2022

Rarethee — Bayesian decision chain for rare-disease early diagnosis

Interactive AI-based digital medical device supporting pediatricians and family doctors in the early diagnosis of rare diseases, built on a Bayesian decision chain.

Rarethee is an interactive, AI-based digital medical device designed to support pediatricians and family doctors in the early diagnosis of rare diseases. The core of the system is a Bayesian decision chain: each observation (symptom, sign, test result) updates a probability distribution over a large catalogue of rare disease hypotheses, and the system suggests the next most informative question or test to narrow the differential.

The project placed third at the 5th Rare Disease Hackathon (2021) and was developed further through 2021–2022, with attention to explainability, clinician trust and safe-fallback behaviour when evidence is thin.

Designing for rare diseases sharpens the usual ML trade-offs: class imbalance is severe, labelled data is scarce and the cost of a missed positive is high. Bayesian reasoning provides a principled language for that regime, and for communicating uncertainty back to a clinician.