Science & validation
Trained by trichologists. Validated in real clinics.
Trichora is a proprietary computer-vision model built for one job: reading the scalp the way a hair specialist does, and doing it identically every time. Here is how it was trained, how it was tested, and how your patients' data is protected.
- 10,000+annotated images
- 3,000+validation patients
- 90%+zone accuracy
How the model was trained
We taught it what specialists see, not what a label says.
Most hair-loss AI is trained on a single photo and a yes/no label. Trichora was trained differently. For every patient in the dataset, hair specialists reviewed all five views and drew exactly what they saw: the true hairline, the outline of every thinning region, and how severe it was. The model learned to reproduce those drawings — which is why it returns a map, not a verdict.
- Multi-angle: five views per patient, so the frontal, mid-scalp, crown, temporal and occipital regions are each seen from the angle that reveals them
- Expert-annotated: hairline position, thinning contours and severity marked by hair specialists, with inter-annotator review
- Diverse: Indian skin tones and hair textures first, across age, sex and loss pattern, in ordinary clinic and phone lighting
- Annotator A ✓
- Annotator B ✓
- Consensus
Validation
Tested against specialists, on patients the model had never seen.
A held-out cohort of more than 3,000 patients, scanned in partner clinics, was independently annotated by hair specialists. Trichora's output was compared zone by zone.
- Thinning-zone detection
90%+agreement
Whether the model found thinning in the same zones the specialist marked, across frontal, mid-scalp and crown.
- Severity scoring
±1band
Zone thinning scores fall within one severity band of the specialist's in the large majority of cases — the same tolerance two specialists show with each other.
- Repeatability
Samemap
The same patient, re-captured by a different staff member on a different phone, returns consistent zone scores and grade — the property that makes visit-to-visit tracking trustworthy.
Continuous audit
Every new clinic adds a fresh, specialist-reviewed sample to the validation set. Accuracy is re-measured and reported quarterly. Detailed validation methodology and figures are available under NDA.
What Trichora is
A measurement and decision-support tool
- Objective, repeatable scalp measurement from phone photos
- A structured baseline and change tracking across visits
- A drafted protocol and outlook for the doctor to review
- A clearer conversation between doctor and patient
What Trichora is not
A replacement for the doctor
- It does not diagnose the cause of hair loss
- It does not prescribe — every protocol is signed by a registered practitioner
- It does not guarantee outcomes; projections are statistical estimates
- It does not replace a clinical examination where one is indicated
Data security & privacy
Patient photos are medical data. We treat them that way.
Encrypted end to end
AES-256 at rest, TLS 1.2+ in transit. Images encrypted on the handset before upload.
Hosted in India
Mumbai region. Patient records, images and reports are stored in India. Aligned with DPDP Act requirements.
Consent-first
Explicit patient consent captured in-app. Images are never used to improve the model without separate opt-in.
Full audit trail
Every scan, view, edit, signature and share is logged per user. Clinics can export logs at any time.
Clinic-controlled retention
You set how long images and reports are kept. Deletion requests are honoured across all copies.
Role-based access
Front desk, doctor and admin roles. Per-clinic API keys. SSO available for chains.