Can pigeons home in on cancer?

We regularly see stories in the media about animals sniffing out their owners’ cancer. And while these stories usually involve dogs, we’ve heard tales of horses, cats and even a donkey helping spot the disease in their owners.

And we’ve talked about the science behind animal’s keen sense of smell before.

Their ability to sniff out volatile molecules given off by tumours is fascinating and has inspired some interesting applications, such as electronic noses.

Milton Keynes University Hospital is even running a trial to see if highly-trained sniffer dogs can help to diagnose prostate and bladder by sniffing urine. But it’s not really practical to have this four-footed ‘lab test’ in every doctor’s surgery.

Today we can add another creature to the list of animals that can spot cancer early: the pigeon.

But rather than their sense of smell, it’s pigeons’ well-documented visual skills, that are under scrutiny here.

But can pigeons really detect the disease? And would we ever expect to see our feathered friends taking up residence in a pathology lab?

The short answer is ‘no’, but – as you’ll see below – we do have much to learn about how our own brains process images, and pigeons could well play a role in understanding this better.

Columba livia

Like humans, pigeons have highly developed visual systems. They can learn how to identify complex patterns and objects in pictures, and sort them into categories. They can even tell all the letters of the alphabet apart.

And this has lead many researchers studying how our own visual systems work, to study the pigeons’ eye view of the world.

Dr Robert Levenson, the study’s lead author, is based at University of California and studies pathology and laboratory medicine.

It was after he heard about pigeons’ exceptional visual memories that he set out to find out what they could do with medical images.

“I was driving into work a few years ago and I heard a news report about a study on visual memory in pigeons,” he recounts, “and for no particularly good reason, it occurred to me to wonder how well pigeons might do when confronted with some pathology-related image tasks.”

Could they tell diseased tissue apart from healthy tissue? And could this tell us anything about how the human brain processes these images?

Levenson’s team’s study, published today in PLOS ONE journal, showed that eight specially trained pigeons were correctly able to tell the difference between microscope images of benign and malignant human breast tissue over 80 per cent of the time.

The researchers trained the pigeons using a technique called food reinforcement – when the pigeon chose correctly they got food, if they didn’t get the right answer no food was delivered. They did this every day for 144 trials.

After the training was complete, the pigeons were tested to see if they could do the same, but with new images and no reinforcement if they got the answer right.

The pigeons’ training environment Image courtesy of Levenson and Wasserman.

Results showed that not only could the pigeons categorise the images, but also successfully demonstrated their ability to learn how to find the key discriminating features necessary to make that classification.

In fact, even when the images were presented in different formats, such as varying magnification, colour versus monochrome, or rotated, the pigeons were still able to spot which images contained signs of cancer.

This told the researchers that the way pigeons visually discriminate between medical images is similar to humans. For example, pathologists often look for certain characteristics, like cell shape, on an image that help determine if it’s benign or malignant. It seems from the study’s results that pigeons may use similar characteristics to differentiate between images.

The team also showed that if they grouped the pigeons together and got them to vote on the images, otherwise known as “flock-sourcing”, they got the answer right 99 per cent of the time.

They also tested the pigeon’s ability to find signs of cancer in mammogram images. Results showed that the pigeons could only memorise the training images. When they were shown new images of mammograms the pigeons couldn’t tell the difference between the ones that showed signs of cancer and the one’s that didn’t.

“Although pigeons are unlikely to be called upon to offer clinical diagnostic support, it does seem quite possible that their discriminative abilities may be turned to a useful purpose,” writes Dr Levenson.

Useful Purpose

The aim here isn’t to bring pigeons into routine clinical use – that remains a flight of fancy.

But Levenson’s work could lead to cheaper and more rapid studies of how the brain perceives images: pigeons could be used to replace humans in certain medical image perception studies, as it would save the cost of paying and recruiting clinicians as subjects.

The researchers also propose that pigeons could help develop computer-assisted medical image recognition tools.

So while pigeons won’t replace pathologists or radiologists in a clinical setting any time soon, (most likely never), they might be able to help researchers develop better ways to spot cancer.

But the idea of trusting them to pigeon-hole cancer in people is probably nothing more than a bird-brained scheme.

-Misha

• Reference: Levenson RM, Krupinski EA, Navarro VM, Wasserman EA (2015) Pigeons (Columba livia) as Trainable Observers of Pathology and Radiology Breast Cancer Images. PLoS ONE 10(11):e0141357. doi:10.1371/journal.pone.0141357

We would like to thank the Royal Pigeon Racing Association for their feedback on this blog post.