Validation of an algorithm for left ventricular segmentation in 150 patients shows potential for further development towards fully automatic segmentation

Background Automatic segmentation of the left ventricle (LV) is desirable to assess the cardiac parameters end-diastolic volume (EDV), end-systolic volume (ESV), ejection fraction (EF) and left ventricular mass (LVM) since manual segmentation is time consuming and observer dependent. A physiologically correct segmentation of the left ventricle requires careful consideration of the long axis displacement and the LV outflow tract which makes the myocardium non-circumferential in the basal slices. To detect the long axis displacement a constraint could be used to keep the LVM fairly constant over the cardiac cycle. However, in order to use this constraint, the error of the segmentation has to be low regarding both endocardial and epicardial borders in the non-basal part of the LV. Therefore, the purpose of this study was to improve and validate an automatic algorithm for LV segmentation in the non-basal part of the LV, as a first step towards fully automatic segmentation.