Using Pigeons to Diagnose Cancer?

Dr. Pigeon examines a microscope image of tissue to see if it is cancerous. (image from the article being discussed)
Dr. Pigeon examines a microscope image of tissue to see if it is cancerous.
(image from the article being discussed)

Pigeons are surprisingly smart birds. Quite a while ago, I wrote about how pigeons can learn to solve probability problems better than some people. In addition, homing pigeons can use sound to find their way back home. Now here’s a new way of looking at the intelligence of pigeons: they can use their pattern-recognition abilities to find cancer in microscopic images of tissue!

In the study, researchers put 16 pigeons in a box with a computer screen (shown above). They were shown microscopic images of breast tissue, some of which indicated the presence of cancer, and some of which did not. There were two “buttons” on the computer screen, one on each side of the image. One button represented the answer “this image shows cancer,” and the other button represented the answer “this image does not show cancer.” Each pigeon was free to choose either button, and if the pigeon was correct, it got a pellet of food.

At first, of course, the pigeons’ answers were random. Over time, however, they learned to look at patterns in the image, and within a matter of hours, they were identifying cancer at a rate that was superior to random guessing. Within a month, they were spotting cancer with about an 80% accuracy rate. The most interesting effect, however, was obtained when several birds were shown an image and their combined answers were used to determine whether or not cancer was present. When that was done, the accuracy of the diagnosis was 99%, which is about as good as a trained person!

One very important aspect of this study is that the birds did all their work alone, in the box. There wasn’t a person present. This is important, because many animals are extremely perceptive, and they can take visual cues from people, even when those cues are unconscious. Thus, whenever an animal can do something novel, there is always the possibility that it is responding to its caretaker, not performing the feat on its own. Since people were not present during the times the pigeons were shown the images in this study, it is clear that the birds were actually learning how to recognize patterns that indicate cancer.

Interestingly enough, the researchers repeated the experiment using mammogram images instead of microscopic images of tissue. When confronted with these images, the pigeons didn’t do nearly as well. The researchers tried to get them to spot suspicious masses in mammogram images, which is a difficult task, even for some trained people. While the pigeons could eventually memorize the correct answer for images that had been shown to them multiple times, they couldn’t look at a new image and correctly identify a problem, as they did with the microscopic images of tissue. Obviously, then, a pigeon’s pattern-recognition abilities are well suited to some types of images but not others.

Now please don’t think that the researchers carried out this study in the hopes of having pigeons do cancer diagnosis one day! They were trying to compare how pigeons learn to how people learn. While it’s not clear what this study can tell us about that, it is clear that pigeons are very good at learning to recognize certain patterns.

It gives a whole new meaning to the term “bird brain,” doesn’t it?

7 thoughts on “Using Pigeons to Diagnose Cancer?”

  1. My first thought is that it’s related to the visual component of their directional/mapping ability. Birds don’t have GPS, but between some sense of geomagnetic location and a visual memory for minute landscape and astrological details, they can perform impressive feats of geolocation.

    I was watching something yesterday regarding the ability of dogs to identify some kinds of cancer by smell. The results are slightly less than 99%, but still impressive.

  2. I once did a postmortem examination of a small finch. That particular bird had a brain that weighed a little over 3% of its body weight. I am seeing figures of around 2% of body weight for the human brain, so your comment about a whole new meaning to the term “bird brain” made me think about those findings. In the case of the finch, it had a brain that was proportionally slightly larger than a human brain.

    1. In The Central Nervous System of Vertebrates by Rudolf Nieuwenhuys, Hans J. ten Donkelaar, and Charles Nicholson (p. 2113), there is a graph of brain weight versus body weight for various orders of birds, and it is amazing how variable it is. The percentage seems to range from about 5% of body weight to about 0.5% of body weight. Most of them are greater than 2%, however, which is the percentage for people.

  3. The chart you mentioned sounds interesting Dr Wile. Do they have the ratio for a ruby-throated hummingbird or any other hummingbird?

    1. They have several dots on their graph for hummingbirds, but they don’t list the species. Instead, they list four references for their bird data, and I don’t have access to those resources. The dots for hummingbirds are on the high side, clustering at about 4%.

  4. Thanks for the information Dr Wile. At one time I was trying to find the weight of a ruby-throated hummingbird eye and couldn’t find any weights for the eye. Using a measured diameter that I found for the eye of a ruby-throated hummingbird I made some rough estimates of the weight. My estimate for an intermediate sized ruby-throated hummingbird eye was about 50 milligrams.

    One piece of literature that I saw suggested that the combined eye weight of the hummingbird was more than the brain weight. That is why I asked the question. An intermediate average weight between males and females according to Wikipedia is about 3.6 grams (3.4g males and 3.8g females). At 4% of the body weight, the brain of the intermediate sized ruby-throat would be about 144 milligrams. If my estimates are correct the two eyes, at about 100mg, would not weigh more than the brain, but they are pretty close.

    If you have seen information about, or know of any actual weights for hummingbird eyes, I would appreciate it if you would let me know what weights you have seen.

  5. Doesn’t surprise me. In fact, some birds might do even better. When birds are flying overhead and scanning the ground for food, they must be able to use visual clues to distinguish an item of interest against a background of non-food vegetation/rocks/ground cover. I would expect this to be enhanced with color vision. Since most microscope slides are stained with dyes to bring out nucleic acid, collagen, or protein components, human pathologists learn to recognize subtle differences in color, pattern, cell size, etc. This would be even easier for an animal that makes its living, so to speak, distinguishing subtle details against a complex background. It would be interesting to compare a bird’s capability in pattern recognition to a computer programmed for the same task.

Comments are closed.