The reason set the classification as a 2 class multi-label problem

nyukat / GMIC

An interpretable classifier for high-resolution breast cancer screening images utilizing weakly supervised localization

https://doi.org/10.1016/j.media.2020.101908

GNU Affero General Public License v3.0

168 stars 48 forks source link

The reason set the classification as a 2 class multi-label problem #20

Closed emma-sjwang closed 3 years ago

emma-sjwang commented 3 years ago

Hi, thank you very much for your sharing. I am newer to mammography classification. I have read several papers but some of them treat this problem as 2-class classification problem. May I know the reason why this paper utilizing multi-label to deal with this problem?

Thank you very much.

seyiqi commented 3 years ago

Hi @emma-sjwang,

We formulated the problem as a multi-label one because we extracted benign labels along with the malignant labels from the pathology reports in our private dataset. It's not a decision based on technical reasons.