search

nyukat / BIRADS_classifier

High-resolution breast cancer screening with multi-view deep convolutional neural networks
https://arxiv.org/abs/1703.07047
BSD 2-Clause "Simplified" License
147 stars 58 forks source link

Birads 3 Categories #13

Closed Emirbz closed 5 years ago

Emirbz commented 5 years ago

Hello , i'm using this repository for my school project 'Breast cancer classification using mammoram pictures' the project is working perfectly my question is why it shows only 3 birads categories while birads got 7 categories ? Thank you for reading my question 👍

kjgeras commented 5 years ago

Hi Amir,

This is because we are only considering the three BI-RADS used in screening mammography (0, 1, 2). In later stages of the diagnostic pipeline, a patient can be assigned other BI-RADS. What it means is that, if you want to use our models, you should do one of two things:

  1. Change your prediction task by mapping all BI-RADS categories into three BI-RADS categories. I suggest to follow the way we describe here: https://cs.nyu.edu/~kgeras/reports/datav1.0.pdf (i.e., 0->0, 1->1, 2->2, 3->2, 4a->0, 4b->0, 4c->0, 5->0).
  2. Drop the last layer of our network and put a new one with however many outputs you want to predict.

Please also check this repository: https://github.com/nyukat/breast_cancer_classifier It contains much stronger models than this one.

Emirbz commented 5 years ago

What does each Number stand for ? 0 probability of what ? same for 1 and 2 💃

kjgeras commented 5 years ago

These are different categories.

BI-RADS 0: "incomplete, needs additional imaging" BI-RADS 1: "normal" BI-RADS 2: "bening"

You can read more about these for example here: https://breast-cancer.ca/bi-rads/

Emirbz commented 5 years ago

thanks for your quick response So to have the others categories : 0- incomplete 1-negative 2-benign findings 3-probably benign 4-suspicious abnormality 5-highly suspicious of malignancy 6-known biopsy with proven malignancy i have to use the other project (breast_cancer_classifier)? because i need the malignancy probability

kjgeras commented 5 years ago

Not exactly. In order to explain this, I would basically need to write a few pages. Please read the website that I gave you a link to first, then this paper: https://arxiv.org/pdf/1703.07047.pdf and this paper https://arxiv.org/pdf/1903.08297.pdf and everything will be clear.