TOPIC 389 - INTELLIGENT SOFTWARE FOR RADIATION THERAPY PLANNING

放射治疗计划的智能化软件 除手术外，放射治疗是局限性前列腺癌最有效的治疗方式。放射治疗的成功源于对肿瘤反应和正常组织耐受性的治疗窗口的利用，这最大化了消灭肿瘤的机会，同时使周围的正常组织免受严重损伤。这需要从对患者的医学图像的检查中生成准确且个性化的辐射剂量分布。放射治疗在技术上是复杂的和劳动密集型的。在整个患者护理过程中，需要密集的人力监督和干预。 人工智能和机器学习技术擅长于自动化工作流程和任务-在这种情况下，癌症治疗计划的制定。我们相信，这种机器学习在解决现有治疗计划工作流程中的固有问题方面具有令人难以置信的潜力。我们的方法将缓解当前治疗计划的问题，与日常操作无缝集成，并将作为一种治疗解决方案，以限制风险并延长生命的方式使患者受益。 该项目的目标是开发一个从开始到结束的前列腺癌治疗计划系统。我们将与三个放射肿瘤学专家团队合作，将现有的患者数据存档，用于人工智能算法的训练。这些算法将使高质量的计划，以轻松生成。在第一阶段完成时，我们将有一个原型界面，允许用户使用人工智能算法创建治疗计划和分割。此外，用户将能够根据其他专家计划和训练数据评估这些计划。

INTELLIGENT SOFTWARE FOR RADIATION THERAPY PLANNING Besides surgery, radiotherapy is the most effective treatment modality for localized prostate cancer. The success of radiotherapy stems from the exploit of a therapeutic window in tumor response and normal tissue tolerance which maximizes the chance of sterilizing the tumor while sparing the surrounding normal tissue from severe damage. This requires an accurate and individualized radiation dose distribution generated from the examination of the patient’s medical images. Radiation therapy is technically complex and labor intensive. Intensive human supervision and intervention are needed throughout the path of patient care. Artificial intelligence and machine learning technologies are adept at automating workflows and tasks—in this case the development of cancer treatment plans. We believe that this machine learning has incredible potential to address inherent problems in the existing treatment planning workflow. Our approach will mitigate the current issues with treatment planning, integrate seamlessly with day-to-day operations, and will serve as a treatment solution that benefits patients in a way that limits their risk and extends their lives. The goal of this project is to develop a start-to-end prostate cancer treatment planning system. We will work with three expert radiation oncology teams to archive existing patient data for use in training of artificial intelligence algorithms. These algorithms will enable high quality plans to be generated with ease. At the completion of Phase I, we will have a prototype interface that will allow users to create treatment plans and segmentation using artificial intelligence algorithms. Additionally, users will be able to evaluate these plans against other expert plans and against the training data.