GEP-NETs radiomics in action: a systematical review of applications and quality assessment

Abstract

Purpose

To provide a comprehensive overview of the applications and quality of radiomics studies in GEP-NETs.

Methods

Embase, Scopus, and PubMed were searched until 2023. Studies that extracted qualitative radiomics features of GEP-NETs were included. Radiomics quality score (RQS) was used to assess the quality of studies. Changes in study quality were analyzed by grouping studies into three categories based on the year of publication. Correlation of impact factor (IF), CiteScore, Scientific Journal Rankings (SJR) and RQS were tested by spearman correlation analysis.

Results

A total of 64 studies were included, focusing on aggressive behavior prediction in tumors (n = 34), differentiation of GEP-NETs from other lesions (n = 18), and prognosis or treatment response prediction (n = 13). Three RQS criteria met most frequently in studies were discrimination statistics, discussing clinical utility and well-documented image protocol. The three RQS criteria met least frequently were prospective design, multiple imaging time points, open data. As time progressed, the 2022–2023 group achieved significantly higher RQS scores compared to the previous groups. IF and RQS (r = 0.29, p = 0.024), CiteScore and RQS (r = 0.22, p = 0.085), SJR and RQS (r = 0.28, p = 0.028) were all weakly associated.

Conclusion

Few studies focused on prognosis or treatment response prediction, indicating potential for future research. While overall improvements have been made, the majority of studies still exhibit low quality. Optimizing dataset quality, model assessment, and reporting of the radiomics workflow remains necessary. The three commonly used journal evaluation metrics may not accurately reflect the quality of a radiomics study.