An artificial intelligence-driven distributed stereotactic radiosurgery strategy for multiple brain metastases management

项目总结 脑转移瘤(BMS)是一种危及生命的神经系统疾病，但目前的治疗方案不能 管理多个(&gt；4)BMS(MBMS)，而不会产生强烈的不良影响。立体定向放射外科(SRS)， 利用强剂量照射骨髓间充质干细胞，并迅速减少对周围组织的照射，已被证明是一种有效的 数量有限、体积较小的骨髓瘤的治疗方案。然而，当SRS不能避免高毒性剂量时 BM是多个、聚集或毗邻关键器官。要安全有效地治疗患有SRS的MBMS，需要 解决这些迫切需要：1)确定最大可耐受SRS剂量；2)研究神经认知 降低和设计策略，以保持患者治疗后的生活质量；以及3)制定和实施 高质量的简化MBMS SRS治疗和后续护理。 为了满足MBMS SRS的管理需求，我们的目标是开发和实施人工智能(AI)- 驱动治疗计划系统(TPS)和进行治疗干预临床试验，两者都致力于 提高MBMS SRS治疗质量和效率。人工智能驱动的TPS，即AimBMS，将有三个人工智能- 基于计算模块，包括用于自动分割的AI-Segtor，用于治疗的AI-Predictor 时空分布式SRS计划优化的结果预测和AI-Planner。AimBMS最初是 基于回顾数据开发并促进MBMS分布的SRS前瞻性I/II期临床 试验，而临床试验将提供关键的临床知识和证据作为反馈，以改进AimBms 性能。该项目的最终目标是将AimBMS转化为常规的临床实践，以改进 MBMS SRS治疗质量、患者治疗后生活质量和临床设施工作流程。 为了响应PAR-18-560，我们已经在放射肿瘤学家和 医学物理学家将开发一种新的人工智能驱动的分布式SRS技术并进行针对癌症的研究 管理MBMS的治疗性干预。该项目的创新包括：1)新颖的SRS治疗方案 基于人工智能的自动分割、治疗结果预测和 时空规划优化；2)新的人工智能学习能力，提高开发的人工智能工具的性能 通过连贯的临床试验。技术的发展将为治疗干预临床提供支持 试验，而临床试验被指定为改善开发的系统性能。这是天衣无缝的 集成开发模式保证了所开发系统的临床实用性。建成后，我们的新 开发的AimBMS将为MBMS SRS管理奠定坚实的基础，并使更多的人受益 患有BMS的患者。此外，基于人工智能的治疗规划和治疗交付基础设施是为 MBMS SRS可以转移到其他肿瘤部位，产生更广泛的临床影响。

PROJECT SUMMARY Brain metastases (BMs) are a life-threatening neurological disease, but current treatment regimens cannot manage multiple (>4) BMs (mBMs) without causing strong adverse effects. Stereotactic radiosurgery (SRS), utilizing potent dose to irradiate BMs and quick dose falloff to spare nearby tissues, has proven to be an effective treatment regimen for limited-number and small-size BMs. However, SRS could not avoid high toxic dose when BMs are multiple, clustered, or adjacent to critical organs. To safe and effectively treat mBMs with SRS requires addressing these urgent needs: 1) to identify the maximum tolerable SRS dose; 2) to study neurocognitive decline and design strategies to preserve patients’ post-treatment quality of life; and 3) to develop and implement high-quality streamlined mBMs SRS treatment and follow-up care. To address mBMs SRS management needs, we aim to develop and implement an artificial intelligence (AI)- driven treatment planning system (TPS) and conduct a therapeutic intervention clinical trial, both dedicated to improve mBMs SRS treatment quality and efficiency. The AI-driven TPS, namely AimBMs, will have three AI- based computational modules, including AI-Segtor for automatic segmentation, AI-Predictor for treatment outcome prediction and AI-Planner for spatiotemporal distributed SRS plan optimization. AimBMs is initially developed based on retrospective data and facilitate the mBMs distributed SRS prospective phase I/II clinical trials, while the clinical trial will provide critical clinical knowledge and evidence as feedback to improve AimBMs performance. The ultimate goal of the project is to translate the AimBMs to routine clinical practice to improve mBMs SRS treatment quality, patients’ post-treatment QoL, and clinical facility workflow. In response to PAR-18-560, we have formed a multidisciplinary collaboration between radiation oncologists and medical physicists to develop a novel AI-driven distributed SRS technology and conduct a cancer-targeted therapeutic intervention for managing mBMs. The project’s innovations include: 1) novel SRS treatment planning technological capability enabled by AI-based auto-segmentation, treatment outcome prediction, and spatiotemporal plan optimization; 2) novel AI learning capability to improve developed AI tools’ performance through the coherent clinical trial. The technology development will support the therapeutic intervention clinical trial, while the clinical trial is designated to improve the developed system performance. This seamlessly integrated development mode ensures the developed system is clinically practical. Upon completion, our newly developed AimBMs will lay a solid foundation for mBMs SRS management and benefit a wide population of patients with BMs. Moreover, the AI-based treatment planning and treatment delivery infrastructure built for mBMs SRS can be transferred to other tumor sites to generate an even broader clinical impact.