RSNA Knee Abnormality Detection is a knee joint medical imaging dataset curated by alenic. Preprocessed from the original RSNA Knee Abnormality Detection data, it converts DICOM images to JPEG format and resizes them to 224 × 224 pixels, aiming to provide standardized image data for computer vision tasks related to knee joint abnormality detection.

The original RSNA dataset includes knee MRI images and associated annotation information, covering various abnormalities such as ligament and meniscus injuries, osteoarthritis, joint effusion, synovitis, bone bruises, and fractures. After image format conversion and size standardization, this dataset is more suitable for deep learning model training, image classification, and medical image recognition experiments.

Dataset Composition

• Data Scale: This preprocessed version contains approximately 820K annotated images, with a total size of about 10.77 GB.
• Data Format: JPEG images accompanied by CSV metadata files.
• Task Objective: Identification and classification of abnormalities based on knee MRI images.
• Preprocessing Method: Reading DICOM images, converting them to uint8 format, and resizing to 224 × 224 pixels.
• Annotation Information: Data is organized by fields and labels provided in the accompanying CSV files; specific annotation content is subject to the files actually provided in this version.