OCT-DeformNet: Optical Coherence Tomography-Guided Biological Tissue Shape Prediction for Robot Palpation in Microsurgery
Guangshen Ma, Tianhao Qin, Jiawei Liu, Mark Draelos
AI summary
Problem
Predicting biological tissue deformation during robot palpation is challenging due to complex tissue mechanics, varying tool geometries, and limited intraoperative sensing, which hinders real-time visualization and surgical planning in microsurgery.
Approach
The authors propose OCT-DeformNet, a data-driven framework that uses a multilayer perceptron to map robot tooltip movements and pre-deformed surface geometry directly to post-deformation shapes, guided by high-fidelity OCT imaging.
Key results
- Developed an OCT-guided data collection platform for palpation-induced deformation
- Created an MLP model predicting post-deformation surfaces from tooltip motion without prior tissue knowledge
- Validated on phantoms and ex vivo tissues with average errors of ~0.15 mm and ~0.52 mm
- Enabled real-time inference for surgical planning and simulation
Why it matters
This framework advances micro-scale deformation sensing and modeling, providing a broadly applicable tool for soft-tissue research and robotic microsurgery planning.
Abstract
In medical robotics, biological shape deformation resulting from arbitrary tool-tissue interaction commonly oc- curs and motivates the need in microsurgery to predict the shapes of tissue structures. However, handling deformation is challenging due to the lack of generalized prediction mod- els for various surgical scenarios, complex tissue properties, and different surgical tool geometries. Limited intraoperative sensors to observe microlevel deformations further underscore this difficulty. This paper proposes a novel geometric data- driven framework that uses only the robot palpation tooltip movement and a pre-deformed surface to predict the tissue deformation by using the optical coherence tomography (OCT) sensor. A multilayer perceptron model is trained to learn tool- tissue physics and predict the shape from the given robot-tool configurations represented as orientations and displacements. We conducted realistic experiments to verify the models using phantoms of various stiffness and three ex vivo tissue types, with average prediction errors of approximately 0.15 mm and 0.52 mm respectively. This OCT-guided data-driven platform en- ables micro-scale palpation data collection and model training, and is broadly applicable to soft-tissue research in biomedical engineering and surgical robotics.