Clinical Screening · Telegram Bot · Beta

Rad Skin Analysis

One Photograph, Read Before the Consultation

Rad Skin Analysis (@rad_skin_bot) is a pre-consultation screening assistant for dermatology and aesthetic clinics, and it lives inside Telegram because that is where the patient already is. One photograph — sent, or taken through a guided camera inside the bot — produces two documents: a readable report for the person who sent it, and a fuller, evidence-led one for the clinic, in Farsi or English, as PDF. The parts that have to be reproducible are not left to a model.

Face landmarks map eleven regions — the forehead, brow, nose, cheeks, the skin around each eye and the mouth, the chin and both jawlines — deterministically, on a CPU, in well under a second and at no cost per scan. Six axes are scored from nought to ten and land in one of five named bands rather than arriving as a bare number: dryness, surface evenness, redness, tone unevenness, lines and pores. The overlay is drawn on the photograph in a single cool hue on purpose, because warm heat colours over skin are almost impossible to tell apart from the redness they are drawn on top of.

The skin-age number is built to survive being compared: the model is asked only for a decade, and the position inside that decade is computed from the scores by a fixed rule, so a re-scan three months later moves only if the skin moved — and every earlier scan is kept, because being compared with the next one is the report's whole job. What it will not do is name a disease. The vision model may describe distribution, colour, border quality and symmetry from a closed vocabulary of what a photograph can show, and has no permission to say the word for any condition; the step from those descriptors to “consider eczema” is a lookup a specialist can read and defend, and it answers the same way twice.

Confidence stops at moderate — there is no “high” band in the code at all, not reserved and not unreachable — and every candidate carries the evidence against it beside the evidence for it. @rad_skin_bot is the bot you can open and try today; a clinic that adopts it is given its own instance, carrying its own name. We analyse.

The doctor decides.

Features

  • One photograph, or a guided camera inside the bot — nothing to install
  • Eleven facial regions mapped by landmarks: deterministic, sub-second, no cost per scan
  • Six axes — dryness, evenness, redness, tone, lines, pores — scored nought to ten and banded
  • An overlay drawn on the photograph itself, in a single cool hue
  • A skin-age number that may be compared: a decade from the model, the position inside it from a fixed rule
  • Every earlier scan kept, because the report is written to be compared with the next one
  • Two reports per scan — one for the patient, one written as evidence for the clinic · Farsi & English · PDF
  • No diagnosis: named conditions come from a table a specialist can read, and confidence stops at moderate