All use cases
Hospitals & health departments
Faster care in district hospitals — the doctor decides.
Scans are checked for quality, urgent cases move up the radiologist's list, and every report starts as a structured draft with measurements. Radiologists edit and sign; discharge summaries and follow-up calls take care of themselves.
In tests · clinical validation with partner hospitals pendingHow it works
From the scanner to the patient's follow-up
- 01Quality checkImages from the hospital system are checked before any model runs.
- 02Urgent firstThe imaging model moves urgent cases up the radiologist's list.
- 03Report draftA structured draft with measurements and key findings highlighted.
- 04SignThe radiologist edits and signs; the EMR / HIS is updated.
- 05Follow-upHindi reminder calls and SMS before each visit or vaccination.
Capabilities
What Medical Imaging does
- 01X-ray triage18-finding classifier with heatmaps, anatomy segmentation, cardiothoracic ratio and mediastinum width.
- 02Eye screeningFundus quality, optic disc and cup, vCDR and the ISNT rule; OCT retinal thickness and the ETDRS map.
- 03CT densityLung density and organ density with automatic quality warnings; uncertain regions left unmeasured.
- 04Report draftsVision-language model plus a prompt template produce a draft the radiologist edits and signs.
- 05Discharge summariesDrafted from case notes for the doctor to review.
- 06ASHA assistantWhatsApp answers for health workers from NHM guidelines, with the source cited.
Blocks used
- Medical imaging
- Vision models
- Image quality check
- Human approval
- HTTP / API
- IVR calls
- SMS
In practice
Who it serves
- Swasthya SahayakDistrict hospitals: patients, doctors and ASHA workers under Medical Health & Family Welfare and NHM.
- Radiology departmentsUrgent cases first; time to radiologist review is a pilot measure.
- Hospital chainsScans and patient data stay inside the hospital network; hospitals can plug in their own ONNX models.
Accountable by design
Built to be trusted
- Identity-free outputsPatient identity tags never travel with the image or the result.
- Every caveat statedScale, coverage and contrast written into the result itself.
- Licences documentedOnly models we may use; no faked or renamed models.
- Clinician confirmsDecision support, not diagnosis — nothing is diagnosed by the system alone.