Hello, Thanks for your great effort in maintaining this challenge and the dataset. I have an inquiry regarding image quality: I did GRAPPA on the whole dataset, I also modified the code to construct the entire k-space, but unfortunately, I can see many images have different kinds of artifacts or squished images. I am attaching several examples, is this normal? P001_sax, MultiCoil, training, full sample, GRAPPA: https://drive.google.com/file/d/1R1KmstfqbitK1JEiqDVxMelLhy43Lsxw/view?usp=sharing P038_sax, MultiCoil, training, full sample, GRAPPA: https://drive.google.com/file/d/1coyz6RRlNQaFetK0EMHohJOCYM2y-ruY/view?usp=sharing

Created by Mohamed Elbayumi MohamedElbayumi
Hello Fanwen, Thank you for the response, What do you mean by groundtruth data? are you providing the reconstructed images, Because the images that I show are from the fullsample folder constructed using the GRAPPA code (no sampling factor applied to these sample images) The artifacts are basically wrap around in most of time, but I am concerned more about the quality, is there a possibility to provide P001_sax and P038_sax groundtruth images that you have, so I can compare the quality? Best, Mohamed
Hello, Thank you for asking. From the jpg (seemed like a coil-combined image), you provided, it's a little unclear what specific artefacts are. Please refer to the groundtruth data we provided in the training dataset and compare. Besides, if the undersampling factor is larger than 4, it's quite common to see the ringing artefacts in the image reconstructed using GRAPPA. For the standard GRAPPA recon code, please refer to https://github.com/CmrxRecon/CMRxRecon2024/blob/main/CMRxReconDemo/mainGRAPPA.m. Hope this can help you. Best wishes, Fanwen CMRxRecon Team

Constructed Images quality concerns page is loading…