An AI-enabled Digital Pathology Platform for Multi-Cancer Diagnosis, Prognosis and Prediction of Therapeutic Benefit

摘要：认识到许多癌症的过度诊断正在导致辅助治疗的过度 随着化疗或放射治疗的加强，越来越多的人认识到需要对预后和 预测分析，以确定哪些癌症患者可以从非强化治疗中受益。而基于多基因表达的测试，如Oncotype DX和Decpher，则可用于识别早期乳腺癌和 前列腺癌患者可以避免辅助治疗，从而减少副作用和并发症， 这些检查的价格(3000-4K美元/患者)使它们超出了大多数中低收入患者的承受能力 国家(LMIC)。具有讽刺意味的是，在像这样的LMIC中，对这些预测和预测性测试的需求更加迫切 印度，那里获得放射和化疗等治疗资源的机会有限，因此需要 对那些将从中受益最多的患者进行明智的管理。 使用计算机视觉和模式识别工具进行复杂的数字病理分析已经成为 从苏木素和曙红中解锁关于肿瘤行为和患者预后的亚视属性 (H&E)-单独染色的幻灯片。凯斯西储大学(CWRU)的Madabhushi团队广泛地 显示了这些方法在预测结果和治疗反应方面的潜力 颈部、肺癌和前列腺癌。Madabhushi团队与合作者Parmar博士和Desai博士在 印度最大的癌症中心塔塔纪念中心(TMC)已经表明，高级病理分析 能够识别乳腺癌独特的预后形态特征，这些特征在南方不同 亚裔(SA)和高加索裔美国人(CA)女性1。此外，CWRU小组已经表明数字病原体 基于图像的分类器可以显著地改善甚至超过预测和预测 昂贵的基因表达分析在乳腺癌(肿瘤型Dx)和前列腺癌(解密)中的表现2。 建立在CWRU和TMC 3之间现有的强大合作和数字领域的良好记录基础上 基于图像的预测和基于预测的分析，我们建议优化和验证支持人工智能的数字 用于多种癌症诊断、预后和治疗效益预测的病理学平台(Adapt)。 ADAPT将涉及在SA的背景下优化CWRU小组先前开发的图像分析 癌症患者。此外，通过将AI-病理工具与广泛使用的数字工具Path Presenter集成在一起 病理图像分析平台，ADAPT将成为具有全球足迹的预测和预测工具。 具体地说，ADAPT将被验证用于预测辅助化疗和放射治疗的结果和益处 在雌激素受体阳性(ER)乳腺癌(BC)和三阴性乳腺癌(TNBC)的背景下， 口腔鳞状细胞癌(OC-SCC)和TMC的前列腺癌 在美国(SWOG S8814，RTOG 0920,0521)和TMC(REST，POP-RT)。成功完成项目 将把Adapt确立为印度癌症患者负担得起的精准医学(APM)解决方案。

SUMMARY: Recognizing that over-diagnosis of many cancers is leading to over-treatment with adjuvant chemotherapy or with radiation therapy boost, there is a growing appreciation for the need for prognostic and predictive assays to identify those cancer patients who can benefit from therapy de-intensification. While multi-gene-expression based tests such as Oncotype DX and Decipher exist for identifying early-stage breast and prostate cancer patients who could avoid adjuvant therapies and hence mitigate side-effects and complications, the price of these tests ($3K-4K/patient) puts them beyond the reach of most patients in low- and middle-income countries (LMICs). Ironically, the need for these prognostic and predictive tests is even more acute in LMICs like India, where access to treatment resources like radiation and chemotherapy are limited and hence need to be administered judiciously to those patients who stand to receive the most benefit from them. Sophisticated digital pathomic analysis with computer vision and pattern recognition tools has been shown to “unlock” sub-visual attributes about tumor behavior and patient outcomes from hematoxylin & eosin (H&E)-stained slides alone. The Madabhushi team at Case Western Reserve University (CWRU) has extensively shown the potential for these approaches for predicting outcome and therapeutic response for breast, head and neck, lung and prostate cancer. The Madabhushi team working with collaborators Dr. Parmar and Dr. Desai at the Tata Memorial Center (TMC), the largest cancer center in India, has shown that advanced pathomic analysis is able to identify unique prognostic morphologic signatures of breast cancer that are different between South Asian (SA) and Caucasian American (CA) women 1. In addition, the CWRU group has shown that digital pathomic based image classifiers can significantly improve and even outperform the prognostic and predictive performance of expensive gene-expression assays for breast (Oncotype Dx) and prostate cancer (Decipher) 2. Building on the strong extant collaboration between CWRU and TMC 3, and a strong track record in digital image based prognostic and predictive based assays, we propose to optimize and validate an AI-enabled Digital Pathology Platform (ADAPT) for Multi-Cancer Diagnosis, Prognosis and Prediction of Therapeutic Benefit. ADAPT will involve optimizing the previously developed image assays by the CWRU group in the context of SA cancer patients. Furthermore, by integrating the AI-pathomic tools with PathPresenter, a widely used digital pathology image analysis platform, ADAPT will have a global footprint for the prognostic and predictive tools. Specifically, ADAPT will be validated for predicting outcome and benefit of adjuvant chemo- and radiation therapy in the context of estrogen receptor positive (ER+) breast cancer (BC) and triple negative breast cancer (TNBC), oral cavity squamous cell carcinoma (OC-SCC) and prostate cancer at TMC via a number of clinical trial datasets in the US (SWOG S8814, RTOG 0920, 0521) and at TMC (AREST, POP-RT). Successful project completion will establish ADAPT as an Affordable Precision Medicine (APM) solution for Indian cancer patients.