OBJECTIVE

To address the challenges of low surgical precision and poor consistency in focused ultrasound ablation surgery (FUAS) for uterine fibroids, which are often caused by variations in clinical experience and operator fatigue, this study aims to develop an intelligent three-dimensional (3D) visualization and navigation system by integrating magnetic resonance imaging (MRI) with real-time ultrasound (US) imaging, thereby improving the accuracy and efficiency of uterine fibroid surgery.

METHODS

MRI and US images from 638 patients were annotated by experienced clinicians. The nnU-Net algorithm was used for preoperative segmentation and 3D reconstruction of MRI images to provide detailed visualization of fibroid morphology. The YOLACT model was applied to achieve rapid delineation of the uterus and key anatomical structures in real-time US images. To enhance the accuracy of lesion localization and navigation, the Iterative Closest Point (ICP) algorithm was employed for the registration of preoperative MRI with intraoperative US images.

RESULTS AND DISCUSSION

Experimental results demonstrated that the system achieved a Dice Similarity Coefficient (DSC) exceeding 90% for the segmentation and identification of anatomical structures such as the uterus and fibroids. The YOLACT model achieved an accuracy greater than 95% in identifying key structures in real-time US images. In 90% of the cases, the system enabled efficient and precise tracking; however, approximately 5% of the cases required manual adjustment due to discrepancies between patient anatomy and preoperative MRI data. The proposed intelligent navigation system, based on MRI-US image fusion, offers an efficient and automated solution for FUAS in treating uterine fibroids, significantly improving surgical precision and operational efficiency. This system demonstrates strong clinical applicability. Future research will focus on enhancing the adaptability of the system, particularly in addressing challenges such as significant tissue deformation and occlusion, to improve its robustness and applicability in complex clinical scenarios.