AI-STREAM: Preliminary Analysis
Artificial intelligence for breast cancer screening in mammography
Published in Nature Communications on March 06, 2025 by Yun-Woo Chang et al. (DOI)
This preliminary analysis of the AI-STREAM study, a prospective multicenter cohort, demonstrates that integrating AI-CAD significantly improves cancer detection rates (CDRs) for breast radiologists (BRs) in a single-read setting without increasing recall rates (RRs). The study found a 13.8% higher CDR for BRs using AI-CAD compared to those without, highlighting AI's potential to enhance diagnostic performance for early breast cancer detection.
Quantifiable Enterprise Impact
Key metrics from the AI-STREAM study showcase the tangible benefits of AI integration in mammography screening.
0
Increased CDR for BRs with AI-CAD
0
p-value for RR (BRs with/without AI-CAD)
0
Increased CDR for GRs with AI-CAD
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Performance Metrics
Clinical Impact
Methodology & Workflow
Examine the core diagnostic accuracy improvements across different reading strategies.
Diagnostic Performance Comparison (CDR & RR)
| Reading Strategy | CDR (%) | Recall Rate (%) |
|---|---|---|
| BRs without AI-CAD | 5.01 | 4.48 |
| BRs with AI-CAD | 5.70 | 4.53 |
| AI Standalone | 5.21 | 6.25 |
| GRs without AI-CAD | 3.87 | 6.31 |
| GRs with AI-CAD | 4.89 | 6.89 |
Conclusion: Breast Radiologists using AI-CAD showed a significant increase in CDR (5.70% vs 5.01%) without a significant change in RR. General Radiologists saw a larger CDR increase but also an increase in RR. Standalone AI had comparable CDRs to BRs but with significantly higher RRs.
Understand how AI-CAD influences the detection of specific cancer types and characteristics.
+13.3%
Increased detection of node-negative cancers with AI-CAD (p=0.001)
+13.3%
Increased detection of small-sized cancers (<20mm) with AI-CAD (p=0.002)
AI-Assisted DCIS Detection Example
Figure 2 illustrates a case where a Breast Radiologist initially interpreted a mammogram as non-recall (malignant scale 2) without AI. Upon automatic presentation of AI-CAD results, which marked an abnormal score of 75%, the BR re-evaluated the mammography. With AI-CAD assistance, the BR then interpreted focal asymmetry as malignant scale 4, leading to a recall decision. Subsequent ultrasonography and guided biopsy confirmed Ductal Carcinoma In Situ (DCIS) with microinvasion, highlighting AI-CAD's role in upgrading initial interpretations and enabling earlier diagnosis.
Key Takeaway: AI-CAD can act as a crucial 'second look' for radiologists, prompting re-evaluation of subtle findings and improving the detection of early-stage cancers like DCIS, which might otherwise be missed during initial interpretation.
Explore the study's design and how AI-CAD integrates into existing screening workflows.
Enterprise Process Flow
25,008 women screened (2021-2022)
→
Exclusions (parenchymal change, consent withdrawal, data errors)
→
24,543 participants included
→
BRs read without AI-CAD (Test 1)
→
BRs read with AI-CAD (Test 2) / GRs read with/without AI-CAD (Exploratory)
→
AI Standalone assessment
→
Recall/Non-Recall Decision
→
Diagnostic Work-up (imaging, biopsy)
→
Pathologically confirmed cancer (1-year follow-up)
→
Final evaluation for interval cancers (2-year follow-up linkage to Registry)
Estimate Your AI ROI
Calculate the potential time savings and cost efficiencies your organization could achieve by integrating AI, tailored to your specific operational context.
Estimated Annual Savings
$0
Annual Hours Reclaimed
0
AI Implementation Roadmap
A phased approach to integrate AI into your breast cancer screening program, ensuring a smooth transition and maximizing diagnostic benefits.
Phase 1: Pilot & Integration
Conduct a small-scale pilot with AI-CAD in a single-read setting. Integrate AI-CAD into existing mammography workstations and train radiologists on its use and interpretation guidance. Establish data collection protocols for performance monitoring.
Phase 2: Expanded Deployment & Monitoring
Roll out AI-CAD to additional screening centers. Continuously monitor CDRs, RRs, and interval cancer rates. Implement regular calibration and quality assurance processes for AI thresholds to maintain optimal operating points.
Phase 3: Workflow Optimization & Advanced AI
Evaluate and refine screening workflows based on AI-CAD insights, potentially incorporating AI for triage or as an independent reader for low-risk cases. Explore integration with DBT and other imaging modalities for enhanced diagnostic accuracy.
Ready to enhance your breast cancer screening program with AI?
Connect with our experts to discuss how AI can transform your diagnostic workflows and improve patient outcomes.
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