No matter how far detection science advances, it seems like we’re still ultimately just relying on dogs. Finding bombs? Dogs. Searching for missing people? Dogs. Identifying cancer? Dogs. We’ve even reached a point where, though we might develop new technologies for such tasks, we’re basing these technologies entirely upon what we know about dogs.
Case in point: Researchers have been working to train dogs to detect aggressive forms of prostate cancer in human urine samples, hoping that an artificial neural network could be developed to mimic their abilities.
With our current prostate cancer detection abilities, we’re in a bit of a bind: One commonly used method, the prostate specific antigen (PSA) screening test, often misses aggressive cases of prostate cancer or falsely identifies it in those who don’t have it at all. Meanwhile, a pilot study from Medical Detection Dogs in Milton Keynes, U.K., the results of which were published in PLOS in February, found that dogs can be trained to detect these aggressive cases by smelling urine samples with 71 to 76 percent accuracy. Not perfect, but not terrible either.
In any case, it’s good enough that researchers then used that data to create an artificial neural network capable of detecting the specific compounds in urine that influenced the dogs’ diagnoses. So in the future, prostate cancer might be better detected using software essentially designed to work like a dog’s sense of smell.
Of course, this isn’t at all the first time we’ve used dogs for such purposes. Even for COVID-19, trainers have attempted to teach dogs how to identify the virus, though it doesn’t appear this has yet been very fruitful. Dogs have also accurately identified cancer in human breath and blood samples, and they’ve reportedly brought attention to lesions on their owner’s body that ended up being cancerous, too.
Considering that smelling someone’s breath or urine is far less invasive than, say, performing a colonoscopy or drawing blood, detection for specific cancers could become much more easily available should we have enough dogs capable of managing it. But since that’s not very likely, developing technology that works similar to dogs is the next best thing.