CMA: Marker-assisted prevention and risk stratification (MAPRS): Artificial Intelligence Endoscopy for Colorectal Cancer Prevention (CMA1)

这一协作式绩效评估应用程序（CMA）旨在推进以下方面的精确管理： 癌症，特别是结直肠癌的标记物辅助预防和风险分层（MAPRS） 癌症（CRC）。CRC是美国第三大常见癌症，占所有癌症的近10%。 退伍军人中的癌症MAPRS源于一组来自VA结肠直肠的研究人员 癌症细胞基因组学协作（VA 4C），在VA现场会议的支持下创建 奖VA 4C旨在推进预防，早期发现， CRCs的诊断、预后和治疗。拟议的CMA旨在打破这些限制 显著推进CRC预防、检测、风险分层和精准治疗， 推进MAPRS。MAPRS-CMA旨在：CMA 1）开发人工智能增强型 结肠直肠癌预防内镜检查; CMA 2）检查基于粘蛋白的标记物，以改善 内镜下肿瘤性息肉的检测、切除、组织学分类和监测; CMA 3）验证源自功能性生物标志物的基于组织和血液的组合生物标志物组 特定途径研究，以改善风险分层;和CMA 4）检查 用于精确CRC治疗的细胞基因组药物反应谱分析。 我们的项目CMA 1的主要目标是在VA内创建和建立基础设施 使我们能够开发、验证和部署机器学习（ML）/人工智能（AI） 增强内窥镜检查的模型。在过去的十年里，生物光子技术出现了爆炸式的发展 以更精确地诊断和治疗结肠肿瘤。然而，结果是， 信息密集的成像，以解释和互动过程中。毫不奇怪， 实践的技术改进仍然局限于学术中心的专家。我们 假设任何操作者都可以通过计算机进行可靠实时息肉组织学检查- 使用ML/AI进行辅助诊断。这种能力最终将为广泛采用打开大门 节省成本，ASGE批准的切除和丢弃和遗留模式， 息肉因此，该项目的具体目标是：目标1：创建一个大型的，可扩展的标签 用于ML/AI研究的内窥镜数据库，包括从多个 退伍军人中心。目标2：利用这个图像存储库开发和验证ML/AI模型， 实现息肉的实时组织学以及目标3：开发用于计算机辅助的ML模型 息肉检测结合基于粘蛋白的荧光生物标记物用于宽视野检测。目的 4：使用ML/AI来帮助预测CRC药物反应，基于组合的临床因素， 细胞基因组数据。

This collaborative merit review application (CMA) aims to advance the precision management of cancers, specifically marker-assisted prevention and risk stratification (MAPRS) of colorectal cancers (CRCs). The third most common cancer in the USA, CRC accounts for nearly 10% of all cancers among Veterans. MAPRS stems from a group of investigators from the VA Colorectal Cancer Cellgenomics Collaborative (VA4C), created with the support of a VA Field-based Meeting Award. The VA4C aims to advance basic/translational research on the prevention, early detection, diagnosis, prognosis and treatment of CRCs. The proposed CMAs aim to disrupt these limitations and significantly advance CRC prevention, detection, risk stratification and precision treatment by advancing MAPRS. MAPRS-CMA aims to: CMA1) develop artificial intelligence-enhanced endoscopy for colorectal cancer prevention; CMA2) examine mucin-based markers to improve endoscopic detection, resection, histological classification and surveillance of neoplastic polyps; CMA3) validate tissue and blood-based combinatorial biomarker panels derived from functional pathway-specific studies to improve risk stratification; and CMA4) examine the potential of cellgenomic drug-response profiling for precision CRC treatment. The main objective of our project, CMA1, is to create and establish within the VA an infrastructure to enable us to develop, validate, and deploy machine learning (ML) /artificial intelligence (AI) models to enhance endoscopy. The past decade has seen an explosion in biophotonic technologies to more precisely diagnose and treat colonic neoplasia. The result is, however, increasingly information-dense imaging to interpret and interact with during procedures. Not surprisingly, technological enhancement of practice has remained restricted to experts at academic centers. Our hypothesis is that reliable real-time polyp histology can be enabled for any operator by computer- assisted diagnosis using ML/AI. This capability would finally open the door to widespread adoption of cost-saving, ASGE-sanctioned resect-and-discard and leave-behind paradigms for diminutive polyps. Thus, the specific aims of this project are: Aim 1: To create a large, scalable labeled endoscopic databank for ML/AI research comprised of clinical image data uploaded from multiple VA centers. Aim 2: To utilize this image repository to develop and validate ML/AI models that enable real-time histology of polyps as well as Aim 3: To develop ML models for computer assisted polyp detection in conjunction with mucin-based fluorescent biomarkers for widefield detection. Aim 4: Use ML/AI to help predict CRC drug response based on combined clinical factors and cellgenomic data.