Bimodal Tactile Tomography with Bayesian Sequential Palpation for Intracavitary Microstructure Profiling and Segmentation

Le résumé fourni par la source

Robotic palpation for in-situ tissue biomechanical evaluation is crucial for diagnosing diseases in luminal organs, particularly in the early diagnosis of bladder cancer. Although commercial surgical robotic systems offer tactile feedback, they lack tactile intelligence and autonomous decision-making, hindering thorough tissue assessment. Endoscopic Optical Coherence Tomography (OCT) offers real-time, three-dimensional visualization of tissue microstructures but does not address the tactile sensing needed for lesion profiling. To bridge this gap, we developed a bimodal robotic palpation strategy using the OCTbased tactile sensor ElastoSight, which employs circumferential and sliding B-scan modes with Bayesian optimization for accurate lesion detection. Our technique localizes the tumor phantom's center within 30 iterations, achieving F1 scores over 0.976 and centroid errors below 0.032 mm. The sliding B-scan allows shape segmentation of hard tissue from soft tissue, with a precision rate of 0.983 and area error below 0.25 mm 2. These results demonstrate that our technique effectively addresses real-time lesion localization and segmentation, showing strong performance in simulations and experiments.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.