Segmentation and Feature Extraction of Infants’ Hip Joint Ultrasound Images

In this paper, we proposed a segmentation method for infant hip joint ultrasound images as well as an extraction method of interested region. The method segments the infant hip joint of ultrasound image twice to achieve the optimal segmentation result. The problem of intensities inhomogeneities existing in medical images is solved by mapping the evolved level set functions in the first segmentation procedure into a new, homogenous gray level image. In the second segmentation, shape prior and regularization term are introduced into the original CV model to optimize the segmentation results. The shape prior is used again after segmentation to extract interested regions and related features from segmented image. Experimental results show that our method achieves better segmentation results than existing methods. Besides, our method is superior to other methods on extracting regions of interest.