肾部分切除手术中图像处理关键问题的研究

Laparoscopic partial nephrectomy (LPN) with segmental renal artery clamping is a new surgical technique developed by Chinese surgeons for the treatment of renal tumors. Compared with the conventional method, the renal tissue with warm ischemia can be minimized by clamping only the segmental renal arteries feeding tumor tissues. The preoperative investigation should be performed to understand the anatomical structures of kidney, renal arteries, collecting system, renal tumor etc.. However, the target arteries and the positions for clamping are determined manually by observing the 3D rendering results of CT dataset in visual. The inter- or intra-observer errors could lead to missed and/or over-clamping. Therefore, a computer-aided system should be developed to measure the feeding area of a selective renal artery quantitatively and determine the target artery as well as clamping positions precisely. To this end, some essential image processing techniques will be studied in this project. The following scientific issues will be studied in this project, (1) to reconstruct a 3D kidney model from multi-phase CT image to demonstrate the anatomy of related tissues, (2) to estimate the blood supply area for a specific renal artery quantitatively and determine the target renal artery as well as the position for clamping, (3) to discover the valuable physiological and pathological information hidden in the datasets of CT and the other image modalities. In this project, two experts from the department of radiology and urology will validate the feasibility and accuracy of the developed algorithms in the clinical practice. This research work will not only improve the accuracy of preoperative planning of LPN, but promote the development of LPN with segmental renal artery clamping.

肾动脉分支阻断的肾部分切除是近年由我国医生提出的肾癌治疗新手术技术，通过阻断肾癌供血动脉分支，避免常规方法造成的正常肾组织热缺血。该技术需在术前充分了解肾脏、肿瘤及肾内部组织的三维解剖结构及位置关系，而目前仅依靠肉眼观察制定手术方案的方式，易造成靶血管或阻断位置判断不准。因此，本课题针对临床亟需解决的实际问题，研究其中涉及的图像处理关键问题，主要包括:（1）基于多期CT序列提取关键组织器官，快速重建直观的三维肾脏模型；（2）提供估计肾动脉分支供血范围的精确定量参数，准确定位靶血管及阻断位置；（3）深入挖掘CT及其他模态数据中隐藏的生理病理学信息。参与本课题的医生将密切配合算法研究，一方面保证研究在临床上的实用性，另一方面为算法的临床验证提供支持。通过本课题研究将有效减少医生工作强度，降低误诊发生的概率，促进诊断及手术技术发展，为准确实施肾动脉分支阻断肾部分切除手术提供可靠保障。

由于肾癌对放疗不敏感，外科手术是局限性肾癌治疗的首选方案。分支肾动脉阻断肾部分切除术是近年来由我国医生提出的一种新型术式，可以在治疗肾肿瘤的同时有效避免正常肾组织的热缺血。本课题针对术前计划依靠人工估计分支肾动脉供血区域，确定术中阻断的靶血管这一不足，利用计算机图像处理及人工智能方法来改善术前手术计划的准确性。主要包括以下研究工作：1、基于深度神经网络的肾脏及肿瘤全自动分割；2、精细肾动脉树分割及供血区域估计；3、基于多期CT图像的肾小球滤过率测量；4、肾肿瘤亚型分类算法研究等。此外，本课题还整理收集了包括近700个病人完整图像数据资料，为本课题及后续研究的开展打下了坚实的基础。本课题研究期间共发表相关论文14篇，其中包括期刊论文8篇，会议论文6篇，授权国家发明专利3项，申请国家发明专利3项。