Item 42 — Provide a reference to the full study protocol or to additional technical details
State where readers can access the full study protocol or additional technical details if this description exceeds the journal’s word limit. For clinical trials, include reference to the study protocol text referenced in item 34. For experimental or preclinical studies, include reference to details of the AI methodology, if not fully documented in the manuscript or supplemental material. This information can help readers evaluate the validity of the study and can help researchers who want to replicate the study.
Explanation and Elaboration: Comprehensive, detailed description of methodology sufficient to enable reproduction of the work is the foundation of science. Unfortunately, the amount of text required to fully relate AI methods and complex clinical trials are often in conflict with the manuscript length limits enforced by journals. In the frequent case where it is not possible to fit all the details of the methods into the main article, this information should be provided through other channels and referenced in the article. Clinical trial studies are best referenced to their registries. Source code for model training may be shared through reference to repositories, such as services enabling version control. Information provided through reference to external sources like these should be expected to remain available for at least 5–10 years; in general personal or lab websites do not mean this durability requirement. Other methodological material may be included as appendices or supplements to the manuscript. Through completely describing their methods authors provide a roadmap for others to fully understand, replicate, and build upon their work, fostering credibility and advancement in the field.
| Examples |
| The study is registered at www.chictr.org.cn (registration no. ChiCTR2300069832). [Material and methods (21), CC BY 4.0] |
| The RadImageNet pretrained models and codes can be accessed at https://github.com/BMEII-AI/RadImageNet. [Material and methods (5), CC BY 4.0] |
| Specific details on the model architecture are described in Appendix E4 (supplement). [Material and methods (81), CC BY 4.0] |