Blinded comparison of computer-aided detection with human second reading in screening mammography

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Screening mammography is crucial for early breast cancer detection.
• False-negative cases in mammography can lead to delayed diagnosis.
• Computer-aided detection (CAD) systems aim to assist radiologists in identifying abnormalities.

Purpose of the Study:

• To compare the effectiveness of a human second reader versus CAD in reducing false-negative mammogram interpretations.
• To evaluate the impact of a human second reader and CAD on recall and biopsy rates.

Main Methods:

• Retrospective review of 6,381 screening mammograms.
• Mammograms were initially interpreted by a primary radiologist.
• The same radiologist then reinterpreted studies with CAD assistance.
• A second human reader, blinded to CAD results, reviewed studies for missed abnormalities.

Main Results:

• Both human second reader and CAD increased recall and biopsy rates.
• The human second reader identified two cancers missed by the primary reader and not flagged by CAD.
• Positive predictive value was 0% for CAD reader and 40% for human second reader.
• Cancer detection rate increased by 15.4% with a human second reader, versus 0% with CAD.

Conclusions:

• Both human second readers and CAD systems can potentially increase mammography cancer detection rates.
• No statistically significant difference was found between human second reader and CAD due to small sample size.
• Human readers offer the advantage of interpreting the significance of findings, unlike CAD systems that only flag potential abnormalities.