Unsupervised Segmentation of B-Mode Echocardiograms

We present a method for unsupervised segmentation of echocardiograms (echo). The method uses an iterative Dijkstra's algorithm, a strategic node selection, and a novel cost matrix formulation based on intensity peak prominence and is thus termed the "Prominence Iterative Dijkstra's" algorithm, or ProID. Although the current analysis focuses on the left ventricle (LV), ProID is applicable to all four heart chambers. ProID was tested using artificial echo images representing five different systems. Results showed accurate LV contours and volume estimations as compared to the ground-truth for all systems. Subsequently, ProID was used to analyze a clinical cohort of 66 pediatric patients, including both normal and diseased hearts. Output segmentations, end-diastolic, end-systolic volumes, and ejection fraction (EF) were compared against manual segmentations from two expert readers. ProID maintained an average Dice similarity score of 0.93 when comparing against manual segmentation. Comparing the two expert readers, the manual segmentations maintained a score of 0.93, which increased to 0.95 when they used ProID. Thus, ProID successfully reduced the inter-operator variability across the two expert readers. Overall, this work demonstrates that ProID yields accurate boundaries across all age groups, disease states, and echo platforms with low computation cost, thereby establishing its clinical usefulness.