Automated quality control in image segmentation: application to the UK Biobank cardiovascular magnetic resonance imaging study

Table 2 Initial reverse classification accuracy validation on 400 random forest segmentations

Class	Acc.	TPR	FPR	MAE
				DSC	MSD	RMS	HD
					mm	mm	mm
LVC	0.973	0.977	0.036	0.020	4.104	5.593	14.15
	0.980	0.975	0.019
LVM	0.815	0.947	0.215	0.044	3.756	4.741	13.08
	0.990	0.987	0.008
RVC	0.985	0.923	0.012	0.030	4.104	5.022	16.63
	0.943	0.914	0.047
Av.	0.924	0.949	0.089	0.031	3.988	5.119	14.62
	0.971	0.959	0.025
WH	0.988	0.979	0.000	0.029	4.445	5.504	15.11
	0.948	0.886	0.047

Classes are LV Cavity (LVC), LV Myocardium (LVM), RV Cavity (RVC), An average over the classes (Av.) and a binary segmentation of the whole heart (WH). First row for each class shows the binary classification accuracy for ‘poor’ and ‘good’ segmentations in the Dice Similarity Coefficient (DSC) ranges [0.0 0.7) and [0.7 1.0] respectively. Second row for each class shows the binary classification accuracy for ‘poor’ and ‘good’ segmentations in the Mean Surface Distance (MSD) ranges [>2.0mm] and [0.0mm 2.0mm] respectively. True-positive and false-positive rates are also shown. We report mean absolute errors (MAE) on the predictions of DSC and additional surface-distance metrics: root-mean-squared surface distance (RMS) and Hausdorff distance (HD)

ISSN: 1532-429X