Apply different intensity transforms per channel

[jpcenteno80]

Hi, I am working with CT and would like to apply a different a_min, a_max using ScaleIntensityRange per channel. For example, load input image with shape 1,64,64,64 (channel,spatial_dim,spatial_dim,spatial_dim), apply RepeatChannel transform to end up with shape 2,64,64,64 and then apply ScaleIntensityRange with (a_min=-300,a_max=300) for one channel and (a_min=130,a_max=3000) for the other channel. I would like this to be part of a dictionary style transform chain. Is this possible with your current transforms? Thank you!

[jpcenteno80]

Found the solution using these transforms after loading the image/label and before creating a tensor inside a Compose object:

           ...
            RepeatChanneld('image', repeats=2),
            SplitChanneld('image', channel_dim=0),
            ScaleIntensityRanged('image_0', a_min=heart_range[0], a_max=heart_range[1],
                                b_min=0.0, b_max=1.0, clip=True),
            ScaleIntensityRanged('image_1', a_min=bone_range[0], a_max=bone_range[1],
                                b_min=0.0, b_max=1.0, clip=True),
            ConcatItemsd(['image_0', 'image_1'], name='image', dim=0),
            DeleteItemsd(['image_0', 'image_1']),
            ...

[Nic-Ma]

Hi @jpcenteno80 ,

Thanks for your interest here.
Actually, I would suggest you to use the CopyItemsd transform instead of RepeatChanneld transform.
After copying, you can apply ScaleIntensityranged on different keys and concat the results.

Thanks.