Adaptive blood pool segmentation in three-dimensions: Application to MR cardiac evaluation

MRI is an established method of imaging the cardiac blood pool in four dimensions and evaluating global cardiac function. However, segmentation of the cardiac blood pool from the myocardial wall continues to be a time-consuming task and is operator dependent. This has hampered the widespread use of cardiac MRI in evaluating global cardiac function. We propose the use of an adaptive threshold-based, three-dimensional region-growing technique to segment the cardiac blood pool from the myocardium and to compute left ventricular blood volumes. It uses a unique double segmentation approach incorporating information from the myocardium to help better detect the edges of the blood pool. The adaptive threshold segmentation technique was evaluated on four human subjects by comparing with manual segmentation. The automated segmentation technique is fast and agreed well with manual segmentation. Further, it requires minimal operator input and is robust and user friendly. The conclusion is that adaptive thresholds and a multidimensional region-growing approach is an appropriate method to segment the left ventricular blood pool.