Latest Trends and AI Innovations in Breast Cancer Screening Guidelines
- The National Comprehensive Cancer Network (NCCN) has updated its breast cancer screening guidelines to introduce AI-based risk assessment using mammograms starting at age 35.
- Under the new guidelines, a five-year risk threshold of 1.7% or higher is now used to define increased risk and guide subsequent clinical action.
- The integration of artificial intelligence into screening is supported by data evaluating the efficacy of systems such as Google’s mammography AI system, version 1.2.
The National Comprehensive Cancer Network (NCCN) has updated its breast cancer screening guidelines to introduce AI-based risk assessment using mammograms starting at age 35. This shift is intended to predict a woman’s five-year breast cancer risk, enabling earlier identification and clinical intervention.
Under the new guidelines, a five-year risk threshold of 1.7% or higher is now used to define increased risk and guide subsequent clinical action.
AI Performance in Mammography Screening
The integration of artificial intelligence into screening is supported by data evaluating the efficacy of systems such as Google’s mammography AI system, version 1.2. In a retrospective study involving 115,973 mammograms from five National Health Service screening services with a 39-month follow-up, the AI system demonstrated superior sensitivity compared to first readers.
The AI achieved a sensitivity of 0.541, while the first reader achieved 0.437. In terms of specificity, the AI was found to be noninferior, recording 0.943 compared to 0.952 for the first reader.
The study found that the cancer detection rate increased from 7.54 to 9.33 per 1,000 women. The AI system detected 25.0% of interval cancers.
The AI’s performance was particularly strong regarding invasive cancers and first-time screens. For first screens, the use of AI resulted in 8.8% higher detection and 39.3% fewer recalls.
Clinical Workflow and Efficiency
Beyond detection accuracy, AI implementation has shown the potential to reduce the burden on clinicians. Simulated second-reader replacement using the AI system reduced reading time by 32%.
Despite the reduction in time, this simulated replacement increased detection by 17.7%.
Implementation Challenges and Safety
While retrospective results are positive, prospective noninterventional feasibility deployment across 12 sites involving 9,266 cases revealed critical implementation requirements. The deployment confirmed that the system is technically feasible, but it also identified a distribution shift.
This distribution shift indicates that AI systems require adaptive calibration and continuous monitoring to ensure equity and safety. No systematic demographic disparities were observed during the evaluation.
Broader AI Applications in Breast Care
The use of AI in breast cancer care extends beyond mammography. AI-based techniques have been used to enhance the application of ultrasound elastography in breast cancer screening.
AI is being utilized to transform other areas of care and research, including the development of AI-enabled trial designs to improve clinical trial enrollment.