Validation and Development of a New Automatic Algorithm for Time-Resolved Segmentation of the Left Ventricle in Magnetic Resonance Imaging

Table 6

Image processing error measurements in second observer set divided into slice sections. Image processing error measurements in the second observer subset () as dice similarity coefficient (DSC) and point to curve (P2C) for second observer manual delineation versus manual reference delineation and for automatic segmentation versus manual reference delineation.

	Automatic segmentation versus manual reference		Second observer versus manual reference
	DSC	P2C	DSC	P2C
Endocardium basal	0.88 ± 0.06	2.7 ± 1.1 mm	0.94 ± 0.02	1.5 ± 0.4 mm
Endocardium midventricular	0.94 ± 0.01	1.6 ± 0.4 mm	0.96 ± 0.01	1.1 ± 0.3 mm
Endocardium apical	0.90 ± 0.03	2.0 ± 0.7 mm	0.94 ± 0.01	1.1 ± 0.3 mm
Epicardium basal	0.89 ± 0.05	3.3 ± 1.2 mm	0.95 ± 0.02	1.5 ± 0.6 mm
Epicardium midventricular	0.96 ± 0.01	1.3 ± 0.4 mm	0.97 ± 0.01	0.8 ± 0.2 mm
Epicardium apical	0.92 ± 0.02	2.3 ± 0.7 mm	0.95 ± 0.01	1.4 ± 0.5 mm

Difference are expressed as mean ± SD. For a perfect overlap between the regions DSC should be 1 and P2C should be 0. Basal, midventricular, and apical sections are defined as 1/3 each of the ventricular length in end diastole and end systole separately.