Figure 1.

Bayesian network for modelling lymphatic metastatic spread as described by (Pouymayou et al.)³¹. It consists of the primary tumor $T$, hidden binary variables $X_v$ for the involvement of LNL $v$ (white circles) and observed (or diagnostic) variables as dark circles ($Z_v^{\mathcal{O}}$, where $\mathcal{O}$ denotes the used diagnostic modality). There are potentially many observations per hidden variable. Annotated arcs depict the direction of lymphatic flow where the parameter next to it ($b$ and $t$) represents the probability of metastatic spread. Not annotated arrows connect the LNLs to their diagnoses via sensitivity and specificity.