Semi-automatic Segmentation of the Myocardium in High-Frame Rate and Clinical Contrast Echocardiography Images

Myocardial perfusion quantification (MPQ) with myocardial contrast echocardiography (MCE) is a technique to estimate myocardial blood flow. Accurate blood flow estimation requires precise myocardial segmentation. Due to large intensity variations and the time-varying nature of real-time MCE sequences, manual segmentations are often required. Generating manual segmentations is time-consuming, has low reproducibility and is heavily dependent on the experience level of the clinician. In this paper, we propose a semi-automatic myocardium segmentation approach for real-time high frame rate (HFR) research and clinical MCE images. The proposed algorithm is an image registration-based method. It also employs post-processing techniques and incorporates temporal heart phase information to improve predicted segmentations of unlabeled frames. Results shows that the proposed methodology was effective in producing accurate and replicable segmentations with a DSC of 0.82 and HD of 26 for both research and clinical sequences. Moreover, utilizing the resulting segmentations, time-intensity perfusion curves were generated for MPQ.