[A comparison between physicians' interpretation and a CAD system's cancer detection by using a mammogram database in...

Area of Science:

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

Background:

• Developing automated detection algorithms for mammographic abnormalities is crucial for improving breast cancer diagnosis.
• Computer-aided diagnosis (CAD) systems aim to assist radiologists in interpreting mammograms, potentially enhancing detection rates.

Purpose of the Study:

• To evaluate the efficacy of a developed computer-aided diagnosis (CAD) system in detecting malignant lesions and clustered microcalcifications.
• To compare the diagnostic performance of the CAD system against the interpretations of 579 physicians using a set of 100 mammograms.

Main Methods:

• A cohort of 100 mammograms, including 21 malignant and 29 benign cases, was analyzed.
• The study compared the cancer detection results of a proprietary CAD system with the interpretations of 579 physicians.
• Physicians' performance was assessed within a self-learning course context.

Main Results:

• The CAD system detected 7 out of 8 malignant lesions missed by physicians with an average sensitivity below 60%.
• Physicians' average sensitivity was 76%, while the CAD system achieved a 90% detection rate.
• A combined approach (logical OR) of physician interpretation and CAD output increased sensitivity to 97%.

Conclusions:

• The CAD system demonstrates significant potential to enhance breast cancer detection rates, particularly for subtle malignant lesions.
• Integrating CAD system outputs as a diagnostic aid can effectively improve the sensitivity of mammogram interpretations, benefiting physicians of all experience levels.
• This technology offers a promising tool for improving the accuracy and efficiency of breast cancer screening programs.