists who read the scans do not.
Kerala's government hospitals hold decades of scans with the radiologist's signed report attached. That pairing, the picture and the expert's answer, is the most valuable material in medical AI, and it sits in the archives doing nothing.
Indian data is the scarce part. ICMR, IISc and ARTPARK built the MIDAS platform for one reason: AI trained on foreign populations underperforms on Indian patients, and Indian annotated imaging data is hard to get. Kerala holds what the country lacks. Every scan a Kerala doctor corrects improves the model, so quality rises the longer it runs, and rises on Indian patients.
Built here, not bought in. The mark-up, checking and technician work behind the model is learnt in months and done on a screen. Kerala has the graduates for it.
Nobody already in the job is replaced. A radiographer who learns to run and check these tools becomes more valuable. In a 2024 survey of Indian radiologists, 95.3% said they wanted to learn more about AI.
Three things, to prove it on Kerala's own scans. The trial runs in shadow mode: the AI reads alongside the existing process, and its output never treats anyone.
The AI's draft is compared against the radiologist's report already on file. Target agreed in writing first.
Gate: agreed accuracy on serious findings, or we stop.The AI reads live scans in parallel. Nobody treating the patient sees its output. This is where we learn whether it holds up across different machines.
Gate: sustained accuracy, and the CDSCO licence in hand. The licence takes 12 to 18 months, so trial and licence run in parallel.Only now does the draft reach a doctor in real time. Contracted through the KSUM startup route.
Gate: departmental review at every renewal.Medical AI has failed elsewhere. We would rather bring you the list ourselves than have it arrive in a question hour.
| Care point | What could go wrong | How the trial prevents it |
|---|---|---|
| Over-trusting the machine | A tired doctor at 3 a.m. accepts the draft without checking it properly. | The doctor signs, every time. Confidence scores flag the weak findings, and every change is recorded. Radiologists set the standard and stay in charge, which is why the department backs it. |
| Who is responsible | No Indian statute is written specifically for AI in medicine, and the Digital India Act remains in draft. | Under current law the signing doctor and the hospital remain responsible, which is why the software must never report on its own. AI software is still regulated, as a medical device: CDSCO's Medical Device Software guidance of 21 July 2026 sets explicit expectations for algorithm documentation, dataset composition and validation on Indian populations. During the trial the software makes no clinical decision at all. |
| Patient privacy | Health data is protected under the Digital Personal Data Protection Act 2023. Replacing a name with a code is not enough. | Old scans, all identifiers removed, including anything printed onto the image. Ethics Committee oversight throughout, State keeps control of the data, written plan submitted before access. |
| Moving too fast | A good trial creates pressure to switch it on for real before the evidence justifies it. | Live use is locked behind the agreed accuracy target and the CDSCO licence, written into the trial agreement. |
| Where we stop | The trial produces a poor result and there is pressure to bury it. | Agreed in advance: if accuracy falls below target on any scanner, we pause and retrain before the software goes near a patient, and we say so publicly. Better a published poor result on Kerala's scans than a repeat of Epic and IBM. |
Kerala has bought the machines. What Kerala cannot buy is the people who read them. They do not exist to be bought.
With Wayanad and Kasaragod, every district in Kerala now has a medical college. ₹45 crore has gone into cath labs and ₹80 crore into ambulances. The buildings and the scanners are in place.
And in those same hospitals, the specialist posts sit unfilled. What is stuck cannot be solved by spending: the specialists are not there to recruit, in Kerala or anywhere in India.
A scanner the department already owns can be read tonight. A radiologist it cannot recruit cannot.
eHealth runs in around 800 government institutions with about 2.62 crore health IDs behind it, and the reports go straight into it. The AI Centre of Excellence in Thiruvananthapuram, set up with IndiaAI, provides the computing power.
Public figures are sourced below. Figures from our own deployments are marked as ours.