•  
  •  
 

Abstract

Radiomics refers to the quantitative extraction of features from medical images, which could potentially be introduced into clinical practice for improvement in diagnostic, prognostic, and predictive accuracy. Issues with regard to repeatability, reproducibility, and robustness challenge the reliability of radiomic features and their potential utility in clinical practice. This is a systematic review of the statistical methods employed in the analysis of features and their reliability cross studies. The databases reviewed were PubMed and Web of Science, with sources from January 2015 to January 2024. Critical identified statistical methods include the intraclass correlation coefficient (ICC), concordance correlation coefficient (CCC), means and standard deviations, the coefficient of variation (COV), Friedman test, Spearman's rank correlation, t-tests, and k-means clustering. Each method's application and effectiveness in measuring the consistency and agreement of feature reliability was discussed. Key results of the review indicate that ICC and CCC methods will be necessary for feature reliability; on the other hand, feature categorization and validation come from clustering and correlations. The review emphasizes the need for highly reliable statistical methods to prove the efficacy of the radiomic features. Therefore, a standardized methodology needs to be followed, which may make it reproducible on a clinical scale.

First Page

62

Last Page

68

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Download

Share

COinS