Fuzzy Connectedness Based Segmentation of Fetal Heart from Clinical Ultrasound Images

Congenital Heart Disease(CHD) is one among the most imperative causes of neonatal morbidity and mortality. Nearly 10 percentile of contemporary infant mortality in India is accounted for CHD. It is optimal to use ultrasound imaging modality for scanning the well-being of growing fetus owing to its non-invasive nature. But then several issues such as the manifestation of speckle noise, poor quality of ultrasound images with low signal to noise ratio and rapid movements of anatomically small fetal heart makes ultrasound prenatal diagnosis of cardiac defects as a most challenging task, which can only be done flawlessly by experienced radiologists. This paper demonstrates testing the method of fuzzy connectedness based image segmentation to detect the fetal heart structures from ultrasound image sequences. The proposed work involves Probabilistic Patch Based Maximum Likelihood Estimation(PPBMLE) based image denoising technique as a pre-processing step to remove the inherent speckle noise present in ultrasound images. The second step is use of Fuzzy connectedness based image segmentation algorithm with predefined seed points selected manually inside the fetal heart structure. The results of Matlab based simulation on fetal heart ultrasound dataset proves that the combination of above mentioned image processing techniques was predominantly successful in delineating the fetal heart structures. Quantitative results of the proposed work is apparently shown to illustrate the efficacy of the PPBMLE preprocessing technique.