AJNR Study-Specific Guidelines for AI in Medical Imaging: Bridging Gaps in Reference Standard and Clinical Evaluation.
Authors
Affiliations (2)
Affiliations (2)
- From the Department of Radiology (J.M.), Johns Hopkins School of Medicine, Baltimore, MD; Department of Radiology (H.V.V., N.P., G.Z.), Stanford University School of Medicine, Stanford, CA; Department of Radiology (S.P.), Columbia University Medical Center, New York, NY; Center for Intelligent Imaging (ci2) (A.M.R.), Department of Radiology & Biomedical Imaging, University of California, San Francisco (UCSF), San Francisco, CA; Neuroradiology Section, Mallinckrodt Institute of Radiology (G.J.GP.-C.), Washington University School of Medicine, St. Louis, MO; Department of Neuroradiology (M.W.), Division of Diagnostic Imaging, The University of Texas MD Anderson Cancer Center, Houston, TX and AdventHealth & AdventHealth Medical Group Central Florida Division (R.F.), Maitland, FL.
- From the Department of Radiology (J.M.), Johns Hopkins School of Medicine, Baltimore, MD; Department of Radiology (H.V.V., N.P., G.Z.), Stanford University School of Medicine, Stanford, CA; Department of Radiology (S.P.), Columbia University Medical Center, New York, NY; Center for Intelligent Imaging (ci2) (A.M.R.), Department of Radiology & Biomedical Imaging, University of California, San Francisco (UCSF), San Francisco, CA; Neuroradiology Section, Mallinckrodt Institute of Radiology (G.J.GP.-C.), Washington University School of Medicine, St. Louis, MO; Department of Neuroradiology (M.W.), Division of Diagnostic Imaging, The University of Texas MD Anderson Cancer Center, Houston, TX and AdventHealth & AdventHealth Medical Group Central Florida Division (R.F.), Maitland, FL [email protected] [email protected].
Abstract
The rapid proliferation of artificial intelligence (AI) applications in neuroradiology can lead to heterogeneous study design and reporting that impede reproducibility and clinical translation. To support authors submitting AI imaging research to AJNR, we introduce six study-type-specific reporting checklists for studies focused on: (1) Classification and Clinical Outcome Prediction, (2) Lesion Detection and Segmentation, (3) Image Transformation (quality enhancement or cross-modality synthesis), (4) Clinical AI Tool Evaluation, (5) Radiology Workflow Optimization, and (6) Technical Developments/Miscellaneous. Drawing on collective experience from manuscript reviews and real-world deployments, we have identified common pitfalls and gaps that these checklists address. In this technical note, we summarize the key elements of each checklist. This initiative complements existing AI reporting guidelines, emphasizing rigorous reference standards and clinical relevance to bridge the gap between algorithm development and radiology practice.