Fig. 4: Zero-shot evaluation of RadFM on the unseen classes in the PadChest dataset.
From: Towards generalist foundation model for radiology by leveraging web-scale 2D&3D medical data
We evaluate the model on the human-annotated subset of the PadChest dataset, and ACC scores are shown for the radiographic findings or diagnosis. The top 100 classes in the test dataset are shown in the figure.
