Towards a device that sniffs a disease comparable to a dog’s nose

Research schema for dog olfactory testing. (A) Two dogs, Florin and Midas, selected to participate in the study. (B) Image of the presentation pot. (C) A test pot placed on a metal arm attached to the carousel. (D) Comparison of indications for biopsy-negative controls and cancer samples in a double-blind study. In this table, of the 21 control samples, Florin showed 5 false-positive signs with 76.2% specificity, while Midas’ 6 false-positive signs showed 70% specificity. It shows that. Both dogs correctly showed 5 of the 7 target samples giving a sensitivity of 71.4. Credits: Guest et al, 2021 (PLOS ONE, CC-BY 4.0, creativecommons.org / licenses / by / 4.0 /)

Numerous studies have shown that trained dogs can detect many types of diseases, including lung, breast, ovarian, bladder, prostate, and possibly COVID-19, by smell alone. I will. In some cases, including prostate cancer, for example, dogs had a 99 percent success rate in detecting the disease by sniffing the patient’s urine sample.

But such training takes time dog, And their availability and time are limited. Scientists have sought a way to automate the amazing olfactory function of the dog’s nose and brain with a compact device. Currently, a team of researchers at MIT and other institutions have devised a system that can detect the chemical and microbial content of air samples with even higher sensitivity than dog noses. They combined this with a machine learning process that could identify the characteristic characteristics of diseased samples.

Research results, researchers say, could one day lead to an automatic odor detection system that’s small enough for researchers to incorporate into mobile phones, are published in the journal today. PLoS ONEIn 18 papers by Clare Guest of Medical Detection Dogs in the United Kingdom, Andreas Mershin, a research scientist at MIT, and Johns Hopkins University, the Prostate Cancer Foundation, and several other universities and organizations.

“Dog has been shown to be the fastest and most accurate disease detector for everything we’ve tried so far for about 15 years.” Mershin says. Also, performance on managed tests can exceed the performance of the best lab tests today, he says. “So far, different types cancer Detected in dogs faster than any other technique. “

In addition, dogs are clearly picking up connections that have previously avoided human researchers. When trained to respond to samples from patients with one type of cancer, some dogs identified several other types of cancer, even if there were no similarities between the samples. It is not obvious to humans.

These dogs can identify “cancers that do not have the same identical biomolecular signature in common and have nothing in the odorant,” Mershin says. Using powerful analytical tools such as gas chromatography-mass spectrometry (GCMS) and microbial profiling, “for example, analyzing samples of skin and bladder cancer, breast and lung cancer has shown that dogs can do it. Detect — nothing in common. ”Still, dogs can generalize from one type of cancer to identify other cancers.

Mershin and his team have continued to develop and improve a small detector system that incorporates a mammalian olfactory receptor that is stabilized to act as a sensor. The data stream can be processed in real time by the functions of general smartphones. He envisions the day when all mobile phones will incorporate scent detectors, just as cameras are as popular in mobile phones. Such detectors, with advanced algorithms developed through machine learning, can detect early signs of illness much faster than normal screening regimes, and can even warn of smoke and gas leaks. He says there is.

In the latest test, the team tested 50 samples of urine from confirmed cases. Prostate cancer Controls known to be disease-free using both dogs trained and handled by Medical Detection Dogs in the United Kingdom and a miniaturized detection system. We then applied a machine learning program to draw similarities and differences between samples where sensor-based systems could help identify the disease. When the same sample was tested, the prosthesis was able to match the success rates of dogs, with both methods scoring over 70%.

The miniaturized detection system is actually more than a dog’s nose in that it can detect and identify small traces of various molecules, as confirmed by the controlled tests mandated by DARPA. Mershin says it is 200 times more sensitive. But in terms of interpreting those molecules, “it’s 100 percent annoying.”that is Machine learning Dogs try to find elusive patterns that can be inferred from the scent, but humans have not been able to figure them out from chemical analysis.

“Dog doesn’t know chemistry,” says Mershin. “They don’t see a list of molecules in their heads. When they smell a cup of coffee, they don’t see a list of names and concentrations and they feel an integrated sensation. The sensation of that scented character. Is dog’s and can be mined. “

Physical devices for detecting and analyzing molecules in the air have been developed for several years with a focus on reducing their size, but until now analysis has been lacking. “We knew that sensors were already better than dogs could do in terms of detection limits, but what we’ve shown so far is that we can train artificial intelligence to imitate dogs,” he says. .. “And now we have shown that we can do this. We have shown that what dogs do can be reproduced to some extent.”

Researchers say the results provide a solid framework for further research to develop the technology to a level suitable for clinical use. Mershin hopes to test far more samples, perhaps 5,000, to identify key indicators of disease in more detail. However, such tests are not cheap. He says it costs about $ 1,000 per sample to collect, document, ship, and analyze clinically tested and certified disease-free urine samples.

Looking back on how he was involved in this study, Marcin recalled his study on the detection of bladder cancer. In this study, dogs continued to misidentify one member of the control group as positive for the disease, despite being specially selected based on hospital tests. There is no illness. Patients who knew about testing dogs chose to have more tests and, months later, found that they had the disease very early on. “It’s just one case, but I have to admit that it rocked me,” says Mershin.