search

sudohainguyen / STAN-small-tumor-aware-network

Unofficial implementation of STAN paper published at ISBI 2020 by researchers from University of Idaho using Tensorflow Keras 2.0.
MIT License
12 stars 1 forks source link

Questions about evaluation indicators #1

Closed ykeivn closed 3 years ago

ykeivn commented 3 years ago

Thank you for sharing the code, I have some questions about the calculation of FPR in your paper. According to the formula of FPR, the result is between 0 and 1. But why your result appears to be greater than 1. Please tell me how you defined and calculated it. Thanks!

sudohainguyen commented 3 years ago

hey thanks for reaching out, sorry but I'm not the author of the paper, this is just my implementation that follows their declaration.

bryarshareef commented 3 years ago

Hello @ykeivn, The FPR is calculated as the ratio between the number of false positives and actual positives, which is different than the commonly used formulation as the ratio between the number of false positives and actual negatives. The false positive rate is calculated as FPR= |χg ∪ χ(p )-〖 χ〗_g |/|χ_g | , where χ_g denotes the ground truth pixels of the tumor region, and χ_p is the predicted pixel sets. Based on this equation, if the size of the false positive regions is larger than the size of the actual positive regions, the FPR will be greater than 1.