Speckle reducing anisotropic diffusion for echocardiography

Echocardiographic segmentation accuracy significantly affects the calculation of important cardiac parameters, which assist in the understanding, diagnosis, and treatment of cardiovascular disease. This paper investigates extensions to speckle reducing anisotropic diffusion (SRAD) for the purpose of improving automated segmentation in echocardiographic imagery. We first investigate windowed SRAD (WSRAD), which attempts to increase edge contrast while preserving object size via the addition of neighborhood statistics to the SRAD diffusion coefficient. We also consider deconvolutional SRAD (DeSpeRADo), which attempts to alleviate blurring artifacts inherent to the ultrasound imaging system. We quantify the segmentation effects of each filtering method via simulated cardiac studies. With respect to unfiltered segmentation results, we measured a 21%, 24%, and 22% increase in myocardial segmentation accuracy with SRAD, WSRAD, and DeSpeRADo, respectively.