h5 and DICOM relationship

[AlbertoUAH]

Hi, is there any relationship between .h5 and .dcm files available in fastMRI dataset? I'd like to get DICOMs with all planes instead of Coronal plane, but I don't find any relationship between .h5 and .dcm names.

Best regards.
Alberto.

[speechle55]

hi, i meet the same question, is there any relationship between h5 and dicom? i don't know how to relate 'file1000001' to dicom info

[Solor-pikachu]

Hi, i meet the same question, is there any relationship between h5 and dicom?

[So-cean]

+1

[wehrlimi]

Hello,

I am reviewing the knee CSV file (knee_file_list.csv) and have the same question raised by others, including @speechle55. Specifically, how can the filenames in the CSV be mapped to the corresponding FastMRI DICOM data or directory structure?

At the moment, it is unclear how an entry such as file1000001 relates to any of the FastMRI DICOM folders (e.g., FB_100030____FB,2817673489\study_0c1e55cc\MR3_33a7c6a2). I have not been able to locate the numeric identifiers (e.g., 1000001) in any folder names or DICOM headers. This makes me wonder whether these filenames may have been generated programmatically (e.g., sequentially within a loop).

If a mapping between the CSV filenames and the original FastMRI DICOM files could be provided, that would be extremely helpful and would allow the labels to be used effectively.

Best regards,
Michael

[So-cean]

@wehrlimi you can convert multicoil knee scan("/fastmri/knee/multicoil_train/file1000001.h5" ) into dicom first, then convert it into nii.gz. Also, the "knee_mri_clinical_seq" is not related to fastmri-plus annotations.

[wehrlimi]

this makes sense! thank you @So-cean