A multicenter study in bronchoscopy combining Stimulated Raman Histology with Artificial intelligence for rapid lung cancer detection - The ON-SITE study

项目概要和摘要 肺癌占美国所有癌症死亡人数的25%，因为它通常在一个月内被发现。 晚期，治疗选择有限。这导致了筛选制度 使用低剂量CT的计划，每年检测到约160万个肺结节， 其中大部分（~80%）位于肺的外周。 组织活检是建立明确诊断的标准治疗方法 规划在手术室对组织活检进行快速现场评估（ROSE）， 细胞病理学家或细胞技术人员可用于确定： 1.已获得诊断质量的组织;已通过ROSE鉴定恶性肿瘤， 显示可减少活检工具和重复手术的数量 2.已获得足够数量的细胞，以便在下一次- 世代测序和分子检测 3.淋巴结中没有转移性疾病证明了更具侵袭性的 外周活检 然而，尽管ROSE具有优势，但它不是护理标准，因为它的质量 由于各研究中心的差异很大，它可能会增加手术时间，而且其成本不太可能完全 报销。我们建议开发一种FDA批准的医疗设备， 显微镜成像的新鲜，未经处理的组织活检在治疗室，并提供 基于深度学习的精确、近实时诊断。 具体地，所提出的系统使用受激拉曼组织学（SRH），其 由PI开创，用于脑肿瘤的术中诊断，证明适用于 通过深度学习进行自动诊断，其性能不劣于 病理学家，并导致了第一个CE认证的设备，用于识别脑肿瘤的边缘 被公司。在这里，我们解决了肺癌的迫切临床需求， 对术中决策的信心，并减少不必要的活检/手术， 患者和付款人。 此外，我们将研究提供准确的术中诊断的可能性， 分子标记物可以使药物的局部递送、消融和其他 治疗，在活检过程中。

Project Summary & Abstract Lung cancer accounts for about 25% of all cancer deaths in the US, as it is often caught at an advanced stage when treatment options are limited. This has led to the institution of screening programs with low-dose CT, resulting in ~1.6 million pulmonary nodules detected every year, most of them (~80%) in the periphery of the lung. Tissue biopsy is the standard of care for establishing a definitive diagnosis for treatment planning. Rapid on-site evaluation (ROSE) of tissue biopsies in the procedure room by cytopathologists or cytotechnicians can be used to establish that: 1. Diagnostic quality tissue has been procured; identifying malignancy by ROSE has been shown to reduce the number of biopsy tools and repeat procedures 2. An adequate number of cells have been obtained to allow molecular profiling by next- generation sequencing and molecular testing 3. The absence of metastatic disease in the lymph nodes justifies the more-invasive peripheral biopsy. Nevertheless, ROSE is not the standard of care despite its advantages, because its quality is highly variable across sites, it can increase procedure times, and its costs are unlikely to be fully reimbursed. We propose the development of an FDA-cleared medical device that allows microscopic imaging of fresh, unprocessed tissue biopsies in the treatment room and provides accurate, near real-time diagnosis based on deep learning. Specifically, the proposed system uses Stimulated Raman Histology (SRH), which was pioneered by the PI, translated for intraoperative diagnosis in brain tumors, shown to be suitable for automated diagnosis via deep learning with a performance that is non-inferior to pathologists, and resulted in the first CE-certified device for identification of brain tumor margins by the company. Here, we address an urgent clinical need in lung cancer, giving doctors confidence in their intraoperative decisions, and reducing unnecessary biopsies/procedures for patients and payers. Further, we will investigate the potential to provide accurate intraoperative diagnosis of molecular markers that could enable local delivery of pharmaceuticals, ablation, and other therapies, in the biopsy procedure.