New-Clear Insights on Throughput: A Systematic Review of Artificial Intelligence's Impact on Radiological Workflow Efficiency

Abstract The exponential growth of medical imaging has increased radiologists' workload, prompting the need for innovative workflow optimization strategies. Artificial intelligence (AI) has emerged as a promising tool to enhance efficiency, reduce turnaround times (TATs), and maintain diagnostic accuracy. This systematic review synthesizes current evidence on the impact of AI-integrated workflows in radiology on reporting efficiency, diagnostic accuracy, and operational performance. To systematically evaluate the effects of AI-assisted workflows compared with standard radiology practices on reporting TAT, radiologist workload, diagnostic accuracy, and workflow efficiency across imaging modalities. A systematic review was conducted according to PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines. PubMed, Embase, and Google Scholar were searched for studies published in English from January 2013 to August 2025 using combinations of “Artificial Intelligence,” “Deep Learning,” “Radiology,” “Workflow,” and “Efficiency.” Eligible studies included randomized controlled trials, cohort studies, and pre-/post-implementation analyses comparing AI-assisted and standard radiology workflows. Data were extracted on study design, AI integration method, workflow outcomes, and diagnostic performance. Risk of bias was assessed using the Cochrane RoB 2 and CASP checklists. Owing to methodological heterogeneity, a qualitative synthesis was performed. Included studies encompassed emergency and routine radiology workflows across computed tomography, magnetic resonance imaging, and X-ray modalities. AI integration reduced median reporting times by 20 to 36% and TATs by 12 to 59 minutes, particularly in emergent settings such as intracranial hemorrhage and pulmonary embolism detection. AI-assisted drafting and triage reduced cognitive load and saved approximately 1 hour of daily reading time. Studies reported improved lesion measurement concordance (66.4 vs. 54%) and fewer clinically significant errors (0.27 vs. 0.38 per case), though not always statistically significant. Sensitivity and specificity remained high (up to 95–97%), indicating that AI improved or maintained diagnostic integrity without compromising safety. Embedded AI systems yielded higher adoption and measurable efficiency gains, while standalone tools showed limited workflow benefit. AI integration in radiology workflows significantly enhances time efficiency, reduces radiologist workload, and maintains diagnostic accuracy. While evidence supports AI as a valuable decision support tool rather than a replacement for radiologists, further multicenter, real-world studies are warranted to evaluate long-term clinical and patient-centered outcomes.