I noticed that the Challenge website mentions the use of 3D k-t Uniform sampling patterns, but the training set only appears to include 2D Uniform sampling. Could you please clarify which sampling pattern will be used for the validation set?
Created by Puyang Wang puyangwang Yes, the uniform sampling mask we provide in MaskTaskAll is 2D. Specifically, we offer 2D uniform, ktgaussian, and ktradial sampling trajectories. During the validation and testing phase, we will only use these three patterns. I'm referring to the masks provided in Mask_TaskAll folder. Example path: ChallengeData/MultiCoil/Cine/TrainingSet/Mask_TaskAll/Center001/UIH_30T_umr780/P001
Will exactly these sampling patterns be applied to validation data? Or Uniform sampling will be replaced by ktUniform? We only provide the training data as fully sampled data; you need to perform undersampling yourself for model training. The specific sampling code can be found at: https://github.com/CmrxRecon/CMRxRecon2025/tree/main/CMRxReconMaskGeneration.