Segmentation and Registration of Ultrasound Images of Uterine Fibroids for USgHIFU

Ultrasound-guided high-intensity focused ultrasound(USgHIFU) is a minimally invasive ablation treatment method for uterine fibroids. It completes the image guidance of the HIFU ablation operation by acquiring ultrasound images of the patient in real time. When HIFU ablates the nourishing arteries of fibroids, it causes arterial vasoconstriction and block blood vessels from delivering nutrients to fibroids and induce shrinkage of uterine fibroids. Since the guidance ultrasound imaging integrated on the USgHIFU treatment head is deeper and will be affected by the flowing water in the water bladder, it is difficult to use the ultrasound probe integrated on the treatment head to collect Doppler color flow imaging. External handheld ultrasound acquisition color Doppler flow imaging(CDFI) is needed to assist the nourishing artery ablation operation. This process requires sonographers to manually identify blood vessels. This study proposes a method to automaticly segment and register USgHIFU guidance ultrasound images and handheld ultrasound images. Firstly, use ReFineNet to segment complete fibroids contours in handheld ultrasound images and manually label upper boundaries of fibroids in guidance ultrasound. Then, use iterative nearest point(ICP) and shape context to register two image. In this study, a clinical ultrasound dataset was established to verify the method. Dice of segmentation can reach 0.879, mean distance error(MDE) of registration is less than 1mm.