What are the Quality Assessment Standards used in Artificial Intelligence Diagnostic Accuracy Systematic Reviews?

Abstract

Data was extracted onto a standardised proforma by two independent reviewers (VS and SJ). Study characteristics extracted were study author, year, institution, country, journal and journal impact factor. Data was collected on use of QUADAS and/or other quality assessment tools, quality assessment tool adherence, modifications to pre-existing tools, use of multiple tools to improve applicability to AI specific studies and any limitations pertaining to quality assessment expressed by study authors.Studies were classified into four clinical categories based upon the type of sample evaluated and upon the diagnostic task: (a) axial medical imaging, (b) non-axial medical imaging, (c) histopathological digital records (digital pathology) and (d) photographic images. Key AI-related extraction items were identified through examination of recently developed AI extensions to existing quality assessment tools. A consensus was reached amongst authors to ascertain vital items. Quality AssessmentThe AMSTAR 2 (A MeaSurement Tool to Assess systematic Reviews) was employed to evaluate the quality of included studies (Supplementary Table 1). 9 LowHigh Unclear