What is a mammogram with Computer-Aided Detection (CAD)?

What is a Mammogram with CAD?

A mammogram with Computer-Aided Detection (CAD) is a breast cancer screening technique where computer software analyzes digitized mammographic images to identify and mark suspicious lesions (masses or microcalcifications) for radiologist review as a "second reader."

How CAD Technology Works

CAD systems digitize mammogram images—either from traditional screen-film mammography or digital mammography—and use algorithms to analyze them for abnormalities suspicious for breast cancer 1. The software marks areas of concern, prompting the radiologist to re-evaluate these regions before making a final interpretation 1.

• Three commercial CAD systems received FDA approval for use with screen-film mammography, with approximately 500 systems installed in the United States as of 2003 1
• Only one commercial CAD system was FDA-approved for use with digital mammography at that time 1
• By 2016, approximately 92% of mammography facilities in the United States had adopted CAD into clinical practice 1

Performance Characteristics and Clinical Impact

Cancer Detection Benefits

CAD can increase cancer detection rates but with important tradeoffs. In the largest clinical series evaluating CAD performance 1:

• Cancer detection increased from 41% to 49% (a 19.5% increase) when radiologists used CAD compared to interpretation without CAD
• Overall detection rate improved from 3.2 to 3.8 cancers per 1,000 women screened
• CAD may aid radiologists in detecting early-stage malignancies with proportionate increases in recall rates 1

Important Limitations and Pitfalls

However, real-world implementation revealed significant concerns about CAD's clinical utility 1:

• CAD generates 1.5-4 false-positive prompts per case, which can adversely affect clinical decision-making 2
• Recall rates increased from 6.5% to 7.7% (an 18.5% increase), meaning more women are called back for additional testing 1
• Specificity decreases substantially, causing additional recall in approximately 6-35% of women 2
• CAD increased detection of DCIS (ductal carcinoma in situ), a precancerous lesion with low mortality risk 1
• CAD has NOT been shown to reduce mortality from breast cancer and has been associated with rising screening costs 1

Comparative Performance

CAD does not perform as well as double human reading in organized breast screening programs where double reading is the standard of care 2. The high number of false-positive prompts can detract from CAD's application in screening practice 2.

Evolution to Modern AI Systems

The original CAD systems from the 1990s-2000s used early AI algorithms 1. Modern AI applications use convolutional neural networks (CNNs), which have greater accuracy than older CAD methods and have renewed enthusiasm for automated mammography interpretation 1.

Clinical Recommendation

Use CAD selectively, recognizing it may help less experienced radiologists and in detecting microcalcifications, but be prepared for increased false-positives and recall rates without proven mortality benefit 1, 3. The technology is most beneficial when radiologists understand its limitations and don't allow false-positive prompts to override clinical judgment 2.