Informatics Tools To Analyze And Model Whole Slide Image Data At The Single Cell Level

IMAT-ITCR协作：开发基于深度学习的方法，以识别循环性疾病的亚型 光学显微镜图像中的肿瘤细胞 项目总结/摘要 母IMAT项目（R21 CA 240185）的目标是开发一个新的分馏和分析平台 的CTC亚群，并阐明CTC的转移潜力。目前，这项工作需要研究人员 为了记录芯片上捕获的细胞的数百个单独的显微镜图像， 与流动流体模拟，并分析三个特征的捕获细胞（包括角位置，归一化 速度和剪切）用于鉴定CTC亚型。这个过程非常劳动密集和耗时， 因为大多数步骤依赖于手动操作。ITCR项目（1U 01 CA 249245）的目标是开发一个 信息学平台iSEE-Cell（基于图像的细胞空间模式探索器），它具有一套 在单细胞水平上，用于组织图像分析、可视化、探索和空间建模的信息学工具。 这一拟议的行政补充申请，以支持IMAT和ITCR之间的合作- 资助的项目旨在开发基于深度学习的方法，从光学图像中识别CTC的亚型。 显微镜图像。这一提议的基本原理是，深度学习方法的发展 将提供在HU结构的微芯片上捕获的CTC的自动表征和分类。这 拟议的合作项目将利用这两个项目开发的技术，这将带来 加强互补技术平台和方法的能力， 癌症研究。本研究的创新点包括：1）多亚型的鉴定 使用它们在HU微芯片上的位置信息，而不进行破坏性免疫染色分析; 2） 新的Restore-GAN模型，以提高CTC捕获实验中获得的显微镜图像的质量， 提高CTC亚型预测的准确性; 3）拟议的信息学工具将提供计算机- 辅助自动化工具，使CTC的研究与人工智能。具体目标包括：目标1： 使用显微镜图像和分析/预测结果（来自IMAT项目）作为数据输入，以测试 从肿瘤组织图像中分类不同类型细胞的算法（iSEE-Cell，在卢旺达问题国际法庭项目中开发） 目标2：应用新的计算方法（Restore-GAN，开发于 卢旺达问题国际法庭项目），以提高从IMAT项目获得的图像的图像质量，并测试它们是否 目标3：开发一个用户友好的界面， iSEE-Cell平台用于远程分析光学/荧光显微镜图像。能够自动 从显微镜图像中捕获的细胞中提取/分析信息是迫切需要的， 提高IMAT项目的产量和工作效率。

IMAT-ITCR Collaboration: Develop deep learning-based methods to identify subtypes of circulating tumor cells from optical microscope images Project Summary/Abstract The goal of the parent IMAT project (R21CA240185) is to develop a new platform for fractionation and profiling of CTC subpopulations and elucidate the metastatic potential of CTCs. Currently, this work requires researchers to record hundreds of individual microscope images of the cells captured on the microchip, integrate all images with flow fluid simulations, and analyze three features of the capture cells (including angular position, normalized velocity and shear) for identification of CTC subtypes. This process is very labor-intensive and time-consuming, as most of the steps rely on manual operations. The goal of the ITCR project (1U01CA249245) is to develop an informatics platform, iSEE-Cell (image-based Spatial pattern ExplorEr for Cells), which features a suite of informatics tools for tissue image analysis, visualization, exploration and spatial modeling at the single-cell level. This proposed Administrative Supplement application in support of collaboration between IMAT and ITCR- funded projects aims to develop deep learning-based methods to identify subtypes of CTCs from optical microscope images. The rationale underlying this proposal is that the development of deep learning methods will provide automatic characterization and classification of CTC captured on HU structured microchips. This proposed collaborative project will leverage the technologies developed by both projects, which will bring together and enhance the capabilities of complementary technology platforms and methodologies to advance cancer research. Innovation of the proposed methods include the following: 1) Identification of multiple subtypes of CTCs using their location information on an HU microchip without destructive immunostaining analysis; 2) Novel Restore-GAN model to improve quality of microscope image obtained in CTC capture experiments and enhance predication accuracy for CTC subtypes; 3) The proposed informatics tools will provide computer- assisted automated tools to empower CTC research with artificial intelligence. Specific aims include: Aim 1: Using the microscope images and analysis/prediction results (from the IMAT project) as data input to test whether algorithms to classify different types of cell from tumor tissue images (iSEE-Cell, developed in the ICTR project) can be applied for microscope images; Aim 2: Apply novel computational methods (Restore-GAN, developed in the ICTR project) to improve image quality of the images obtained from the IMAT project, and test whether they can improve prediction accuracy for CTC subtypes; Aim 3: Develop a user-friendly interface to incorporate the iSEE-Cell platform for analyzing optical/fluorescent microscope images remotely. The ability to automatically extract/analyze information from captured cells in the microscope images is urgently needed and will dramatically enhance the throughput and work efficiency of the IMAT project.