Automatic entropy-based femur segmentation and fast length measurement for fetal ultrasound images

Abstract

Accurate fetal biometric ultrasound measurements are important for high quality obstetrics health care, used to determine the gestational age and the growth rate of the fetus, and hence served as important diagnostic tools. In modern ultrasound scanners, fetal measurements are made by manually extracting diameters or contours from ultrasound images. However, manual measurements are usually inaccurate and inconsistent. An automatic technique for acquiring biometric fetal measurements and robust to the pose and appearance variations exhibited in fetal ultrasound scans is highly desirable. In this paper, we focus on the femur measurement, and introduce a new automatic method for femur segmentation from ultrasound images and an automatic method for quickly computing the length of the segmented femur. In evaluation, the presented work is evaluated using the clinical dataset of the `Challenge US: Biometric Measurements from Fetal Ultrasound Images as part of ISBI 2012'. The challenge data consisted of 90 fetal ultrasound images acquired at 21, 28 and 33 weeks of gestation. Experimental results show that the presented approach is effective for the purpose of fetal biometric measurements of the femur and achieved 1st place on the automatic femur segmentation sub-challenge in ISBI 2012.