Microglial Ground Truth Dataset for StainAI

This dataset is utilized by a deep learning tool, StainAI, that converts immunohistochemistry images into quantitative maps for direct and high-throughput quantification of microglia. The development of StainAI involves a four-step process: (1) image pre-processing, (2) image curation and creation of a ground truth dataset, (3) development of the deep learning system, and (4) morphological mapping and analysis to evaluate microglia in low-magnification (20X) 2D IHC slides. The ground truth datasets include manually labeled microglia for training, testing, and validating the deep learning models for cell detection, segmentation, and classification.Guidelines for cell labeling procedure: StainAI_annotation guidelines.docxDatasheet for annotation guidelines: StainAI_AnnotationDatasheet.xlsxDataset for cell detection by YOLO/MaskR-CNN: train_yolo_maskRCNN.zipDataset for cell segmentation by UNet: train_Unet.zipDataset for cell classification by C50: train_C50.zip<br>

本数据集由深度学习工具StainAI使用，该工具可将免疫组织化学（immunohistochemistry, IHC）图像转换为定量图谱，实现小胶质细胞的直接、高通量定量分析。StainAI的开发流程分为四个步骤：(1) 图像预处理；(2) 图像遴选与真值数据集（ground truth dataset）构建；(3) 深度学习系统开发；(4) 形态学映射与分析，用于评估低放大倍率（20X）二维免疫组织化学切片中的小胶质细胞。该真值数据集包含经人工标注的小胶质细胞，用于训练、测试与验证用于细胞检测、分割与分类的深度学习模型。细胞标注流程指南：StainAI_annotation guidelines.docx；标注指南数据表：StainAI_AnnotationDatasheet.xlsx；用于YOLO/Mask R-CNN细胞检测的数据集：train_yolo_maskRCNN.zip；用于UNet细胞分割的数据集：train_Unet.zip；用于C50细胞分类的数据集：train_C50.zip。