Reproducibility dataset for:
GAN-based bone suppression using a combined loss function (2026)

This record contains the full reproducibility package associated with the accepted publication:

Jochymek L., Vašinková M., Doleží V., Gajdoš P.
GAN-based bone suppression using a combined loss function.
(2026)

The archive includes:

• Jupyter notebooks implementing GAN, Autoencoder, and U-Net models

• Configuration settings corresponding to the published experiments

• Trained model weights for best-performing configurations

• Dependency specification (requirements.txt)

• Citation metadata and licensing information

Dataset Information:

The experiments were conducted using the publicly available JSRT chest radiograph dataset:

Japanese Society of Radiological Technology (JSRT)
Standard Digital Image Database
http://db.jsrt.or.jp/eng.php

The original JSRT dataset is not redistributed in this archive due to licensing restrictions.
Users must obtain the dataset directly from the official source.

Purpose:

This reproducibility package ensures transparency, methodological validation, and long-term archival of the experimental configuration reported in the paper.

Technical Environment:

Python 3.6.8
TensorFlow 2.6.2
Segmentation Models v1.0.1
CUDA 11.4

The package enables full replication of the results reported in Tables 1–8 of the publication.

Funding:

This work was supported by:

• Center for Artificial Intelligence and Quantum Computing in System Brain Research (CLARA), Grant No. 101136607
• Research Platform for Digital Transformation and Society 5.0, Grant No. CZ.02.01.01/00/23 021/0012599

Related publication:

Jochymek L., Vašinková M., Doleží V., Gajdoš P. (2026).
GAN-based bone suppression using a combined loss function.
Frontiers in Artificial Intelligence, 9, 1761336.
https://doi.org/10.3389/frai.2026.1761336

This record corresponds to the final published version of the article. Earlier versions may reflect preprint submissions.