AI Research Analysis
AI in Diagnostic Radiology: What Happens When Algorithms Are Updated
Published by Radiation on 26 January 2026. Authors: Martine Rustøen Skregelid, Kasim Ibrahim-Pur, Flemming Skjøth, Malene Roland Vils Pedersen, Helle Precht. DOI: 10.3390/radiation6010004
Interpretation of radiographs is prone to diagnostic errors. Artificial intelligence (AI) has shown promising results in fracture detection, although systematic evaluation of software updates remains limited. This study compares the diagnostic performance of two versions of an AI-based fracture detection software in hand and ankle radiographs and assesses the influence of AI output on diagnostic decisions. Methods: This retrospective diagnostic accuracy study included 193 hand and ankle examinations obtained during routine clinical practice at Lillebaelt Hospital, Denmark. Radiographs were analysed using two versions of the same AI software and compared with the diagnostic report as the reference standard. Diagnostic performance of both versions was assessed using diagnostic accuracy metrics. Exploratory subgroup analyses were conducted to further investigate the difference in performance. The influence of AI was evaluated by the proportion of reports revised after review of AI output. Results: The newest software version demonstrated higher diagnostic performance than the older one (accuracy 0.933 vs. 0.824; p < 0.001). Similar improvements were observed across patient subgroups. Excluding radiographs containing casts resulted in only minimal changes in performance (accuracy in version 2: 0.930 vs. 0.933). In 8 of 15 discordant cases, reporting radiographers revised the initial assessment upon reassessment. Conclusions: The newest version demonstrated higher overall diagnostic performance, indicating that software updates can enhance the accuracy of AI-assisted fracture detection. The proportion of revised assessments suggests that radiographers' decisions may be influenced by AI output.
Executive Impact & Key Findings
This study rigorously evaluated the impact of software updates on AI performance in diagnostic radiology, specifically for fracture detection in hand and ankle radiographs. The latest AI version significantly improved diagnostic accuracy and consistency across diverse patient populations. Critically, AI's output was shown to influence radiographer decisions, highlighting its potential as a decision support tool but also underscoring the need for careful implementation and continuous evaluation. These findings suggest that regular algorithm updates can substantially enhance clinical utility and patient outcomes, warranting systematic assessment of new AI versions in healthcare settings.
Key Insights for Enterprise AI:
0
Accuracy Improvement (v2 over v1)
0
Specificity Improvement (v2 over v1)
0
PPV Improvement (v2 over v1)
0
Radiographer Decisions Influenced
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
This category focuses on the practical diagnostic accuracy and efficacy of AI in real-world clinical settings, comparing different software versions and their impact on radiographer decisions.
0.933
V2 Diagnostic Accuracy
The newest AI software version demonstrated an accuracy of 0.933 for fracture detection, significantly outperforming the previous version and reinforcing AI's potential in diagnostic radiology.
Clinical Workflow Integration
Radiograph Acquisition (2023)
→
V1 AI Analysis & Radiographer Report (with V1 output available)
→
V2 Re-evaluation (2025)
→
Comparison & Reassessment of Discordant Cases
The study integrated two versions of AI into a retrospective analysis of clinical workflow. Initial reports were generated with V1 AI assistance, followed by a blind re-evaluation with V2 AI, and finally a radiographer reassessment of discrepancies.
Software Version Performance Comparison
| Metric | Version 1 | Version 2 | Improvement |
|---|---|---|---|
| Accuracy | 0.824 | 0.933 | Significant (p < 0.001) |
| Sensitivity | 0.818 | 0.909 | Moderate (p = 0.061) |
| Specificity | 0.828 | 0.952 | Significant (p = 0.002) |
| PPV | 0.800 | 0.941 | Significant (Diff 0.14) |
| NPV | 0.844 | 0.925 | Moderate (Diff 0.08) |
Conclusion: Version 2 consistently demonstrated superior diagnostic performance across key metrics, indicating the effectiveness of software updates in enhancing AI capabilities.
This category explores how AI output affects human diagnostic decisions, examining instances where radiographers revised their initial assessments after reviewing AI findings.
8 of 15
Cases Revised by Radiographers
Out of 15 discordant cases between v2 AI and original radiographer reports, 8 (53.3%) were revised upon reassessment, suggesting a direct influence of AI output on diagnostic decisions.
Impact of AI on Radiographer Reassessment
Problem: Discrepancies arose between the latest AI version (v2) and initial radiographer reports, necessitating a reassessment process.
Solution: Radiographers re-evaluated 15 discordant cases. In 8 instances (53.3%), the original assessment was revised, primarily from positive to negative, after considering the v2 AI output.
Outcome: This demonstrates AI's significant influence as a decision support tool, potentially refining diagnostic accuracy by prompting human re-evaluation in uncertain or conflicting cases. However, it also highlights the need for a robust understanding of AI limitations to prevent overreliance.
This category delves into the technical improvements and robustness of the AI algorithm across versions, including its performance in challenging scenarios like radiographs with casts.
0.002
Accuracy Change (with/without casts) for V2
Excluding radiographs with casts resulted in only a minimal accuracy change of 0.002 for version 2 (0.930 vs. 0.933), demonstrating the algorithm's robustness.
Performance with/without Casts
| Version | Original Accuracy (with casts) | Accuracy (without casts) |
|---|---|---|
| Version 1 | 0.824 | 0.815 |
| Version 2 | 0.933 | 0.930 |
Conclusion: The improved version 2 maintained high diagnostic accuracy even when radiographs with visible bandages or casts were excluded, suggesting enhanced robustness and less susceptibility to such imaging artifacts compared to version 1.
Calculate Your Potential ROI with Enterprise AI
Estimate the impact of AI on your operational efficiency and cost savings. Adjust the parameters below to see the potential returns tailored to your organization.
Estimated Annual Savings
$0
Hours Reclaimed Annually
0
Your Enterprise AI Implementation Roadmap
A typical AI integration journey, from initial assessment to full-scale deployment and continuous optimization, designed for minimal disruption and maximum impact.
Phase 1: Discovery & Strategy
Comprehensive assessment of your current infrastructure, identification of key AI opportunities, and development of a tailored implementation strategy with clear KPIs.
Phase 2: Pilot Program & Validation
Deployment of AI solutions in a controlled environment, rigorous testing, performance validation against benchmarks, and collection of initial user feedback.
Phase 3: Full-Scale Integration
Seamless integration of validated AI solutions across your enterprise, comprehensive training for your teams, and establishment of robust monitoring systems.
Phase 4: Optimization & Scaling
Continuous monitoring and refinement of AI models, identification of new applications, and strategic scaling of solutions to maximize long-term ROI and efficiency.
Ready to Transform Your Enterprise with AI?
Unlock the full potential of artificial intelligence to drive innovation, enhance efficiency, and achieve unparalleled competitive advantage. Our experts are ready to guide you.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat