Can a Breath Test Detect Disease? Electronic Noses and Breath Tests Explained

Breath tests can already detect several conditions reliably, and doctors use them every day. They check for H. pylori stomach infection (the urea breath test), lactose intolerance and gut bacterial overgrowth (hydrogen and methane tests), airway inflammation in asthma (FeNO), and of course alcohol. Electronic noses go further. These sensor arrays, paired with machine learning, try to recognize the mix of chemicals in breath linked to cancers, lung disease, Parkinson's and infections. A widely reported 2016 study trained one device to tell 17 diseases apart with about 86 percent accuracy. Electronic-nose disease tests remain mostly research tools, though, not products you can buy for a diagnosis. Use breath tests your doctor orders, and treat consumer "health breathalyzers" with caution.

Ethan's father had a breath test at his doctor's office for a stomach complaint: he drank a small liquid, breathed into two bags half an hour apart, and a day later learned he had an H. pylori infection that a course of antibiotics would clear. That same week, Ethan saw an old headline about a device that "detects 17 diseases from your breath" and asked Jake whether he could buy one. Jake, who reads more about sensors than most people, gave him the honest answer: some breath tests are routine medicine, the 17-disease nose is real research, and the gap between the two is where most of the hype lives. This page lays out that gap, so you know which breath tests to trust, how electronic noses work, and what to ask before relying on one.

This page explains technology and research for general understanding. It is not medical advice; for any symptom or test result, speak with a qualified health professional.

Why breath can reveal disease

Every breath you exhale carries more than air. Alongside oxygen, carbon dioxide and water vapor, it holds hundreds of volatile organic compounds, or VOCs, at tiny concentrations, often parts per billion. Some come from what you ate, drank or breathed in. Others are produced by your body's own chemistry: cells breaking down fats, bacteria in the gut digesting sugars, inflamed airways making nitric oxide. When disease changes that chemistry, the mix of compounds in breath can change too.

Physicians have known this for a very long time. Doctors once noted the fruity, acetone smell on the breath of people with uncontrolled diabetes, the ammonia-like smell of kidney failure and the musty smell associated with severe liver disease. What has changed is the ability to measure those compounds precisely, and to look not at one chemical but at the pattern of many at once. That is the idea behind both modern breath tests and electronic noses.

Breath has obvious appeal as a sample. It is painless to give, available in unlimited amounts, and can be collected anywhere, without needles, labs or specialist staff. The challenge is that it is also easily influenced by things that have nothing to do with disease, which is the theme of much of this page.

Compounds researchers look for

A few breath compounds come up again and again in research, because their links to body chemistry are well understood:

• Acetone: rises when the body burns fat for fuel, as in fasting, low-carbohydrate diets and uncontrolled diabetes.
• Ammonia and related amines: can rise with kidney problems, when the body clears nitrogen waste less effectively.
• Sulfur compounds: linked to severe liver disease, and also to bad breath from the mouth itself.
• Isoprene: related to cholesterol production and affected by exercise.
• Light hydrocarbons such as pentane and ethane: studied as markers of oxidative stress, when cell membranes are damaged.
• Nitric oxide: produced by inflamed airways, the basis of FeNO.

None of these is unique to one disease, which is why researchers look at combinations and patterns rather than any single compound.

Diabetes and breath: what is real

Breath acetone and diabetes are often mentioned together, with some reason. When insulin is severely lacking, the body breaks down fat rapidly and produces ketones, including acetone, which can give breath a fruity smell; in extreme cases this signals diabetic ketoacidosis, a medical emergency. Researchers have studied whether breath acetone could track blood sugar or ketones without finger pricks.

The reality so far is that breath acetone relates to ketone levels more closely than to blood sugar, and varies with diet, fasting and exercise. Blood glucose meters and continuous glucose monitors remain the tools for managing diabetes, and blood or urine ketone tests are used when ketones matter. Anyone with diabetes who notices fruity breath with nausea, vomiting or confusion should seek urgent medical care rather than rely on any breath device.

Breath tests doctors already use

Notice what these have in common: each measures one specific compound linked to one specific process, usually after a controlled step such as a drink or test meal. That design makes them reliable. Electronic noses take a different and much harder approach, trying to read a disease from the whole chemical pattern of unprepared breath.

The urea breath test for H. pylori

Helicobacter pylori is a bacterium that lives in the stomach lining of a large share of the world's population. Many people never notice it, but it is a major cause of stomach ulcers and is linked to stomach cancer. The urea breath test is one of the standard ways to detect it, and to confirm it is gone after treatment.

