Stability of Radiomic Models and Strategies to Enhance Reproducibility

Abstract

Radiomics is a progressive field aiming to quantitatively assess the diversity within and between tumors using image analysis. It holds tremendous promise for tracking tumor treatment progress over time. This review summarizes recent advances in ensuring the stability, repeatability, and reproducibility of radiomic analyses. It covers various factors influencing the radiomics process and potential variables that can affect stability. The study also proposes strategies to enhance the reliability of both radiomic features and models. Additionally, we highlight the importance of stability in each radiomic phase to achieve the cut-off stable model. Moreover, we discuss the details of using the radiomics quality score (RQS) to evaluate radiomics research, guiding researchers in formulating reasonable research designs to promote more stable radiomic models.