Artificial Intelligence-Based Computer-Aided Detection in Breast Cancer Diagnosis: Variation by Breast Density,...

Area of Science:

• Radiology
• Artificial Intelligence in Medicine
• Oncology

Background:

• Artificial intelligence computer-aided detection and diagnosis (AI-CAD) systems are increasingly used in medical imaging.
• Understanding factors influencing AI-CAD scores is crucial for accurate interpretation and clinical integration.

Purpose of the Study:

• To investigate if an FDA-approved AI-CAD system assigns different case scores based on imaging features, tumor characteristics, and breast density.
• To assess the relationship between AI-CAD output and clinicopathologic variables.

Main Methods:

• Retrospective multisite study of patients undergoing biopsy after abnormal screening tomosynthesis.
• A commercial AI-CAD tool generated case scores; imaging features, breast density, tumor size, grade, and pathology were collected.
• Univariable and multivariable linear regression analyses were used to assess associations.

Main Results:

• AI-CAD scores were significantly higher for malignant cases compared to benign ones.
• Scores were highest for calcifications and mass with calcifications imaging features.
• Multivariable analysis revealed higher scores associated with non-dense breasts, higher tumor grade, larger tumor size, older age, and specific imaging features.

Conclusions:

• AI-CAD scores are influenced by breast density, imaging presentation, and tumor characteristics.
• AI outputs appear to reflect a combination of imaging findings and underlying tumor attributes.
• Context-informed interpretation of AI-CAD outputs is essential for effective clinical application.