Performance of the Gradient-Boosted Decision Tree (GBDT) model trained with Input C and tested on the external dataset.

(A) displays the normalized confusion matrix. The greatest difficulty this model faced was in distinguishing early arterial phase (E-AP) from late arterial phase (L-AP), even more so than in the original dataset. The model had comparable accuracy distinguishing late arterial phase (L-AP) from portal venous phase (PVP) to the original dataset. (B) displays the precision-recall curves (PRC) using a One vs. Rest (OvR) approach (See Fig 5 for details on interpretation). Similar to in the original dataset, the PRC shows that the model has the most difficulty classifying E-AP scans.