Reproducibility of computer-aided detection marks in digital mammography

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Digital mammography is a key tool for breast cancer screening.
• Computer-aided detection (CAD) systems aim to improve mammogram interpretation accuracy.
• Serial imaging without compression release is used in specific clinical scenarios.

Purpose of the Study:

• To assess the performance and consistency of a CAD system.
• To evaluate CAD system accuracy on serial mediolateral oblique (MLO) digital mammograms.
• To analyze CAD system reproducibility without releasing breast compression.

Main Methods:

• A CAD system was applied to full-field digital mammograms of 82 patients with diagnosed breast cancer.
• Two MLO views were analyzed per patient, taken serially without compression release.
• Sensitivity and reproducibility of CAD marks for masses and microcalcifications were evaluated.

Main Results:

• CAD sensitivity was 71% for masses and 80% for microcalcifications on initial MLO views.
• Sensitivity decreased on second MLO views (68% for masses, 17% for microcalcifications).
• Reproducibility of CAD marks was high for masses (84%) but very low for microcalcifications (0%).

Conclusions:

• The CAD system demonstrates good sensitivity for detecting breast lesions.
• Reproducibility of microcalcification detection is a significant limitation in serial MLO views without compression release.
• Image variations due to patient positioning or movement impact CAD performance for microcalcifications.