cai2r / fastmri_prostate Goto Github PK

Hi,
first of all thank you for extending this great project even more by adding the classification code.

I wonder if you could also give the resulting scores of your training pipeline, just so people can compare other approches quickly, without having to run all your preprocess steps.

Thank you in advance!

Hi,

Thanks for the great work.

For T2 reconstruction, I found the code using the rss-then-avg pipeline, so 3 k-space data correspond to the final reconstruction. This is a little different from the fastMRI knee and brain dataset and will make a difference in training new machine learning tools.

To make it consistent for the other dataset and convenient for training, I tried to use avg-then-rss, which can result in a one-to-one correspondence between k-space and reconstruction. However, I found in this way, the image looks smoother.

What's your opinion on the difference between the two pipelines? Do you have any suggestions? Thank you.

Hi,

I'm seeking clarification on the process of determining the 'max' and 'norm' values for NYU patients. These values are crucial for intensity scaling during data analysis. Understanding their origins is pivotal to ensure accurate results.

We've attempted to calculate these values independently, but they don't match the values present in the h5 attributes.

While our code reads the 'max' and 'norm' values from the h5 file, it currently doesn't use them. However, an RIM reconstruction model relies on these values for intensity scaling.

This code reads in the max and norm value, but is not used further. However, an RIM reconstruction model uses these values for intensity scaling.
with h5py.File(fname, "r") as hf:
kspace = hf["kspace"][:]
calibration_data = hf["calibration_data"][:]
hdr = hf["ismrmrd_header"][()]
im_recon = hf["reconstruction_rss"][:]
atts = dict()
atts['max'] = hf.attrs['max']
atts['norm'] = hf.attrs['norm']
atts['patient_id'] = hf.attrs['patient_id']
atts['acquisition'] = hf.attrs['acquisition']

Your insights into this matter are greatly appreciated. Looking forward to your response.

Hello,

I could only find the volume level annotations in this dataset. If anybody can help, it would be awesome.

The repository contains many great resources, but it seems to be currently missing a link to the arXiv paper at https://arxiv.org/abs/2304.09254 in the README.