Omniscient Diagnostic Network

Clinical Evidence

Certainty threshold 60%

Posterior Probabilities

Leading diagnosis

Low-acuity

Diagnostic certainty

Evidence active

0 / 6

Above threshold

How it works

Each candidate diagnosis D has a prior probability P(D) (its base rate before any evidence). Each finding E_j has a likelihood P(E_j | D) — how often that finding occurs in patients who truly have D. The "omniscient" engine here is a naive Bayes classifier: it treats every toggled finding as conditionally independent given the diagnosis and multiplies their likelihoods together.

P(D | E₁..E₆) ∝ P(D) · Π P(Eⱼ | D)^[on] · (1−P(Eⱼ | D))^[off]

then normalise across all 6 diagnoses so probabilities sum to 1.

Diagnostic certainty is 1 minus the normalised Shannon entropy of the posterior distribution — it is 0% when every diagnosis is equally likely (maximum confusion) and approaches 100% when one diagnosis completely dominates.

Evidence buttons — toggle findings (chest pain, dyspnea, fever, tachycardia, hypoxia, diaphoresis) on or off; every toggle re-runs Bayes' rule instantly.
Random Patient — samples a true underlying diagnosis from the priors, then samples each finding from that diagnosis's own likelihood table, exactly like a real (simulated) case vignette.
Certainty threshold — highlights the leading diagnosis's bar and 3D pylon when its posterior clears the slider value, the same "confidence gate" clinical decision-support tools use before surfacing a suggestion.

This is a simplified version of the reasoning behind real Bayesian clinical decision-support systems (e.g. DXplain, Isabel) — the physics/math is the genuine Bayes' theorem, only the diagnosis set and likelihood table are illustrative.

Active evidence node Inactive evidence node Leading diagnosis pylon

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.