Dogs are teaching machines to sniff out cancer | Digital Trends Spanish


With their legendary sense of smell, dogs are adept at identifying the characteristic scents of cancer in breath, urine and feces.

A study published in the scientific journal PLOS One has shown how these animals can “train” an artificial intelligence to detect aggressive prostate cancers.

Prostate cancer affects one in nine men, making it the second most common type of cancer in this population.

The prostate specific antigen (PSA) test is one of the most widely used to identify it, but it often fails to do so or misdiagnoses.

While the ability of dogs to recognize cancers has been successfully tested since the 1990s, it is not entirely practical for large-scale procedures.

For this reason, a team of researchers from Medical Detection Dogs from the United Kingdom proposed combining the strengths of canine smell and detection methods based on artificial intelligence, with the aim of developing an application for smartphones.

Teach the machines

Pixabay

For the study, two dogs had to identify different types of cancer from urine samples. According to the results, they successfully recognized 71 percent of the positive cases, and between 70 and 76 percent of the negatives.

The researchers collected the volatile scent compounds from each of the urine samples and discovered distinctive chemicals. Using this data, they trained an artificial neural network to identify the portions of the spectroscopy that contributed to the dogs’ diagnoses.

The final goal of the team is to apply the algorithm to “an electronic nose”, which allows to bring this discovery to smartphone applications. “We have shown that it is possible to replicate dog performance as sensors and brains,” he said. Claire Guest from the Medical Detection Dogs.

Before this tool is ready for smartphones, scientists acknowledge that more samples are needed to improve detection. Either way, the preliminary results are already promising.

Editor’s Recommendations






You may also like...

Leave a Reply