Acquisition variables

What acquisition variables are available ? Specifically, can we have the scanner information for each patient ? In the public dataset , there is only the clinical information.

Hi Elisavet,

The following three acqusition variables are available for every case in the training, validation and testing datasets of the PI-CAI challenge:

• Scanner manufacturer (e.g. SIEMENS)
• Scanner model name (e.g. Skyra)
• Diffusion b-value (e.g. 1400)

For the training datasets, these values are available in the imaging metadata of each scan (via their standard DICOM attribute). Diffusion b-values are only available in the imaging metadata of high b-value sequences (i.e. files ending in _hbv.mha). For example:

import SimpleITK as sitk

t2w_img = sitk.ReadImage('10001_1000001_t2w.mha') 
scanner_manufacturer = t2w_img.GetMetaData("0008|0070")
scanner_model_name = t2w_img.GetMetaData("0008|1090")

hbv_img = sitk.ReadImage('10001_1000001_hbv.mha') 
diffusion_b_value = hbv_img.GetMetaData("0018|9087")

For the validation and testing cohorts, these values are available through the clinical-information-prostate-mri.json interface of grand-challenge.org. Please refer to Lines 104-107 of the process.py script of the baseline U-Net GC algorithm, for an example on how they can be extracted while running your Dockerized AI algorithm on grand-challenge.org (during validation/testing).

Hope this helps.

Thank you very much for your quick reply.