Synthetic data only
Scan-report follow-up, across every hospital

No follow-up silently lost.

AI reads scan reports, tracks every recommended follow-up across hospitals, and reaches the patient in their own language — a clinician approves every step.

See the patient app
  • Human review at every handover
  • Urgent findings: always a phone call
  • Synthetic data only

The gap

What no single hospital sees.

A scan report at Northbridge recommends a follow-up at Valeford. Each hospital's system only sees its own bookings — so the request falls between them.

Northbridge Hospitals
Its own booking system
Handed over

The request left Northbridge's worklist when care moved to Valeford.

Valeford Hospitals
Its own booking system
No record

The report was written at Northbridge, so Valeford never received it.

ResultWise
Sees every hospital
Overdue — escalated to nurse

No booking at either hospital, so the clinician was alerted — and 14 days later a nurse phones the patient.

What it does

One loop, from report to patient.

Every piece is built so a person can check it in seconds.

AI reads the report

Pulls out the test, due date, urgency and hospital — and highlights the exact sentence it used.

Conclusion: LVEF within normal limits with a small fall from baseline. Continue trastuzumab. Repeat echocardiogram in 12 weeks as per surveillance protocol; this will be performed at Valeford Hospitals cardio-oncology service.

Heart scan (echocardiogram)Due 02 Nov 2026RoutineValeford Hospitals

Clinician in control

Confirm or correct every field. Nothing is tracked until a clinician ticks it off.

I have checked this against the report
Confirmed by Dr Priya Rao

Cross-hospital tracker

Looks for a booking at every hospital, every day — and escalates when there isn't one.

  1. 1Watching Northbridge and Valeford
  2. 2Overdue: clinician alerted
  3. 314 days later: nurse phones the patient

Urgent means a phone call

Suspicious or urgent findings never get an automated message. A clinician phones the patient.

Automated messages blocked

Phone the patient, then mark “I have phoned the patient”.

8 languages, made for reading

Plain words in the patient's language, with easy read, large text, high contrast and read aloud. Right-to-left scripts too.

EnglishPolskiاردوதமிழ்বাংলাਪੰਜਾਬੀالعربيةRomână
Large text High contrast Read aloud Easy read

Patients can reply

One tap — and the reply lands live in the clinician's inbox.

Please call me
Clinician inbox · new reply

FHIR Task export

Every confirmed follow-up exports as a standard FHIR Task, ready for another hospital's system.

{
  "resourceType": "Task",
  "status": "requested",
  "intent": "order",
  "priority": "routine",
  "code": {
    "text": "Heart scan (echocardiogram)"
  },
  "for": {
    "reference": "Patient/halina",
    "display": "Halina Nowak"
  },
  "owner": {
    "display": "Valeford Hospitals"
  },
  "restriction": {
    "period": {
      "end": "2026-11-02"
    }
  }
}

Full audit log

Every AI, system, clinician and patient action is written to the case's timeline.

  1. 10 Aug · SystemReport received from Northbridge Hospitals cardiology.
  2. 10 Aug · AIFound 1 follow-up: echocardiogram in 12 weeks at Valeford (90% confidence).
  3. 11 Aug · ClinicianDr Priya Rao checked against report and confirmed.
  4. 11 Aug · SystemTracking at Northbridge and Valeford. Care handed over to Valeford.

Safety

AI suggests. A person decides.

Clear lines, built into the product rather than left to policy.

What the AI does

  • Reads free-text scan reports and finds the recommended follow-up
  • Pulls out the test, due date, urgency and hospital
  • Quotes the exact sentence it used, so a clinician can check it in seconds
  • Flags anything uncertain for a careful human check
  • Drafts a plain-language translation, with a back-translation to compare
  • Sorts patients' questions — serious ones first — and drafts a reply for a clinician to edit

What the AI never does

  • Never: Tracks or sends anything a clinician hasn't confirmed
  • Never: Messages a patient about an urgent or suspicious finding — a clinician phones
  • Never: Puts a diagnosis in a patient message
  • Never: Changes a date, phone number or name in translation — the safety check blocks sending
  • Never: Closes a loop on its own — a person verifies every booking
  • Never: Answers a patient's question itself — a clinician approves every reply

On a small synthetic test set of 10 held-out reports, the AI found every recommended follow-up; keyword rules found 22%. Real validation is needed before clinical use. How accurate is the AI?

Please read: what this is, and isn't

  • • Hackathon prototype — not a product in clinical use.
  • • Synthetic data only: every patient, hospital and number is fictional.
  • • Not a medical device. Not clinically validated.
  • • Never enter real patient or health information.

See the whole loop in 60 seconds.

Follow Halina's heart scan from a free-text report to a checked message in Polish.