Curvelet Based Seed Point Segmentation for Abnormality Detection in Fetus us Images

Objectives: A new method for automatic segmentation is designed for improving accuracy and abnormality detection in fetus Ultra Sound images at an early stage. Methods: Curvelet-based Seed Point Selection (S-CSPS) method is developed for improving abnormality detection rate using 𝑘-means segmentation algorithm. Correct identification of regions for each pixel that belongs to the objects in US images is obtained through seed point evaluation reducing the speckle and therefore resulting in the improvement of abnormality being detected. Findings: The proposed S-CSPS is implemented in MATLAB platform using ultrasound images of anomalies of fetal spine dataset. The performance analysis is conducted on factors such as noise, segmentation accuracy abnormality detection rate and segmentation time with respect to a different number of fetus images. Improvement: The simulation analysis results shows that S-CSPS method offers better performance with an improvement of segmentation accuracy and improving the abnormality detection rate compared to state-of-the art methods.