Table 2. Best classification metrics for each feature-level approach.

Approach	Handcrafted and Radiomic Features	Pretrained Resnet18	Pretrained DenseNet121	Combined Deep Features	All Features	Selected Handcrafted, Radiomic, and Deep Features
Accuracy	0.892	0.886	0.917	0.912	0.925	0.941
Precision	0.895	0.899	0.932	0.926	0.926	0.944
Recall	0.892	0.886	0.917	0.919	0.925	0.941
F1 Score	0.892	0.882	0.914	0.910	0.922	0.940