Catalyst Project: Quantification of immunohistochemistry images of neuroglia

Catalyst项目为历史上的黑人学院和大学（HBCU）提供支持，以建立教师的研究能力，以加强科学、技术、工程和数学（STEM）本科教育和研究。预计该奖项将进一步提高教师的研究能力，提高学院的研究和教学水平，并让本科生参与研究经验。该奖项授予霍华德大学，支持使用人工智能和机器学习来开发改进的技术，以实现对神经胶质细胞的无偏、高通量形态学检测和定量。预计所提出的模型将克服现有方法的局限性，新的深度学习系统将揭示神经胶质细胞高分辨率分割中以前未表征但基本的局限性。提出了一种新的深度学习系统，以解决脑病理学中量化区域特异性形态学和细胞亚类的困难。该提案旨在表征神经胶质细胞内的标志性特征，包括一些和扩展的神经胶质过程，这是已知的神经胶质细胞的病理决定因素。该方法采用复合深度学习系统和卷积神经网络进行细胞检测，然后使用专用分割分类器进行单细胞分割，以识别背景中具有区域特异性特征的异质胶质细胞。形态学参数将被确定并用于预测基于最广泛使用的2D， 20X免疫组织化学图像的神经胶质细胞的激活表型。为了提高模型的性能，将开发一个用户友好的网络工具箱，用于快速数据管理，并将其集成到一个综合的注释细胞形态和表型数据库中，用于模型训练和测试。提出的模型将与病理专家整理的黄金标准手工数据和其他现有的计算机辅助方法进行比较，包括基于规则的半自动方法和基于深度学习的方法，以控制有效性、一致性和计算效率。自动化分析、功能丰富的可视化、数据库集成和易于社区访问的开源分发，将使该系统成为神经胶质图像分析的常规工作流程。这项新技术将通过提供定量组织学的全自动工具而使神经科学研究界受益。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Catalyst Projects provide support for Historically Black Colleges and Universities (HBCU) to work towards establishing research capacity of faculty to strengthen science, technology, engineering, and mathematics (STEM) undergraduate education and research. It is expected that the award will further the faculty member's research capability, improve research and teaching at the institution, and involve undergraduate students in research experiences. This award to Howard University supports the use of artificial intelligence and machine learning to develop improved technology for the unbiased, high throughput morphological detection and quantification of glial cells. It is anticipated that the proposed model will overcome the limitations existing methodologies, and the new deep learning system will reveal previously uncharacterized yet fundamental limitations in high-resolution segmentation of neuroglia cells.Development of a novel deep learning system is proposed to resolve the difficulties in quantifying region-specific morphology and the cell subcategories in brain pathology. The proposal aims to characterize the hallmark features within glial cells, include the some and extended glial processes, which are known pathological determinants of glial cells. The approach involves a composite deep learning system with a convolutional neural network for cell detection, followed by a dedicated segmentation classifier for single cell segmentation to identify the heterogeneous glial cells with appearance of region-specific features in the background. Morphological parameters will be determined and used to predict the activation phenotypes of glial cells based on the most widely used 2D, 20X immunohistochemistry images. To enhance the model performance, a user-friendly web toolbox will be developed for fast data curation and integrated into a comprehensive database of the annotated cell morphology and phenotypes for model training and testing. The proposed model will be compared to the gold-standard manual data curated by pathology experts, and other existing computer-aided methodologies, including the rule-based semi-automatic methods, and deep learning-based methods to control for effectiveness, consistency, and computational efficiency. The automated analysis, feature-rich visualization, database integration and open-source distribution with easy access from the community, will allow this system to become a routine workflow for glial image analysis. This new technological will benefit the neuroscience research community by providing a fully automated tool for quantitative histology.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.