Algorithm A1 Inference algorithm of the cascade graph encoder.

1:Input:  Barcode-Like Timelines B; Stride Size s; Kernel Size w; Network Weights ${\theta }_{{\mathrm{E}}_{\mathrm{all}}}$; Set of Substructures $R_{\mathrm{heart}}$, $R_{\mathrm{vessels}}$, $R_{\mathrm{others}}$ 2:Output:  Sub-graph Chart Diagrams $G_{\mathrm{heart}}$, $G_{\mathrm{vessels}}$; Main-graph Chart Diagram $G_{\mathrm{main}}$ 3:$i=1$ 4:while$(i-1)s+w\le T_{\max }\left(B\right)$do 5:        $g_{\mathrm{heart}}^i(B\left(R_{\mathrm{heart}}\right);{\theta }_{{\mathrm{E}}_{\mathrm{heart}}})={\mathrm{E}}_{\mathrm{heart}}(K(i,B\left(R_{\mathrm{heart}}\right));{\theta }_{{\mathrm{E}}_{\mathrm{heart}}})$ 6:        $g_{\mathrm{vessels}}^i(B\left(R_{\mathrm{vessels}}\right);{\theta }_{{\mathrm{E}}_{\mathrm{vessels}}})={\mathrm{E}}_{\mathrm{vessels}}(K(i,B\left(R_{\mathrm{vessels}}\right));{\theta }_{{\mathrm{E}}_{\mathrm{vessels}}})$ 7:        $G_{\mathrm{heart}}$⟵ add point $g_{\mathrm{heart}}^i(B\left(R_{\mathrm{heart}}\right);{\theta }_{{\mathrm{E}}_{\mathrm{heart}}})$ 8:        $G_{\mathrm{vessels}}$⟵ add point $g_{\mathrm{vessel}\mathrm{s}}^i(B\left(R_{\mathrm{vessels}}\right);{\theta }_{{\mathrm{E}}_{\mathrm{vessels}}})$ 9:        $v=\left(g_{\mathrm{heart}}^i(B\left(R_{\mathrm{heart}}\right);{\theta }_{{\mathrm{E}}_{\mathrm{heart}}}),g_{\mathrm{vessels}}^i(B\left(R_{\mathrm{vessels}}\right);{\theta }_{{\mathrm{E}}_{\mathrm{vessels}}}),{\mathrm{E}}_{\mathrm{others}}(K(i,B\left(R_{\mathrm{others}}\right));{\theta }_{{\mathrm{E}}_{\mathrm{others}}})\right)$ 10:        $g_{\mathrm{main}}^i(B;{\theta }_{{\mathrm{E}}_{\mathrm{sub}}},{\theta }_{{\mathrm{E}}_{\mathrm{main}}})={\mathrm{E}}_{\mathrm{main}}(v;{\theta }_{{\mathrm{E}}_{\mathrm{main}}})$ 11:        $G_{\mathrm{main}}$⟵ add point $g_{\mathrm{main}}^i(B;{\theta }_{{\mathrm{E}}_{\mathrm{sub}}},{\theta }_{{\mathrm{E}}_{\mathrm{main}}})$ 12:end while