Image-Free Tumor Segmentation of Soft Tissue Using a Minimally Invasive Robotic Palpation System

Abstract

Tumor segmentation is crucial for surgical planning and precise tumor resection for effective treatment. Traditionally, tumor localization has been performed using medical imaging techniques such as CT and MRI or through direct palpation by surgeons. However, in minimally invasive robotic surgery (MIS), these methods have limitations, including registration errors with imaging and inaccuracies caused by the subjectivity of palpation by surgeons. In this study, we introduce a robotic palpation system and an image-free process for MIS tumor segmentation using a robot. Our proposed system enables precise tumor shape differentiation through direct robotic palpation. For this, the robotic palpation system collects surface shape information through the proposed process, allowing tissue palpation at specific depths according to surface curvature. Additionally, it visualizes stiffness maps, enabling image-free tumor segmentation. In experiments using this system, evaluation of planar and curved phantom models demonstrates precise segmentation at targeted sites, with sensitivities of 0.9634 and 0.9729, and specificities of 0.9646 and 0.9878, respectively. Validation on ex-vivo porcine liver models further confirms the efficacy of our approach.