Prostate deformation analysis during ultrasound-guided radiotherapy procedures with transformative models

前列腺癌的发病率很高（每7名男性中就有1名），是北美男性癌症死亡的第二大原因。为此，高剂量率（HDR）近距离放射治疗提供了一种剂量调制的方法来治疗非肿瘤的能力，以优化临床性能，同时避免可能干扰患者健康的敏感区域。工业合作伙伴Elekta Ltd专门从事放射治疗和放射性近距离治疗系统，用于乳腺、肺、肝或肺癌患者的局部治疗。Elekta开发的技术的一个关键优势是整合术前成像，以创建剂量计划并根据患者的MRI治疗肿瘤，在治疗阶段调整剂量调制。然而，HDR近距离放射治疗广泛使用的一个主要限制在于介入过程中前列腺的不自主运动或局部变形.事实上，这些放射治疗系统依赖于在放射治疗期间对肿瘤的准确跟踪，这可能受到前列腺变形的高度影响。这可能导致局部治疗对周围组织的毒性，并增加局部持续或复发的风险。该项目的目标是开发计算机化工具，该工具将基于计算机视觉和机器学习方法测量前列腺中引起的变形，最终补偿图像引导治疗期间的变形。本项目的目标如下：（1）创建一个接口来加载术前（MRI）和术中采集（2）实现前列腺的3D形状建模工具;（3）通过使用非参数化几何方法开发前列腺配准算法来计算4D变形场;（4）临床验证，以验证术前和术中采集之间的前列腺变形。这项工作的一个主要影响将是使组织变形分析在介入环境中精确的剂量绘画结合多参数磁共振成像，潜在地减少复发，并尽量减少正常前列腺过度照射的风险。

With its high incidence rate (1 in 7 men), prostate cancer is the second leading cause of cancer death in men inNorth America. Towards this end, high-dose-rate (HDR) brachytherapy offers the ability to treat prostatecancer with a dose-modulated approach for optimized clinical performance, while avoiding sensitive zones thatmight interfere with the patient's well-being. The industrial partner, Elekta Ltd, is specialized in radiotherapyand radioactive brachytherapy systems used for focal treatments in patients with breast, lung, liver or prostatecancer. A key advantage of the technology developed by Elekta is the integration of pre-operative imaging tocreate a dose-plan and treat the tumor based on the patient's MRI, adapting the dose modulation during thetreatment phase. However a major limitation in the widespread use of HDR brachytherapy resides in theinvoluntary movement or local deformation of the prostate gland during the intervention. Indeed, theseradiotherapy systems depend on accurate tracking of the tumor during radiation treatment, which can be highlyinfluenced bu the deformation of the prostate. This can cause possible toxicity of local therapy to surroundingtissue and increase the risk of local persistence or recurrence. The goal of this project is to develop acomputerized tool which will measure the deformation induced in the prostate based on computer vision andmachine learning approaches, ultimately compensating the deformation during image-guided treatments. Thisproject will have the following objectives: (1) Create an interface to load the pre-operative (MRI) andintra-operative acquisitions (3D ultrasound or MRI); (2) Implement a 3D shape modelling tool of the prostate;(3) Compute a 4D deformation field, by developing a prostate registration algorithm using a non-parametricdiffeomorphic approach; (4) Clinical validation to verify prostate deformation between pre-operative andintra-operative acquisitions. A major impact of this work will be to enable tissue deformation analysis in aninterventional environment for precise dose-painting in combination with multiparametric MR imaging,potentially reducing recurrence and minimizing the risk that normal prostate is excessively irradiated.