1. You fast for a few hours beforehand, and your doctor may ask you to pause certain medicines, such as acid-reducing drugs and antibiotics, for a period before the test, because they can cause false negatives.
2. A baseline breath sample is collected.
3. You swallow a small dose of urea labeled with a carbon isotope, usually non-radioactive carbon-13.
4. If H. pylori is present, its enzyme, urease, splits the urea and releases labeled carbon dioxide, which passes into the blood and out through the lungs.
5. A second breath sample, usually after about 15 to 30 minutes, is analyzed for labeled carbon dioxide.

The test is accurate, painless and avoids an endoscopy for many patients. It is exactly the kind of breath test Ethan's father had, and a good example of how well breath testing can work when it targets one known chemical process.

What a clinic breath test appointment is like

1. The clinic gives preparation instructions, which may include fasting, diet changes and pausing certain medicines; follow them exactly, and ask before stopping any prescribed medicine.
2. You arrive rested; smoking, vigorous exercise and chewing gum are usually not allowed beforehand.
3. A baseline breath sample is taken, often by breathing through a straw into a bag or tube.
4. You drink the test solution, if the test uses one.
5. Further samples are collected at set times, from one more sample to one every 15 to 30 minutes for a few hours.
6. Samples are analyzed on site or sent to a lab, and your doctor explains the result.

Longer tests mean sitting in the clinic for hours, so bring something to read. The tests are not painful; the hardest part for most people is the fasting.

If anything about the preparation is unclear, call the clinic beforehand; a test done without proper preparation often has to be repeated.

Hydrogen and methane tests for gut problems

People with lactose intolerance cannot fully digest the sugar in milk. The undigested sugar reaches the colon, where bacteria ferment it and produce hydrogen, some of which is absorbed into the blood and breathed out. A hydrogen breath test measures that: you drink a set amount of lactose, then give breath samples every 15 to 30 minutes for two to three hours. A clear rise in hydrogen suggests the sugar was not absorbed.

The same principle, with different test sugars such as lactulose or glucose, is used to look for small intestinal bacterial overgrowth, where bacteria normally found in the colon multiply higher up in the gut. Many tests also measure methane, because some people's gut microbes turn hydrogen into methane, which would otherwise hide a positive result. Preparation matters: a special low-fiber diet the day before, fasting, and avoiding recent antibiotics, all as directed by the clinic.

Results are read as a curve over time rather than a single number, and interpretation follows clinical criteria, so a home gadget's single reading is no substitute for the full test.

FeNO: measuring airway inflammation

Fractional exhaled nitric oxide, or FeNO, measures nitric oxide in exhaled breath, in parts per billion. Inflamed airways of a particular kind, the allergic or "type 2" inflammation common in asthma, produce more nitric oxide. A FeNO test takes a few minutes: you breathe out slowly and steadily into a handheld device for about ten seconds.

Doctors use FeNO alongside breathing tests to support an asthma diagnosis, to judge whether inhaled steroids are likely to help, and to check whether treatment is controlling inflammation. It does not diagnose asthma by itself, since smoking, recent meals, infections and some medicines affect the reading, but it adds objective evidence to a condition that is otherwise judged largely on symptoms.

FeNO devices are small enough for a family doctor's office or a specialist clinic, and some health systems now use them routinely when assessing breathing problems in adults and children old enough to follow the breathing instructions.

Breathalyzers: the most familiar breath test

A breathalyzer estimates blood alcohol from alcohol in breath. Alcohol in the blood passes into the air in the lungs in a fairly predictable ratio, so measuring breath alcohol gives an estimate of the blood level. Police devices typically use fuel-cell sensors or infrared spectroscopy, which are specific to alcohol and accurate when properly calibrated. Cheaper consumer devices often use semiconductor sensors, which can react to other substances and drift over time, so their readings should not be relied on to decide whether it is safe to drive.

Breathalyzers also show the practical rules that apply to any breath test: wait after eating, drinking or using mouthwash, since alcohol in the mouth gives falsely high readings, and calibrate regularly.

Bad breath and the halimeter

The most common reason people worry about their breath is not disease detection but bad breath, or halitosis. In most cases it starts in the mouth, from bacteria on the tongue and between teeth that produce volatile sulfur compounds. Dentists sometimes use a portable meter, often called a halimeter, that measures those sulfur compounds to judge how severe bad breath is and whether treatment is working.

Persistent bad breath despite good brushing, flossing and tongue cleaning is worth mentioning to a dentist or doctor, since it can occasionally point to sinus, throat, stomach or other problems. But for the vast majority of people, the answer is dental hygiene, not a disease breathalyzer.

