Ten rules to reduce noise and increase precision in imaging studies in rheumatology: a methodological viewpoint...

Imaging studies are ubiquitous in rheumatology. Research in imaging has throughout the years mainly focused on improving the sensitivity of imaging techniques and scoring methods. Remarkably little attention has been paid to constraining unwarranted noise (acquisition, measurement, and analytical error) in spite of strict regulatory guidance about noise reduction for new drug approval. Noise reduction in imaging depends on the application of several methodological rules. Scientific credibility of imaging results should not depend on the complexity and incomprehensibility of the reported statistical model (plus eye-catching P values), but rather on optimal signal-to-noise ratio, usually meaning less 'noise' at a given 'signal'. This viewpoint deals with noise reduction. Repeated scoring of existing images, correction of inappropriate prior scoring, correction of fallible previous analytical procedures and more, can all contribute to a more trustworthy and better result. In the current world of patient-reported clinical information, imaging data still stand out as being rather 'objective'. But the influence of scoring and analysis in terms of unwarranted noise by readers and data analysts cannot be disregarded. This viewpoint proposes 10 expert-based rules for noise reduction at the level of image scoring, database handling, and data analysis. These rules should be considered not-yet-validated methodological hypotheses that require empirical validation through comparative methodological research.