Preserving Erectile Function by Quantifying the Nerve-Sparing step of the Robotic Prostatectomy

摘要 外科医生之间的表现差异导致患者结局的差异，但 如果外科医生不了解外科手术的技术考虑因素， 这将有助于他们改善结果。作为一个典型的例子， 前列腺癌根治性前列腺切除术（RARP）可导致患者的 勃起功能（EF）恢复（10-50%）。客观评估外科医生的可靠方法 与患者结果密切相关的性能通常缺乏。 在这个项目中，作为一个测试案例，量化外科医生的表现，以改善病人 结果，我们将重点评估外科医生的神经保留（NS）剥离质量， 通过手术视频和患者结局评价RARP。细致入微的NS步骤是一个 良好的测试用例，因为它是可量化EF结局的主要决定因素，RARP是 常见手术（约145，000例/年），手术视频可随时用于分析。 我们将通过三个独立但互补的目标来实现我们的目标。目标1：我们 寻求通过专家共识确定共同的技术考虑因素， 这是最佳执行NS步骤以恢复EF所必需的。目标2：我们将开发一个 通过手术的计算机视觉分析的自动化性能评估管道 视频.目标3：我们将开发和验证一个技能反馈评估工具，以证明 概念NS专用VR仿真。 拟议工作的主要区别在于，我们将量化最相关的技术 组织剥离的考虑因素驱动患者报告的结局。外科医生 参与这项研究的人不仅可以通过他们的手术视频提供数据， NS步骤，但他们也将提供RARP中的真实的患者EF结局数据， 建立外科医生技能、患者因素和EF结果之间的关系。 统计建模将描述外科医生技能和患者因素对EF的不同影响 结果。此外，我们将利用基于深度学习的计算机视觉来全面捕捉所有 NS技术和技能的许多方面，以帮助确定它们如何有助于 最终的EF结果。 拟议的工作将实现可扩展和可操作的反馈，使外科医生能够 有价值的知识，以最大限度地提高手术效果。RARP后NS阶跃和EF恢复 将作为我们未来自动化评估的测试案例，以改善任何 外科手术

ABSTRACT Variation in performance between surgeons leads to differences in patient outcomes, but surgeons cannot improve if they are not aware of the technical considerations for a surgical procedure that will allow them to improve outcomes. As a prime example, robot-assisted radical prostatectomy (RARP) for prostate cancer can lead to highly variable rates of patient erectile function (EF) recovery (10-50%). Yet reliable means of objectively assessing surgeon performance, that strongly associate with patient outcomes, are generally lacking. In this project, as a test case for quantifying surgeon performance to improve a patient outcome, we will focus on assessing a surgeon’s nerve-sparing (NS) dissection quality during RARP through the evaluation of surgical video and patient outcomes. The nuanced NS step is a good test case because it is the primary determinant of the quantifiable EF outcome, RARP is a common procedure (~145,000 cases/year), and surgical video is readily available for analysis. We will accomplish our objective with three independent, yet complementary aims. Aim 1: We seek to determine through expert consensus the common technical considerations necessary to optimally perform the NS step for EF recovery. Aim 2: We will develop an automated performance assessment pipeline through computer vision analysis of surgical video. Aim 3: We will develop and validate a skills feedback assessment tool for a proof-of- concept NS-specific VR simulation. The primary differentiator of the proposed work is we will quantify the most relevant technical considerations for tissue dissection driving a patient reported outcome. Surgeons participating in this study will not only provide data through surgical videos of them performing the NS step, but they will also contribute real patient EF outcome data from the RARP to establish the relationship between surgeon skill, patient factors, and EF outcome. Statistical modeling will delineate the differential impact of surgeon skill and patient factors to EF outcome. Further, we will harness deep learning-based computer vision to holistically capture all the numerous facets of NS technique and skill to help determine how they contribute to the ultimate EF outcome. The proposed work will enable scalable and actionable feedback, empowering surgeons with valuable knowledge to maximize surgical outcome. The NS step and EF recovery after RARP will serve as our test case for future automated assessments to improve outcomes in any surgical procedure.