STAN - Small Tumor-aware Network

Unofficial implementation of STAN paper published at ISBI 2020 by authors from University of Idaho using Tensorflow 2.0.

STAN architecture

Dataset

We use the Dataset B in this implementation. Access permission needed, visit the Release Requirement for more details.

TODOs

• [x] Dataset B Generator
• [x] Model implementation
• [x] Training code using click
• [x] Example for Dataset B
• [x] Tversky loss function
• [x] Dice loss function
• [ ] Lovasz loss function
• [ ] Focal loss function
• [ ] Smarter training procedure with mlflow and hydra
• [ ] Inference code
• [ ] Evaluate code