面向虚拟钻骨手术的操作力模型构建及力触觉反馈模拟研究

Bone drilling is the essential surgical skill and widely used in orthopedic surgery. The surgeon must be very careful to apply appropriate forces and to operate the tools at appropriate speeds to avoid over-operation. Becoming a skillful surgeon with sensitive force perception requires rigorous training and repeating practice. Traditionally, the medical students used the models, animals or cadavers to train their bone machining skill. But this method is unrealistic, non reusable or expensive. A visuo-haptic surgical simulator provide the virtual environments for simulating real-world, and employ the haptic devices for force feedback. It can provide a safe, cost-effective, and repeatable training method for surgical training. In order to provide the high fidelity haptic rendering for bone drilling, this research focuses on the bone drilling force model and real time haptic rendering. With two data samples with different confidence level, the artificial neural network will be trained with two steps. The simulation samples are from the finite element analyses results and the true samples are the measured data in the cadaveric bone. In this way, the convergence rate and the predictive precision of the force model can be guaranteed. Furthermore, the voxel based model with independent bone density value from CT images is constructed for the bone. By employing parallel algorithm, the collision detection between the cutting lips of the drill bit and the voxels of the bone model can be accelerating accomplished. Moreover, The evaluation and validation of the bone drilling simulation will be conducted to demonstrate the skill training effects. This research will provide the a raining alternative in bone drilling for novices.

骨科医生必须经过严格训练和反复实践后才能掌握手术操作的精准“手感”，具备敏感的力感知能力。传统的基于模型、动物或尸体的手术训练方式因效率低、成本高、灵活性差等弊端，导致医生的成长周期长。为了让医生高效、精确地掌握钻骨操作手感，本项目致力于虚拟钻骨操作的精确模拟研究，解决力预测模型的精确建立及力触觉交互的实时模拟两个关键问题。针对骨组织不均匀性及个体差异特性，项目拟建立与骨密度值相关的钻骨操作力预测模型，通过大样本低置信度有限元分析模拟数据和小样本高置信度尸体测量数据，对力预测模型进行二次学习，以保证模型收敛速度和预测精度；建立具有独立体素特征的骨组织物理模型，通过并行算法加速解析钻头切削刃与体素间碰撞检测，结合用户输入参数，通过力预测模型实时计算钻骨操作力，实现力触觉交互模拟；通过系统信效度分析来评价钻骨操作模拟的精度和训练效果，为骨科手术提供一种精确、高效、灵活的虚拟训练方式。

针对传统虚拟钻骨手术训练存在的诸多问题，本项目基于力反馈技术和虚拟手术技术，致力于为受训者提供精准的虚拟钻骨力并实现实时的力触觉交互反馈。本项目通过对钻骨过程进行有限元仿真和钻骨过程实验数据采集与分析，确定了钻骨力与主要影响因素的关系，同时通过神经网络模型得到关于骨密度、钻头钻削速度和转速的钻骨力预测模型。在力触觉交互的实时模拟方面，针对骨组织不均匀特性，建立了与骨密度值相关的具有独立体素特征的骨组织模型，基于此模型设计了快速碰撞检测、体素去除等算法，保证虚拟钻骨过程中反馈力的准确性与稳定性。项目开发完成了“脊柱椎弓根钉植入手术训练系统”，并开展了系统信效度分析，评价了钻骨操作模拟的精度和训练效果。分析结果表明该系统为骨科手术提供了一种更为高效灵活的训练方式，为解决我国手术训练成本高效率低等问题提供了全新解决思路，并提供了核心算法与技术支持。通过