How an electronic nose works

An electronic nose, or e-nose, copies the way a biological nose works rather than the way a laboratory instrument does. Your nose does not have one receptor for each smell; it has hundreds of receptor types that each respond partly to many molecules, and your brain recognizes a smell from the overall pattern. An e-nose does the same with an array of chemical sensors.

1. Sensor array: a set of sensors, often between a handful and a few dozen, each made of a material that responds slightly differently to different compounds. Common types include metal-oxide sensors, conducting polymers, and nanomaterials such as gold nanoparticles coated with organic molecules.
2. Sample: breath is blown into a bag or directly into the device, sometimes after being concentrated, and passed over the sensors under controlled temperature and humidity.
3. Response pattern: each sensor's electrical resistance or another property changes, giving a "fingerprint" of numbers for that breath.
4. Pattern recognition: machine learning compares the fingerprint with fingerprints from people known to have, or not have, a condition, and outputs a classification or probability.

The key point is that an e-nose usually does not identify individual chemicals. It recognizes patterns, which makes it cheap and fast but also means its accuracy depends entirely on the quality of the data it was trained on.

Breath-testing terms in plain words

• VOC: volatile organic compound, a carbon-based chemical that evaporates easily.
• Breathomics: the study of all the compounds in breath and what they reveal.
• E-nose: an electronic nose, a sensor array with pattern recognition.
• GC-MS: gas chromatography with mass spectrometry, the lab reference method.
• ppb: parts per billion, the concentration scale for many breath compounds.
• End-tidal breath: the last part of an exhalation, from deep in the lungs.
• Validation: testing a model on people it was not trained on.

Knowing these few terms makes research summaries and product claims much easier to judge. When a description uses none of them and talks only about "AI that smells disease", treat it as marketing until you see the details.

How breath is collected for analysis

How a sample is taken can change the result as much as the disease does. Breath from the start of an exhalation is mostly air from the mouth and windpipe; the last part, called end-tidal or alveolar breath, comes from deep in the lungs, where blood and air exchange compounds. Many studies capture only that last part. Samples are collected into inert bags, onto sorbent tubes that trap compounds for later lab analysis, or straight into a device.

Room air matters too: a hospital corridor smells of disinfectant, and some of those compounds end up in every breath sample taken there. Careful studies sample the room air at the same time and subtract it, ask participants to rest and fast beforehand, and record smoking, diet and medicines, so the model learns the disease rather than the setting.

The machine learning behind e-noses

Turning sensor readings into a diagnosis is a classic supervised learning problem. Researchers collect breath from people with a confirmed diagnosis and from comparison groups, record each sensor's response, and train a model to separate the groups. Older studies used statistical methods such as discriminant analysis and principal component analysis; newer ones use support vector machines, random forests and neural networks. Our introduction to AI and machine learning explains these ideas without the math.

The same weaknesses that affect any machine learning project apply with extra force. Breath studies often have a few hundred participants but dozens of sensor features, which makes overfitting easy: a model can learn quirks of one hospital's patients rather than the disease. Careful projects therefore test their model on people it never saw during training, ideally from a different hospital, and report accuracy only on that independent group. Our guide to cleaning data and handling outliers covers why messy inputs matter so much.

One practical consequence: an e-nose trained in one hospital may need retraining for another, because the equipment, room air and patient mix differ. Methods for detecting unusual readings, such as anomaly detection, help flag samples that do not fit what the model learned, rather than forcing a confident but wrong answer.

The 2016 "17 diseases" study

The headline this page originally covered came from a study by a team at the Technion, Israel Institute of Technology, led by Professor Hossam Haick, published in the journal ACS Nano in December 2016. The team collected breath from 1,404 people across several countries, covering 17 different conditions, including lung, colorectal, gastric, head-and-neck, ovarian, kidney and prostate cancers, Parkinson's disease, multiple sclerosis, Crohn's disease, ulcerative colitis, chronic kidney disease, pulmonary hypertension and pre-eclampsia, plus healthy controls.

They used an array of nanomaterial sensors, often called the Na-Nose, combined with pattern-recognition software, and reported that the system distinguished the conditions from one another with about 86 percent accuracy. They also analyzed samples by gas chromatography and mass spectrometry, finding that each disease had a distinct combination of compounds, with 13 compounds contributing to the patterns.

It was an important proof of concept: breath patterns differ across many diseases, and a single type of sensor array can learn to separate them. It was not a product, a home test or a replacement for medical diagnosis, and the researchers presented it as a step toward future screening tools.

