Are we supposed to use Clinical Information Descriptors for training?

Hi,

I am trying to understand the dataset.

From my understanding, the input of our algorithm should be images, and output is the detection map+likelihood score. The Clinical Information Descriptors are provided in the label github repo. Are we supposed to incorporate Clinical Information as part of the input? if so, there is a gleason score in that table, does that correlated to the label?

Thanks, Chacha

Hi Chacha,

Clinical/acquisition information can be used as part of the input to your AI model. These parameters only include the ones mentioned under the "Clinical and Scanner Information" section in the "Datasets: Imaging, Labels" page. You cannot use the Gleason scores as input, because that information is part of the labels or annotations (i.e. they won't be available for validation/testing cases).

Hope this helps.

Hi Anindo,

Thank you so much for your prompt reply!

Just one quick clarificaition question. In your provided nnUnet and nnDetection baseline implementations, did you use those clinical informations?

Thanks, Chacha

Hi Chacha,

For the sake of simplicity, we haven't used that information in any of our baseline models.

Hope this helps.