Limits of Technology - Man Beats Machine in Assessing Breast Masses

Keeping Abreast of Computer-Aided Breast Cancer Diagnosis

Twenty years ago, the late Philip Olson, Minneapolis radiologist and quarterfinalist in the National Amateur Golf Championships, said to me, “It takes a radiologist to interpret technology-generated images.” Phil was speaking about two computer-aided technologies – CT Scans and MRIs.

Phil was as good at reading the future as he was in reading greens. At least that’s what a New England Journal of Medicine article shows It says radiologists, acting alone without computer aid, can assess mammograms more accurately alone than with computer help.

The Journal study of 429, 345 screening mammograms read from 1998 to 2002 at 43 facilities in three states found that,

The use of computer-aided detection is associated with reduced accuracy of interpretation of screening mammograms. The increased rate of biopsy with the use of computer-aided detection is not clearly associated with improved detection of invasive breast cancer.

CAD costs more, doesn’t enhance cancer pick up rates, creates more false-positives, and leads to more needless biopsies. The study concluded,

• Regular mammography and CAD found roughly the same number of cancers.
• Rate of false positives was 31% higher in CAD.
• 157 women with CAD had to be called back for re-exam
• Of those 157, 15 got a biopsy.
• If every breast screening center in the U.S. were to adopt CAD, costs for screening would rise 18%, or $550 million.

Of the study, Dr. Ferris Hall, a radiologist at Boston’s Beth Israel, said,

This is a substantial hit for this technology. This study suggests that it may actually be more harmful.

Of his group’s practice, Hall commented,

We have been using CAD for three or four years. We got it not because we are saving lives but because (insurers) are paying for it. The payback time (on the cost of the system) is only three years if you are a busy practice, and as a radiologist I am extra bucks just for pressing a button.” Hall added that reimbursement for CAD was mandated by Congress in 1998 despite little evidence to support its use and after heavy lobbying by CAD equipment makers.

Phil would have understood. He knew computers couldn't handle shades of gray, had no right brain, and possessed no soul or feel for human consequences. Yes, computers are great at piling up data, sifting through it, and even predicting where data is likely to lead in the future, but when it comes to deciphering human images, computers don’t help.

References

1. Fenton, Joshua, and 10 co-authors, Influence of Computer-Aided Detection on Performance of Screening Mammography, New England Journal of Medicine, 356:1399-1409, April 5, 2006.
2. Hall, F.M, Breast Imaging and Computer-Aided Diagnosis, New England Journal of Medicine,356:1464-1466, April 5, 2006.
3. Armstrong, David, Study Shows Flaw in Mammography Tool: Popular Computer-Aided Method Yields More False Positives than Less Expensive Approach; Risk of Unnecessary Procedures, Wall Street Journal, April 5, 2007