RSNA Knee MRI is a knee MRI processing dataset for the RSNA Knee Abnormality Detection Challenge, designed to provide standardized input data for the training and evaluation of knee abnormality detection models. This dataset contains preprocessed 3D volumetric arrays of all 24,371 RSNA MRI sequences. All sequences have been compressed into a single .npz file, which can be directly loaded as a PyTorch tensor.