Development of optical flow and segmentation algorithms for foetal motion

The development of segmentation and optical flow algorithms for foetal motion estimation using foetal ultrasound images is explained. The current paper work explains the development of enhancement and segmentation algorithms as the ultrasound diagnosis highly depends on the quality of the data. Developed algorithms segment the foetus and helps in analyzing the anatomical structure. Applying optical flow algorithms on the segmented image will help in analyzing the motion along the anatomical parts of the image. This makes an easy diagnostic tool for the obstetricians to analyze motion of the foetus and the data can be stored in the database for further analysis.

Median filtering is used to remove speckle noise in the ultrasound image for better visualization. Removing the noise will also help in segmenting the image. Adaptive thresholding and Prewitt's edge detection are used to segment the foetus after which optical flow algorithms are implemented for contour motion analysis to help analyzing and estimating the motion of the foetal anatomical parts like the head, neck, limbs etc and the entire foetal motion in the mother's womb.

In the present study, segmentation algorithm for the analysis of foetal anatomical structure along the contour with Gradient based Lucas Kanade optical flow algorithm using Sobel operator and Savitzky Golay filter and Phase based optical flow algorithm is developed and implemented on the segmented and unsegmented foetal ultrasound image. Sobel operator and Savitzky Golay filter perform smoothening and differentiation operation for estimating the motion which is compared on its computation time and accuracy to arrive at the best operator for Lucas Kanade method for motion estimation. The optical flow algorithms are implemented on foetal ultrasound image sequences and other image sequences to compare the performance of both the algorithms. The present study shows that even in the presence of local noise, the Lucas Kanade optical flow algorithm using Savitzky Golay filter gives better results for foetal ultrasound images (segmented and unsegmented) compared to results from phase based optical flow algorithm.