New framework READII-2-ROQC identifies confounding factors in imaging AI models

A new open-source framework called READII-2-ROQC has been developed to address confounding factors in radiomics and imaging foundation models. This framework uses volume-preserving negative controls to evaluate whether extracted features truly represent tumor biology or are influenced by factors like tumor volume or acquisition artifacts. When applied to public cancer imaging cohorts, READII-2-ROQC demonstrated that some models maintain performance even after spatial structure is destroyed, indicating volume-driven confounding, while others show sensitivity to perturbations. AI

IMPACT This framework could improve the interpretability and reliability of imaging AI models, leading to more accurate diagnostic biomarkers.

RANK_REASON The cluster contains an academic paper detailing a new framework and methodology for evaluating AI models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• PyRadiomics
• READII-2-ROQC

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →