Artificial intelligence reading digital mammogram: enhancing detection and differentiation of suspicious...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Oncology

Background:

• Grouped microcalcifications on mammograms can indicate early breast cancer.
• The morphology and distribution of calcifications are key indicators of breast disease.
• Accurate detection and classification of microcalcifications are vital for effective breast cancer management.

Purpose of the Study:

• To evaluate the impact of artificial intelligence (AI) on improving the sensitivity of digital mammograms.
• To assess AI's role in the detection and specification of grouped microcalcifications.
• To determine if AI can enhance the diagnostic performance for suspicious microcalcifications.

Main Methods:

• Retrospective analysis of grouped microcalcifications in 447 patients.
• Correlation of detected microcalcifications with AI applied to initial mammograms.
• AI provided heat maps, demarcation, and quantitative suspicion scoring; histopathology confirmed malignancy.

Main Results:

• AI demonstrated a 67.5% correlation with malignant microcalcifications (P<.001).
• Mammography sensitivity for microcalcifications was 94.7%, with a 82.1% negative predictive value.
• AI showed potential in detecting suspicious lesions, though human expertise is needed for final specification.

Conclusions:

• AI systems can enhance mammogram sensitivity for detecting suspicious microcalcifications.
• AI serves as a valuable decision support tool for microcalcification detection and classification.
• While AI shows promise, expert human interpretation is essential for accurate diagnostic specification.