Each patient signal (age risk, biomarker level, symptom severity) enters the input layer. Every hidden layer computes a weighted sum of the previous layer's activations plus a bias, then squashes it with a sigmoid; the output layer turns its raw scores into probabilities with softmax.
a(l) = σ(W(l)·a(l-1) + b(l))
P = softmax(z_out)
- Age risk / Biomarker level / Symptom severity — the three input activations, shown as the bottom ring of nodes.
- Hidden layers — two dense sigmoid layers; brighter nodes carry a stronger signal, and only connections above a strength threshold are drawn.
- Output layer — three classes (Healthy / At-risk / Disease); node brightness is the softmax probability.
- Decision threshold — the probability a class needs to clear before it becomes the predicted diagnosis; raising it makes the network more conservative about flagging risk or disease.
This is the same feedforward pattern used by real clinical ML models for triage and risk scoring, just with three inputs instead of hundreds of lab values and imaging features.