Checklist for Artificial Intelligence in Medical Imaging (CLAIM): A Guide for Authors and Reviewers

T he advent of deep neural networks as a new artifi- cial intelligence (AI) technique has engendered a large number of medical applications, particularly in medical imaging.Such applications of AI must remain grounded in the fundamental tenets of science and scientific publication (1).Scientific results must be reproducible, and a scientific publication must describe the authors' work in sufficient detail to enable readers to determine the rigor, quality, and generalizability of the work, and potentially to reproduce the work's results.A number of valuable manuscript checklists have come into widespread use, including the Standards for Reporting of Diagnostic Accuracy Studies (STARD) (2-5), Strengthening the Reporting of Observational studies in Epidemiology (STROBE) (6), and Consolidated Standards of Reporting Trials (CONSORT) (7,8).A radiomics quality score has been proposed to assess the quality of radiomics studies (9).Peer-reviewed medical journals have an opportunity to connect innovations in AI to clinical practice through rigorous validation (10).Various guidelines for reporting evaluation of machine learning models have been proposed (11)(12)(13)(14).We have sought to codify these into a checklist in a format concordant with the EQUATOR Network guidelines (15,16) that also incorporates general manuscript review criteria (17,18).To aid authors and reviewers of AI manuscripts in medical imaging, we propose CLAIM, the Checklist for AI in Medical Imaging (see Table and downloadable Word document [supplement]).CLAIM is modeled after the STARD guideline and has been extended to address applications of AI in medical imaging that include classification, image reconstruction, text analysis, and workflow optimization.The elements described here should be viewed as a "best practice" to guide authors in presenting their research.The text below amplifies the checklist with greater detail.