Platform
PITHIA.ai supports rapid donor kidney biopsy review with automated pre-screening, consultant verification, and traceable reporting.
The purpose is not to automate diagnosis. It is to make expert review available in the transplant decision window, with clear accountability for the final decision.
Automated pre-screening
Most digital pathology viewers are a window onto an image. The user still does the time-consuming work of finding structures and making counts or measurements by hand.
In PITHIA.ai, automated pre-screening prepares a draft assessment before a consultant opens the case. We call this the "Ghost User": AI model-generated measurements and annotations presented for verification.
When a biopsy arrives, the system proposes key regions and structures of interest and computes draft quantitative measures. The consultant reviews those outputs, corrects any inaccuracies, and signs off the final assessment.
A traceable chain of review
For a transplant team waiting in theatre, speed matters. Trust matters more. The report needs to show who reviewed the biopsy, what was measured, and when the decision was made.
PITHIA.ai records an audit trail for each case. Reports can include the model version used for pre-screening, verification actions taken, and the named consultant who signed off.
Clinical responsibility remains with the consultant. Automated analysis informs the draft; the signed report reflects human verification.
Designed for urgent, interrupted work
Renal transplant decisions are often made out of hours. Reviews are done under pressure and in between other clinical responsibilities.
The interface is designed for interrupted work: start a case on a tablet, step away, and return later without losing context.
Corrections are recorded as part of quality improvement. Over time, verified review data can be used to refine automated pre-screening under appropriate governance.
Extending to other services
Our immediate focus is a reliable national renal pathway in the UK. Once established, the same workflow can be adapted to other time-critical services where remote expert review and traceable verification are required.