深度学习3D肌肉控制人体下肢生物力学模型及其应用研究

The human biomechanical model is an important common basis for studies of human injury mechanisms, pathologies, and corresponding prevention and treatment device design. Diabetic foot is one of the most common and extremely serious long-term complications of diabetes. Reasonable distribution of plantar pressure is one of the main methods to solve this problem. However, there are still some problems like the influences of individual muscle strength differences on the mechanism of diabetic foot ulcers, which lead to unreasonable design of the special insole. Therefore, this project focuses on the study of 3D muscle individuality modeling based on the differences in the characteristics of living muscles and the construction of human lower extremity biomechanics model coupled with deep learning closed-loop muscle control method, in order to explore the mechanism of diabetic foot ulcers under gait loading and optimal design of the individualized insoles. The project firstly measures the characteristics of human living muscles in motion by combining ultrasound mechanical imaging and other techniques, and compares the motion data to drive the individualized 3D modeling of muscles; then the 3D active muscle human body lower extremity model is built by coupling PD closed-loop control method based on deep learning to achieve human gait modeling by finite element simulation and reveal the effects of muscle forces on the formation of diabetic foot ulcers in gait. At the same time, taking into account the computational efficiency, an individualized insole rapid optimization design method based on the rational distribution of dynamic plantar pressure is established through an agent model, and complete engineering verification.

人体生物力学模型是人体损伤机理、病理及相应预防与治疗器械设计等研究的重要共性基础。糖尿病足是一种最常见且极为严重的糖尿病长期并发症，合理分配足底压力是解决该问题的主要方法之一，但尚存在个体肌肉力差异对糖尿病足溃疡形成影响机理不明确等问题进而造成专用鞋垫适配设计的不合理。因此，本项目重点研究基于活体肌肉特性差异的3D肌肉个性化建模，及耦合深度学习闭环肌肉控制方法的人体下肢生物力学模型构建，以探究步态加载下糖尿病足溃疡机理及个性化鞋垫的优化设计。项目首先结合超声力学成像等技术在线测量运动中人体活体肌肉特性，对比运动数据以驱动肌肉的个性化3D建模；继而耦合基于深度学习的PD闭环控制方法建立3D主动肌肉人体下肢模型，实现人体步态的有限元模拟，以揭示步态中肌肉力作用对糖尿病足溃疡形成的影响规律；同时兼顾计算效率，通过代理模型建立基于足底动态压力合理分布的个性化鞋垫快速优化设计方法，并完成工程化验证。

人体生物力学模型是人体损伤机理、病理及相应预防与治疗器械设计等研究的重要共性基础。糖尿病足是一种最常见且极为严重的糖尿病长期并发症，合理分配足底压力是解决该问题的主要方法之一，但尚存在糖尿病足溃疡形成机理不明确等问题进而造成专用鞋垫或足矫形器适配设计的不合理。本项目结合医学影像学与机械加载实验研究了人体活体骨骼肌与足底软组织三维建模参数的个性化反求；耦合神经动力学算法、肌骨动力学与有限元分析方法建立了人体下肢的动态生物力学分析方法，实现了足踝步态动力学分析与内部组织应力应变响应的同步监测；基于批量临床患者实验明确了糖尿病患者与正常人下肢肌肉及足底软组织的差异，并结合步态仿真分析揭示了糖尿病患者足溃疡发生的机理；进而基于自主提出的足底压力中心线自适应调节概念完成了适用于糖尿病、扁平足及糖尿病合并扁平足患者的仿生鱼骨足矫形器设计与数字化制造，并通过>60名患者临床实验验证了其有效性与可靠性。研究成果解决了下肢康复医疗器械性能评估与优化设计缺乏虚拟仿真评估工具与方法的部分难题，为高端、个性化康复医疗器械的设计提供了理论基础与方法参考。