Question

[Zanbo0]

Hello, I read your paper. That is excellent. But I have a question, how can we get the feature_map.mha and the mask.mha.
I would be very grateful if you reply.

[satishev]

Feature_map would be the output of computing a pixel-wise texture map from an input image.
Mask would be the annotation of the region of interest that you are analyzing (so RADISTAT is limited to within that region)

[Zanbo0]

I appreciate your reply very much. I am a CS student, a beginner in RADIOMIC. I have the .dcm and nii.gz, should I transform them to .mha？ And the feature_map and mask need my own calculations or they are generated with software.

[satishev]

You can try to extract feature_map using from the DCM/NII.GZ file using tools such as pyradiomics - the result of which should be a pixel-wise map that you can save as an MHA file. Mask is a file that specifies the ROI - usually defined by the user in some fashion (bounding box/lesion area etc.)

[Zanbo0]

Hello, after studying I have some questions again. Now I can transform mask.mha and feature_map.mha, and use them to generate the label_mask. As @jta35 said that there are incorrect arguments passed to slic() in radistat.py, and more issues will happen when I try to fix them. Such as generating the bar charts and computing the texture_vec, spatial_vec.

I don't know how to use the mask to filter out the feature_map, and the generation of bar charts has errors. Here is the error information.

ValueError: shape mismatch: objects cannot be broadcast to a single shape. Mismatch is between arg 0 with shape (3,) and arg 1 with shape (0,).

And I noticed that the shape of your data is (222,222), but my data's is (512,512), I don't know if this has any effect.