ESTRO 2024 - Abstract Book

S5085

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

and the maximum possible score for each criterion were calculated. Furthermore, the frequency of reader disagreements was assessed for studies that provided data from multiple readers.

Results:

A total of n=91 systematic reviews were incorporated into this analysis (n=35 published prior to 2022). The overall average RQS was 9.16 (pre-2022, average RQS = 8.05; 2022-present, average RQS = 9.85). Detailed scoring information was accessible for 67/91 (73.6%) reviews, constituting a total of 1764 individual quality scores, and multiple reader data was available for 20/91 (22.0%) reviews. The resulting percentage max score, adherence, and disagreement for all 16 criteria are reported in Figure 1. Both feature reduction and discrimination statistics showed some of the highest average maximum score (70.0%/85.0%) and adherence (66.4%/88.5%), respectively. However, assessment of discrimination statistics appeared to have the highest frequency of disagreements between readers (28.7%). Furthermore, when calculating the average RQS from raw score data and comparing them with the reported RQS, out of the 67 reviews that could be assessed, 18 overestimated their RQS and 10 underestimated it. These were primarily caused by errors in applying the criteria and incorrectly adding each criterion when reporting the final RQS for reviewed papers.

Conclusion:

The RQS remains an invaluable tool to push the field towards clinical translation. However, additional educational resources for authors who wish to use it in their reviews, and for manuscript reviewers and editors, may be necessary moving forward. Furthermore, some of the descriptions of the RQS criteria provided in the original paper by Lambin