Radiologists review interval cancers on DBT and AI detection performance

In a review of 46 interval DBT cancers, 17% were missed (false negative). AI correctly identified 35%, but only 22% of cancers with high mammographic visibility, suggesting about 20% of interval cancers could be diagnosed earlier using AI.

• Radiologists classified 22% of interval cancers as minimal-sign significant and 20% as minimal-sign non-specific, with 41% true negative.
• The AI model gave high-risk scores to some cases and incorrectly located 24% of cancers.
• The study was a retrospective informed consensus-based review of 46 DBT screening examinations.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.