Infarct quantification with cardiovascular magnetic resonance using

Table 1 Results overview

Method	Mean error % units (accuracy)			Variability (CoV) %precision)
	ROI pos	ROI size	SD	ROI pos	ROI size	SD
Naïve, all data	5 ± 9	7 ± 13	4 ± 13	47 ± 46	48 ± 72	61 ± 42
Spurious removal, all data	3 ± 9	6 ± 13	3 ± 12	43 ± 45	45 ± 72	57 ± 40
Spurious removal, high quality	5 ± 8	9 ± 12	5 ± 11	41 ± 38	46 ± 66	54 ± 37
Spurious removal, good quality	3 ± 9	5 ± 13	3 ± 13	44 ± 50	43 ± 77	57 ± 41
Spurious removal, poor quality	− 3 ± 9	− 1 ± 12	− 2 ± 12	44 ± 38	51 ± 61	62 ± 50

Mean error and variability for the different versions of the algorithm and different sets of the data. Naïve refers to not removing spurious pixels, no-spurious refers to removing spurious pixels. High quality, good quality and poor quality refers to that the analysis was only performed on the data subsets with the corresponding image quality rating. A positive mean error (accuracy) represent overestimation compared to reference infarct size and negative mean error an underestimation. Mean error is expressed in scar percentage units. Variability (precision) is expressed as coefficient of variability, i.e. standard deviation of scar sizes divided by reference infarct size

ISSN: 1532-429X