Skip to Content

Cancer Breathalyzer

SUNY researchers are working on a small, easy-to-use disease detector.
March 20, 2006

A little over a year ago, a British Medical Journal study showed that dogs could smell cancer. And this month a study in the journal Integrative Cancer Therapies showed similar results. Both articles provoked headlines and quite a bit of skepticism.

They also brought attention to the odors, or volatile organic compounds (VOCs), that the human body exudes in its day-to-day functioning and during a disease. These compounds, big and small, are found in human breath, and researchers at the State University of New York at Buffalo are building a chemical sensor that will examine a person’s breath to detect diseases. “It’s tantamount to trying to measure the surface of the earth, from Mount Everest…all the way down to the height of a raisin, and be able to see it all in detail,” says Frank Bright, professor of chemistry at SUNY Buffalo, who’s leading the research team.

Past research in this field has linked specific combinations of chemicals to specific diseases, Bright says. For example, acetone, ethane, and hydrogen peroxide are associated with diabetes, and mono-methylated alkanes with breast cancer. These compounds can be detected using expensive, bulky machines, such as gas chromatographs and mass spectrometers.

In contrast, the Buffalo device is designed to be a tube the size of a roll of quarters, containing an array of xerogel sensors and a complementary metal oxide semiconductor (CMOS) detector. “Everything can be integrated on a silicon chip with processing circuitry, to make things more compact, less expensive, and lower power,” says Albert Titus, a professor of electrical engineering at SUNY Buffalo who’s working on the sensor.

Xerogels are glass-like materials with nanoscopic pores into which the researchers can infuse tiny, chemically sensitive fluorescent dye molecules. Using pin printing, a genomics technique, they can deposit hundreds of these molecules, each about 10 micrometers in size, on top of a light-emitting diode, which stimulates the sensors to emit light. In the presence of a targeted VOC, the fluorescent molecules respond in a particular way, such as by emitting light with a certain color or intensity.

Ideally, one xerogel sensor would detect one chemical compound, Bright says. But since many different compounds are made of the same elements, or can have similar chemical structures, a sensor could make mistakes in attempting to differentiate between two similar compounds. To deal with that, the researchers are making many sensors that respond to the same VOC in different ways, so if one makes a mistake, most of the others will likely identify a VOC correctly. Although other people are working on similar devices, no one has anywhere near as many sensing elements, Bright says. So far, the Buffalo team has built sensors that respond to some 100 different chemicals.

The CMOS detector will convert the sensors’ optical patterns into electrical signals, which will be analyzed by pattern-recognition software – similar to that used in recognizing handwriting. In this way, the system will identify different chemicals based on their electrical signal patterns.

But the identification of chemicals goes only so far in detecting disease, because a given chemical is often associated with different diseases. For instance, acetone is associated with both diabetes and early-stage lung cancer. Rather, it’s the combination of many different compounds that points to one particular disease.

In order to get information about such combinations, the researchers plan to employ a form of neural networks, which is software that learns from experience. Collaborating with researchers at the Roswell Park Cancer Institute in Buffalo, they will ask patients diagnosed with various cancers to blow into a prototype sensor. Then they’ll use this data to train the neural-network software to recognize chemical patterns present in an individual with lung cancer, for instance, and to distinguish those patterns from ones in the breath of a person with breast cancer.

In the end, the pattern-recognition software will be able to recognize specific chemicals, while the neural networks will identify the pattern of a mixture of chemicals. This makes their sensor novel, says Peter Mazzone, who has studied the detection of lung cancer with chemical sensors at the Cleveland Clinic in Ohio. Most chemical sensors respond to a variety of chemicals, he says. They can detect patterns of chemicals, but cannot identify individual chemicals in the mixture – that typically requires gas chromatography and mass spectrometry (GC-MS). But the SUNY sensor system “seems to be able to detect patterns of chemicals and to identify the chemicals themselves,” he says. “If [it] can do this without GC-MS, then it would be a step forward in this field.”

The Buffalo sensor falls into a category of devices known as “electronic noses,” commonly used in the food industry and environmental monitoring. Many of them measure the change in resistance across a thin polymer film when it absorbs a targeted chemical, as opposed to measuring optical change, says Amy Ryan, who conducts research into electronic noses at NASA’s Jet Propulsion Laboratory. While she hasn’t seen xerogels applied to disease sensing in the past, she says, since pattern recognition software has been used for visual purposes, there’s no reason xerogels and pattern recognition shouldn’t work together.

Bright says that all the individual parts of their sensor system are in place and that the prototype should be ready for testing within a year. He believes the xerogel sensors offer the advantages of robustness and stability, as well as being easy to apply to different chemicals. Yet the CMOS chip’s detection limits need more work to be able to detect subtle differences in the sensors’ emissions, he says, and the engineering team is investigating different ways to process the signals.

If the device works as the researchers intend, it should be useful for screening diseases early on, to be followed by a more thorough medical examination. As a dog’s judgment could never serve as the final say in a person’s medical condition, so likewise the sensor would not be used for making definitive judgments. “An electronic nose is not an analytical instrument,” Ryan says. “But it’s an easy, relatively inexpensive, noninvasive tool which can be used for lots of different kinds of initial screening.”

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.