Dogs that detect cancer? Check. Fruit flies can do it? Check.
Well, yes. Pigeons have excellent vision and, it turns out, can be trained to identify malignant and benign breast tumors pretty much as a radiologist or a pathologist would — by looking at a mammogram or a slide from a biopsy.
Researchers at the University of Iowa and elsewhere experimented with 16 birds. All had participated in studies before, so presumably they knew the drill: They get food when they pick right, nothing when they pick wrong. Figure out what the nice people are looking for, or stay hungry.
The scientists put the pigeons in little boxes with touch-sensitive monitors that showed slides of breast tissue. Then the scientists trained them to peck a blue rectangle when they saw benign tissue and a yellow one when the slide showed malignant tissue.
In one experiment, the handlers gave the birds slides in color and black-and-white, and in varying degrees of magnification, randomly presenting benign and malignant tissue. In another, the pigeons were tested to see if, on a mammogram, they could discern calcifications, which are associated with cancer and which radiologists often find difficult to recognize.
In a third test, the birds were shown mammograms to see if they could distinguish a benign from a malignant mass. The study was published in PLOS One.
The pigeons were quick to learn to discriminate benign from malignant breast tissue in the first experiment, averaging 87 percent correct scores on slides they had trained with and 85 percent even with slides they had never seen.
On the second test, they did just as well, learning to see subtle signs of calcifications with 85 percent accuracy after two weeks of training.
But they didn’t fare as well with the masses on the mammograms. Some were accurate 80 percent of the time, others only 60 percent. One birdbrain never managed to do better than chance.
Pigeons are not going to replace radiologists or pathologists anytime soon. But, the authors wrote, they may be useful as surrogates for human subjects in medical image perception studies because they can be used in repetitive ways that few people could tolerate. Pigeons may also be useful in exploring the effect of technical aspects of color parameters — for example, contrast and brightness — on the accuracy of perception.
(Click here to read the full article, written by Nicholas Bakalar, NY Times)