Towards A Illness-Sniffing Machine That Rivals A Canine’s Nostril

Trained dogs can detect cancer and other diseases by smell. A miniaturized detector can analyze trace molecules to mimic the process.

Numerous studies have shown that trained dogs can detect many types of diseases – including lung, breast, ovarian, bladder, and prostate cancers, and possibly Covid-19 – simply by smell. In some cases, such as prostate cancer, the dogs had a 99 percent success rate in detecting the disease by sniffing the patient’s urine samples.

But it takes time to train such dogs and their availability and time are limited. Scientists have been looking for ways to automate the amazing smelling abilities of a dog’s nose and brain in one compact device. A team of researchers at MIT and other institutions has now developed a system with which the chemical and microbial content of an air sample can be measured even more sensitively than a dog’s nose. They coupled this with a machine learning process that can identify the characteristic features of the disease-causing samples.

The results, which researchers say could one day lead to an automated odor detection system small enough to fit into a cellphone, are published today in PLOS One magazine in an article by Claire Guest of Medical Detection Dogs published in the UK. Research scientist Andreas Mershin from MIT and 18 others at Johns Hopkins University, the Prostate Cancer Foundation and several other universities and organizations.

“Dogs have proven to be the earliest and most accurate disease detectors for anything we’ve ever tried for about 15 years.Says Mershin. And their controlled test performance has in some cases outperformed the best current laboratory tests, he says. “So far, many different canine cancers have been discovered earlier than any other technology.

Additionally, the dogs appear to be making connections that human researchers have so far eluded: when they were trained to respond to samples from patients with one cancer, some dogs identified several other cancers – although the similarities between the samples weren’t for not visible to people.

These dogs can identify themselves “Cancer types that do not have identical biomolecular signatures in common, nothing in the odorous substances” Mershin says. With powerful analysis tools such as gas chromatography-mass spectrometry (GCMS) and microbial profiling “If you analyze the samples of skin cancer and bladder cancer, for example, and breast cancer and lung cancer – all things that the dog has been shown to detect – they have nothing in common. “Still, the dog can somehow generalize from one type of cancer in order to identify the other.

In recent years, Mershin and the team have developed and further developed a miniaturized detector system that contains mammalian olfactory receptors that are stabilized as sensors and whose data streams can be processed in real time with the capabilities of a typical smartphone. He envisions a day when an odor detector will be built into every phone, just as cameras are now ubiquitous in phones. Such detectors, equipped with advanced algorithms developed through machine learning, could potentially detect early signs of disease much earlier than typical screening regimes – and even warn of smoke or a gas leak.

In the latest testing, the team tested 50 urine samples from confirmed cases of prostate cancer and controls known to be disease free, using both dogs trained and handled by Medical Detection Dogs in the UK the miniaturized detection system. They then applied a machine learning program to find similarities and differences between the samples that could help the sensor-based system identify the disease. When testing the same samples, the artificial system was able to match the dogs’ success rates, with both methods achieving greater than 70 percent.

The miniaturized detection system is 200 times more sensitive than a dog’s nose, according to Mershin, because it can recognize and identify tiny traces of various molecules, which is confirmed by controlled tests prescribed by DARPA. But in terms of the interpretation of these molecules, “It’s 100 percent dumber.” This is where machine learning comes in to try to find the elusive patterns dogs can infer from smell, but humans couldn’t grasp from chemical analysis.

“The dogs don’t know any chemistry,” says Mershin. “You don’t see a list of molecules in your head. When you smell a cup of coffee, you don’t see a list of names and concentrations, you feel an integrated feeling. The dogs can reduce this feeling of smell. ”

While the physical device for detecting and analyzing the molecules in the air has been under development for several years with the main emphasis on reducing their size, the analysis has so far been absent. “We knew that the sensors were already better than what the dogs can in terms of detection limit, but what we haven’t shown before is that we can train an artificial intelligence to mimic the dogs.” he says. “And now we’ve shown that we can do it. We have shown that what the dog does can be recreated to a certain extent. “

This achievement, the researchers say, provides a solid framework for further research to bring the technology to a level suitable for clinical use. Mershin hopes to be able to test a far larger set of samples, perhaps 5,000, to pinpoint the significant indicators of disease. Such tests aren’t cheap, however: it costs about $ 1,000 per sample to collect, document, ship, and analyze clinically tested and certified samples of disease-transmitting and disease-free urine, he says.

When Mershin pondered how he got involved in this research, he recalled a study on the detection of bladder cancer in which a dog kept identifying a member of the control group as positive for the disease despite being selected as being on the basis of hospital tests be disease free. The patient who knew about the dog test decided to have more tests and a few months later it was found that the disease appeared very early. “Although it’s only one case, I have to admit that it influenced me,” says Mershin.

The team consisted of researchers from MIT, Johns Hopkins University in Maryland, medical detection dogs in Milton Keynes, UK, Cambridge Polymer Group, Prostate Cancer Foundation, University of Texas at El Paso, Imagination Engines and Harvard University. The research was supported by the Prostate Cancer Foundation, the National Cancer Institute, and the National Institutes of Health.