Cancer diagnosis has certainly got more interesting. A pair of scientists is has led a study that involves pigeons detecting breast cancer.
The ambitious research was carried out by University of California Davis pathologist Richard Levenson and University of Iowa psychologist Edward Wasserman. Together they experimented on sixteen pigeons, which they closely trained to spot the difference between malignant and benign breast tissue samples.
According to their study, which they have published in PLOS ONE, the flock of pigeons functioned as "trainable observers" to obtain a better understanding on the human behavioral activities that are critical in making correct diagnosis especially on breast cancer tissues.
Pathology training is not only expensive but also extensive, since tissues can appear differently from one another. For example, there are tissues that may appear malignant but are actually benign. A wrong analysis, however, can spell the difference between ruining and saving a life.
The pigeons were chosen based on their excellent visual abilities. They can sense more colors than humans, and in general, they practice the "what they see is what they get" approach when it comes to shapes that they're seeing.
Nevertheless, training them to work similarly as pathologists is difficult and tricky. The scientists didn't want to overdo it that it could not provide a more accurate diagnosis if there's a slight change in tissue sample slides, for example. Further, animals have the tendency to respond according to the reaction and based on their observation on humans.
To avoid these problems, the experiment was designed to limit the pigeon in a box with a computer screen at the front. The screens would show different slides of breast tissues with accompanying buttons of yellow and blue, which can stand for malignant or benign. The slides were shown at random and colored buttons for cancerous tissues were different among the birds. These were all done to remove bias and reduce interaction with humans.
The birds were trained only once per day, and they were rewarded with a pellet, which was automatically given by the setup if they got their answers right.
Based on the results, individually pigeons scored lower than humans in terms of accuracy, but if their guesses were viewed collectively, their accuracy increased to 99%. Further, after one day of training, their accuracy already rose to 80%.
Despite the pleasing performance, we might not be seeing pigeon pathologists anytime soon.
