Fig. 8: Confusion matrices for non-gated model performance on the MESA test set based on ground truth method chosen.
From: Automated coronary calcium scoring using deep learning with multicenter external validation
The left matrix (a) compares model prediction to scores derived from manual segmentations on non-gated chest CT exams while the right matrix (b) compares model predictions to scores derived from gated coronary CT exams, the clinical reference standard for CAC scoring. Note the performance differences (higher on the left) based on ground truth chosen.
