CADICA: a new dataset for coronary artery disease
收藏资源简介：
The CADICA dataset is an annotated Invasive Coronary Angiography (ICA) dataset of 42 patients. In ICA imaging, lesion degree assessment is commonly done by visual estimation, which implies a subjective factor and interobserver variability. Accurate recognition of lesions is crucial for a correct diagnosis and treatment. This motivates the development of computer-aided systems that can support specialists in their clinical procedures. This dataset can be used by clinicians to train their skills in angiographic assessment of CAD severity, by computer scientists to create computer-aided diagnostic systems to help in such assessment, and to validate existing methods for CAD detection approaching solutions for clinical settings. The CADICA dataset includes ICA images, manually labeled lesion bounding boxes, and selected clinical features. The CADICA dataset becomes a directory that contains the "metadata.xlsx" file, which is the file where the clinical data is located, as well as two main folders that differentiate the videos selected by the medical team for each patient: "nonselectedVideos" and "selectedVideos". Inside each folder, there are several sub-directories with the naming convention "pX", where X is the ID of each patient, and "vY", where Y is the ID of the video of that patient. The folder "pX" contains the following information: "vY": several sub-directories with the videos selected for that patient. "lesionVideos.txt": includes the IDs of the videos chosen where appears at least one lesion which is labeled. "nonlesionVideos.txt": contains the IDs of the selected videos with no visible lesions. The folder "vY" contains the following information: "input": a sub-directory containing a separate PNG file for each video frame. "pX_vY_selectedFrames.txt": includes the IDs of the keyframes for the medical team for all the selected videos. "groundtruth": a sub-directory available only if there are lesions in that selected video. The folder "groundtruth" contains the following information: "pX_vY_000ZZ.txt": contains the bounding boxes and their category in each row. There are such files as frames in "pX_vY_selectedFrames.txt". Bounding boxes are specified in the format [x,y,w,h], where (x,y) are the pixel coordinates of the top left corner, w is the width, and h is the height of the bounding box. "pX_vY_groundTruthTable.mat": contains a table with the ground truth information of that video. Inside the "selectedVideos" folder, you can find the "CADICAprojections.json" file, containing video projections. Please note that maybe the total videos does not match the total of selected videos, as manual discarding can be done.
CADICA数据集是一款包含42名患者的标注型侵入式冠状动脉造影（Invasive Coronary Angiography, ICA）数据集。在ICA成像流程中，病变程度评估通常依赖目视估算，该方式存在主观偏差与观察者间变异问题。病变的精准识别对正确诊断与治疗方案制定至关重要，这推动了可辅助临床医师开展诊疗工作的计算机辅助系统的研发。本数据集可供临床医师用于训练冠状动脉疾病（Coronary Artery Disease, CAD）严重程度的血管造影评估技能，也可供计算机科学家开发用于辅助此类评估的计算机辅助诊断系统，还可用于验证面向临床场景的冠状动脉疾病检测现有方法。 CADICA数据集包含ICA图像、人工标注的病变边界框（bounding boxes）以及精选临床特征。 CADICA数据集以目录结构组织，内含存储临床数据的"metadata.xlsx"文件，以及两个用于区分医疗团队为每位患者筛选视频的主文件夹："nonselectedVideos"（未选视频）与"selectedVideos"（已选视频）。每个文件夹内包含若干以"pX"命名的子目录（X为患者ID），子目录下另有以"vY"命名的子目录（Y为该患者对应视频的ID）。 "pX"文件夹包含以下内容： "vY"：对应该患者筛选视频的多个子目录； "lesionVideos.txt"：收录至少包含一处标注病变的视频ID； "nonlesionVideos.txt"：收录无可见病变的已选视频ID。 "vY"文件夹包含以下内容： "input"：一个子目录，其中存储对应视频每一帧的独立PNG文件； "pX_vY_selectedFrames.txt"：收录医疗团队为所有已选视频选定的关键帧ID； "groundtruth"：仅当该已选视频中存在病变时才会包含的子目录。 "groundtruth"文件夹包含以下内容： "pX_vY_000ZZ.txt"：每行包含一个边界框及其类别信息，文件数量与"pX_vY_selectedFrames.txt"中列出的帧数一致。边界框以[x,y,w,h]格式指定，其中(x,y)为左上角顶点的像素坐标，w为边界框宽度，h为边界框高度； "pX_vY_groundTruthTable.mat"：包含该视频的真值信息表格。 在"selectedVideos"文件夹中，可找到"CADICAprojections.json"文件，其中存储了视频投影相关信息。请注意，总视频数与已选视频数可能存在差异，因为可能会进行人工筛选剔除。
