Fig. 5: Normalized predicted swelling values from ML models & Experts.

Each image ground truth swelling is set to be 1.0, unfilled symbols denote images that fail the image confidence score filtering. Predictions are on the 20 image Cavity Test set.

Each image ground truth swelling is set to be 1.0, unfilled symbols denote images that fail the image confidence score filtering. Predictions are on the 20 image Cavity Test set.