output labels

[anooyih]

In the mat file, the predicted tissue class in pclass_mat or cell type in nuclei_mat are stored as integer. I could not find how to map those integers to actual meaningful label such as epithelial, lymphocyte, plasma, etc., for cell type or adipose, abnormal epithelial, normal epithelial, etc., for tissue type. Could you provide the dictionary for each integer somewhere?

Thanks in advance.

[carloalbertobarbano]

Same problem here :-) Can you provide a way to map the values? thank you

[jw235]

Are you using the .mat or .dat file to visualize the results ? and do you use Qupath ?

[simongraham]

I believe it should be:

nuclei_dict = {
    0: 'nolabel',
    1: 'neutrophil',
    2: 'epithelial',
    3: 'lymphocyte', 
    4: 'plasma',
    5: 'eosinophil', 
    6: 'connective'
}

tissue_dict = {
    0: 'background',
    1: 'abnormal_epi',
    2: 'normal_epi',
    3: 'inflammatory', 
    4: 'muscle',
    5: 'stroma', 
    6: 'debris'
    7: 'mucin'
    8: 'adipose'
}

Please double check.

[Himanshunitrr]

You can also look at https://github.com/TissueImageAnalytics/cerberus/blob/master/dataset.yml