基于多模态影像融合的肿瘤精准消融方法与系统

Ablation therapy has been a curative method for liver tumors following surgical resection. Conformal and complete tumor eradication and accurate evaluation for treatment effect are two critical points for ablation therapy. Liver is a flexible and moving organ with large deformation, which affects the accuracy of ablation treatment. This project focuses on the precise requirements of ablation treatment, aiming to solve the following complex technical problems using multimodal images techniques. We will introduce self-learning regularization restriction with high generalization performance and prior knowledge to solve the segmentation of the tumor and its adjacent structures, especially the segmentation of tumors and vessels from dynamic ultrasound. We will model structure-preserving mathematics of optimal transmission to realize accurate and rapid intraoperative registration. A functional minimization method will be established under the constraint of high-dimensional data statistics to study the morphological changes of organs and tumors before and after the ablation therapy. The homogeneous perturbation mathematical theory and a new multi-step gradient acceleration algorithm will be developed to simulate the special area's thermal field rapidly. Finally, we will establish a precise and intelligent ablation system for hepatic tumors based on multimodal image fusion, to realize precise localization, intraoperative planning, accurate and rapid calculation and monitoring of thermal field. Through the inter-discipline of mathematics and medicine, this project will not only promote the development of mathematics, but also facilitate ablation therapy access to the era of precise quantification.

消融已成为继手术之后肝脏肿瘤另一种根治性治疗方法，在三维空间上精准适形完全灭活肿瘤和准确评估疗效是消融治疗的关键。肝脏作为形变较大的柔性运动脏器是影响消融精准性的难点所在。本研究针对肿瘤消融的精准化需求，利用多模影像技术重点解决以下问题：利用先验知识，通过引入高泛化能力自学习正则化约束数学方法，解决肿瘤及周边结构的快速精准分割问题，特别是基于动态超声图像的病灶分割与血管提取问题；构建保解剖结构的最优传输数学模型，实现术中精准快速配准；建立高维数据统计约束下的泛函极小化方法，研究消融前后脏器及肿瘤形态变化的规律；通过建立同质化微扰数学理论，设计新的多步梯度加速算法，实现特定区域温度场的快速模拟与仿真；最终建立多模态影像肝癌精准智能消融系统（肿瘤消融精准定位、术中规划、温度场精准快速计算与温控系统）。本项目不仅促进数学的发展，形成新的生长点，而且通过数学与医学交叉助力消融治疗进入精准量化时代。

消融已成为继手术之后肝脏肿瘤另一种根治性治疗方法，在三维空间上精准适形完全灭活 肿瘤和准确评估疗效是消融治疗的关键。肝脏作为形变较大的柔性运动脏器是影响消融精准性的难点所在。本研究针对肿瘤消融的精准化需求，利用多模影像技术重点解决了以下难题：（1）提出了一种理想参数下微波消融热场的解析解方法，该方法计算局部热场仅耗时0.18秒。以解析解温度场先验结合深度学习方法计算个性化消融区与真实消融损伤区域的重叠率（Dice系数）为90.3%，整体预测时间约为0.45秒。（2）采用ICP（迭代最近点）算法和序列最小二乘优化策略，实现CT坐标系与术中导航标记贴坐标系之间的精确配准，单点配准的误差小于0.5毫米。（3）通过完全无监督的方式，提出图像端到端的配准模型，术前术后肝脏配准后平均Dice系数为0.92。（4）对导航系统进行了动物实验和初步的患者评估，综合评估消融治疗术前高效规划、术中精准引导、术后客观评估三个阶段在肿瘤治疗的应用价值。在临床115例HCC术后评估实验中，术后残癌检测率的敏感性和特异性分别达到83.3%和95.2%，较临床医生常规的二维影像评估登指数提升了20.2%。

内容获取失败，请点击重试