Skip to content

ZongyuLi-umich/DblurDoseNet

Repository files navigation

DblurDoseNet

https://github.com/ZongyuLi-umich/DblurDoseNet

PyTorch implementation of DblurDoseNet method for SPECT dosimetry, as described in the paper "DblurDoseNet: A deep residual learning network for voxel radionuclide dosimetry compensating for SPECT imaging resolution", by Zongyu Li, Jeffrey A Fessler, Justin K Mikell, Scott J Wilderman, Yuni K Dewaraja; to appear in Medical Physics, 2021. DOI: 10.1002/mp.15397.

Dataset

https://doi.org/10.7302/ykz6-cn05

Training

python3 train.py --batch [batch size] --lr [learning rate] --epochs [# of epochs]
For example, with batch size set to 32, learning rate set to 0.002, epochs number set to 200, the training command is
python3 train.py --batch 32 --lr 0.002 --epochs 200

Testing

To use the best checkpoint (the ckpt having the lowest validation loss), run the following command:
python3 test.py --batch [batch size] --is_best

About

PyTorch implementation of DblurDoseNet

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages