SemiDose

Semi-supervised learning for dose prediction in targeted radionuclide therapy: a synthetic data study.

👉 The paper has been accepted in Physics in Medicine & Biology.

👉 Code:

• requirements.py, necessary Python libs
• data_load.py.py, load dataset
• model_load.py, load model
• hyper_parameter.py, hyper-parameters
• train_regfixmatch.py, model training, validation and test process

⭐ Highlights:

• We develop an effective radiotherapy dose estimation method based on semi-supervised deep learning.
• We propose a new pseudo-label generation algorithm in SSL that can better adapt to regression-based tasks.

Requirements

• Python 3.*
• Pytorch 1.x or 2.0

Citation

@article{Zhang_2026,
doi = {10.1088/1361-6560/ae36df},
url = {https://doi.org/10.1088/1361-6560/ae36df},
year = {2026},
month = {jan},
publisher = {IOP Publishing},
volume = {71},
number = {2},
pages = {025005},
author = {Zhang, Jing and Bousse, Alexandre and Pham, Chi-Hieu and Shi, Kuangyu and Bert, Julien},
title = {Semi-supervised learning for dose prediction in targeted radionuclide therapy: a synthetic data study},
journal = {Physics in Medicine & Biology},
}