ESTRO 2025 - Abstract Book

S3401

Physics - Machine learning models and clinical applications

ESTRO 2025

Results: The model card consists of 5 sections: 1) Model basic information; 2) Model technical specifications (i.e. architecture, software and hardware); 3) Training data, methodology, and information; 4) Evaluation data, methodology, and results (i.e. commissioning); and 5) Other considerations, including fields about ethical use, risk analysis, and monitoring. Fields can have two status: required or optional. To allow evaluations on different datasets, section 4 can be repeated multiple times with specific indexes. Sections 3 and 4 contain task-specific fields for the considered RT applications. Besides the Microsoft Word format (Fig.1), a Markdown (.md) prototype was implemented in HuggingFace [4], to easen model card filling with dropdown lists and standardised content.