Get a Whiff of This

Can electronic noses come close to the real thing?

Sep 1, 2012
Kerry Grens

Artificial noses have orbited the Earth in spacecraft, inhaled in doctors’ offices, and sniffed in food-processing plants, all in an effort to surpass the sensitivity and specificity of the mammalian olfactory organ. Sure, dogs have been known to smell a cancer, but can they tell you what kind it is, too? Hossam Haick, a professor at the Technion-Israel Institute of Technology, has developed a device that can do just that. By collecting a breath sample from patients, Haick’s electronic nose can determine whether that person has lung cancer, as opposed to breast, prostate, or head and neck tumors, and even whether it’s non-small cell or small cell lung cancer.

Cancer patients emit a suite of volatile organic compounds in their breath that is different from the composition of healthy patients’ breath—and that differs from cancer to cancer. “If we develop an artificial nose which can detect very tiny amounts—at the range of parts per billion or parts per trillion—of these biomarkers of cancer, then we can provide a very simple and inexpensive way to detect cancer,” says Haick. “And most importantly, this is not invasive.”

Haick’s cancer sniffer is currently in clinical trials, but there are already some electronic noses on the market. Alpha MOS, a French company, markets electronic noses that are used in food and beverage quality control, in plastics and packaging manufacturing to detect contaminants, and in flavor and perfume development.

If we develop an artificial nose which can detect very tiny amounts of these biomarkers of cancer, then we can provide a very simple and inexpensive way to detect cancer.
—Hossam Haick, Technion-Israel Institute of Technology

Pretty much all electronic noses are based on the same approach, “sort of a fingerprint pattern recognition, like in the human process of olfaction,” says Alpha MOS spokesperson Marion Bonnefille. The mammalian sense of smell uses a combinatorial code composed of responses from different olfactory receptors. Rather than have an odor receptor specific to a particular odor, mammals interpret different scents by the pattern of receptors stimulated and the neural responses they excite. “In this way, 1,000 different receptors can recognize a million different odorants,” says Nate Lewis, a professor at CalTech and a pioneer in developing electronic noses.


Lewis’s own technology works through an array of tiny sensors made of polymer film that act like sponges. Each one responds to an odor slightly differently, and the amount of swelling of the “sponge” in the presence of a vapor changes its electronic resistance. The pattern of resistance changes is distinct for each odor, giving the electronic nose the ability to distinguish between good wine and bad wine, toluene and benzene, and even between mirror images of the same molecule. Lewis has been able to detect compounds diluted down to the tens of parts per trillion.

But there is a limit to the seemingly endless uses for artificial noses. “What we are not good at . . . is [breaking] down a complex mixture into hundreds of different compounds,” says Lewis. Gas chromatography-mass spectrometry, and the human nose to some degree, can tell you the specific composition of a sample, whereas an e-nose can only tell you whether or not the sample matches a particular profile. “It would be very good to know what are the biomarkers found inside [a breath sample], but this would require further studies and further redevelopment of the device,” says Haick.

Perena Gouma of the State University of New York, Stony Brook, has made artificial noses that can detect particular components of a person’s breath. “We have arrays of sensors, each of which can target a specific biomarker . . . or class of chemicals,” she says. In such a way, her lab is developing gadgets to measure the likelihood of? having a health condition. Whereas the electronic nose uses the overall differences between healthy breath and diseased breath to distinguish between them, Gouma’s approach requires knowledge of the particular biomarkers in advance so that sensors can be developed specifically to detect them. (See “Vital Signs,” The Scientist, August 2011.)

The inability to describe the composition of a sample aside, Tufts University chemist David Walt says that during his research on electronic noses “we pretty much couldn’t find a problem that we couldn’t solve.” That is, except for one very big challenge: for each problem an e-nose can solve, say, to distinguish spoiled milk from fresh or safe packaging from contaminated packaging, “every one of those different problems is a training and then validation that can eat up a huge amount of money,” Walt says. Additionally, recalibrating each device might be a massive undertaking if it requires, for example, bringing in numerous patients with a certain disease. “That’s not trivial.”

Walt gave up on researching electronic noses several years ago because he could not find investors to move the technology into the real world, presumably because of these challenges. But he is somewhat optimistic for those who remain in the field—if they can figure out how to make it easier to train the devices to recognize signals of particular odors. “Lots of opportunities are there,” he says.