DEVELOPMENT OF ARTIFICIAL INTELLIGENCE (AI) TOOLS TO UNDERSTAND AND DUPLICATE EXPERTS RADIATION THERAPY PLANNING FOR PROSTATE CANCER

开发人工智能（AI）工具来理解和复制专家的前列腺癌放射治疗计划

• 批准号: 10027481
• 负责人: PRAVEEN SINHA
• 金额: $ 29.94万
• 依托单位: ONCOSPACE, INC.
• 依托单位国家: 美国
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-09-16 至 2020-06-15
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10027481
• 关键词: Anatomy; Artificial Intelligence; Cancer Patient; Categories; Clinical; Consensus; Data Set; Databases; Dose; Generations; Goals; Knowledge; Malignant neoplasm of prostate; Manuals; Methods; Modeling; Outcome; Patients; Prostate; Prostate Cancer therapy; Quality Control; Radiation Dose Unit; Radiation Oncologist; Radiation therapy; Risk; Savings; Structure; Structure of base of prostate; System; Time; Training; Validation; base; high risk; improved; knowledge base; personalized medicine; programs; prostate cancer risk; tool; treatment planning

Use of the Oncospace framework for AI-based prostate radiotherapy treatment planning The goal of this program is to establish a fully automated treatment planning method for low, intermediate, and high-risk prostate cancer patients undergoing radiation therapy. The program will use artificial intelligence (AI) methods trained with our existing database of over 1600 prostate treatment plans including clinical outcomes. Results inherently reflect the expert judgement of the original planners regarding dosimetric tradeoffs. The methods will include integrity checking of normal anatomy and target volume structures to identify potential manual or automatic contouring errors. The system will be used to automate treatment plan generation from the point when structures have been drawn until the plan requires expert assessment/approval. Plans for patients in each risk category will be validated against those generated and approved by three expert radiation oncologists, where assessment criteria is decided upon a priori by consensus between experts. Benefits include time savings via use of knowledge-driven versus trial and error-based plan objectives, more personalized treatment plans and improved quality control.

使用Oncospace框架进行基于AI的前列腺放疗治疗计划 该计划的目标是为接受放射治疗的低，中，高风险前列腺癌患者建立一个完全自动化的治疗计划方法。该计划将使用人工智能（AI）方法，并使用我们现有的1600多个前列腺治疗计划（包括临床结果）的数据库进行训练。结果本质上反映了原计划者关于剂量权衡的专家判断。这些方法将包括正常解剖结构和目标体积结构的完整性检查，以识别潜在的手动或自动轮廓绘制错误。该系统将用于自动生成治疗计划，从绘制结构开始，直到计划需要专家评估/批准。针对每个风险类别的患者的计划将根据由三名放射肿瘤专家生成和批准的计划进行验证，其中评估标准由专家之间的共识决定。受益包括通过使用知识驱动而不是基于试验和错误的计划目标节省时间，更个性化的治疗计划和改进的质量控制。