Human–Robot Shared Control Strategy for Ultrasound Scanning Based on Lesion Position Prediction

In minimally invasive breast surgery, it is crucial to identify the optimal scanning plane of the lesion and plan the surgical instrument’s trajectory within this plane. When the surgical instrument interacts with breast tissue, the optimal scanning plane may shift, making consistent tracking challenging. Previous studies have explored visual servoing to enhance tracking performance. However, image features alone are insufficient to address the complex and continuous deformation of soft tissues. To overcome the challenges of unstable scanning plane tracking and increased puncture errors during breast puncture, a human-robot shared ultrasound scanning framework is proposed. The framework integrates PF-Net-based point cloud completion with a modified PointNet lesion prediction model to achieve dynamic tracking of the lesion’s optimal scanning plane. Compared with traditional leader-follower control, the proposed method shortens the probe trajectory by 45.7%, increases the lesion cross-sectional area ratio by 13.9%, and improves ultrasound image confidence by 6.4%, thereby enhancing the operator’s overall scanning performance. This approach enables stable tracking and accurate imaging of internal lesions under ultrasound guidance, providing technical support for precise lesion localization and subsequent minimally invasive excision. Note to Practitioners—This study aims to address the challenges of lesion displacement and the difficulty in continuously tracking the optimal scanning plane during ultrasound-guided breast interventions. During puncture, the interaction forces between the needle and tissue can cause lesion movement, while the lack of realistic haptic feedback in leader-follower control makes it difficult for the surgeon to perceive and adjust the probe position in time, thereby affecting the accuracy of lesion plane tracking and localization. To overcome these issues, we propose a human-robot shared control framework for ultrasound scanning based on real-time lesion position prediction. The framework endows the follower robot with local autonomy to adaptively track the optimal scanning plane, effectively improving the accuracy of lesion localization and the operational stability during minimally invasive breast surgery.