Why breath-test headlines oversell

News stories about breath tests tend to follow a pattern: a striking number, such as "detects 17 diseases" or "90 percent accurate", a photo of a small device, and a suggestion that home testing is around the corner. The study behind the story is usually more careful, describing a research prototype tested on selected patients. The gap comes from compressing years of work into one line, and from comparing clearly sick patients with healthy volunteers, which is far easier than finding early disease in ordinary people.

The original version of this page fell into the same trap with its exclamation-filled headline. The research deserved attention; it did not deserve the implied promise.

Why e-noses are not in every clinic yet

Ten years after that study, electronic noses are still mainly in research, for reasons that apply across the field:

• Confounders: food, coffee, smoking, alcohol, medicines, exercise, time of day and even the air in the room change breath chemistry.
• Sensor drift: chemical sensors change with age, temperature and humidity, so the same breath can give different readings months apart.
• Collection differences: how hard and how long someone breathes, and which part of the breath is captured, matter.
• Small, single-center studies: many report impressive accuracy that falls when tested on new groups.
• Specificity: many diseases share inflammation-related compounds, so separating one disease from another, rather than sick from healthy, is hard.
• Regulation: a diagnostic device must prove safety and accuracy to regulators, which takes large, costly trials.

Reviews pooling dozens of e-nose studies have found high average accuracy, often around 90 percent sensitivity and specificity, but they also note that most included studies were small and few were validated independently. High numbers in early studies are a reason for interest, not yet a reason for clinical use.

Lung cancer and nodules: a practical target

Lung disease is a natural focus for breath research, since the breath comes straight from the lungs. For lung cancer, the established screening tool is low-dose CT scanning for people at high risk, such as long-term smokers in certain age groups. CT finds many small spots, called nodules, most of which are not cancer, and deciding which need further tests is a real clinical problem.

Several research groups are testing whether breath analysis can help sort nodules into likely benign or likely malignant, reducing unnecessary procedures. That is the kind of narrow, practical question where a breath test could add value alongside existing tools, rather than trying to replace them. Anyone eligible for lung cancer screening should follow their doctor's advice on CT screening; breath tests are not a substitute.

Breath research also covers chronic obstructive pulmonary disease and asthma flare-ups, where tracking changes over time in one person may prove more useful than comparing them with others.

What accuracy numbers really mean

Headlines often quote one accuracy figure, but medical tests are judged by several:

Predictive values depend heavily on how common a disease is. Imagine a screening test with 90 percent sensitivity and 90 percent specificity, used on 1,000 people where 1 in 100 has the disease. It catches about 9 of the 10 people who are ill, but it also wrongly flags about 99 of the 990 who are healthy. So only around 1 in 12 positive results is real. That is why a test that sounds accurate in a study comparing patients with healthy volunteers can produce many false alarms when used to screen the general public.

How good studies avoid fooling themselves

Because breath data is so easy to over-interpret, the best studies follow a few rules. They decide in advance what they will measure and how they will judge success. They keep the people analyzing samples unaware of who has the disease, so expectations cannot creep in. They match patients and comparison groups on age, smoking and other factors. They collect samples the same way, at the same places, for both groups. And they test the final model on a separate group, ideally from a different hospital, reporting that result as the headline.

When you read about a breath test, look for those signs. A study of a few dozen patients compared with healthy volunteers from the same lab is an early clue; a multicenter trial with independent validation is real evidence.

Funding and authorship are worth a glance too: studies run entirely by a device's developer are common and legitimate, but independent replication carries more weight.

Breath tests for COVID-19

The pandemic gave breath testing a real-world trial. In April 2022, the U.S. Food and Drug Administration granted emergency use authorization to a COVID-19 breath test that used gas chromatography and mass spectrometry to detect compounds associated with infection, with results in a few minutes. Several countries tested similar devices at airports and events. These tests showed that breath analysis could be deployed at scale, while also showing the usual trade-offs: positive results needed confirmation with a standard molecular test, and performance varied with variants and settings.

Separately, trained detection dogs were studied for spotting COVID-19 by smell, with some promising results, which underlines the underlying idea that disease changes body odor in ways a sensitive enough nose, biological or electronic, can detect.

Those deployments also produced useful lessons for later breath tests: how to train non-specialist staff, how to manage sample throughput, and how to communicate that a quick test is a screening step rather than a final answer.

What sniffer dogs teach us about disease odor

Dogs have roughly hundreds of millions of scent receptors, far more than humans, and have been trained in studies to detect certain cancers, low blood sugar in people with diabetes, and infections, from breath, skin or urine samples. The results vary widely between studies and dogs, and dogs tire, need constant training and cannot be standardized like a device. But they prove that disease odors exist and can be detected, which is exactly what electronic nose researchers are trying to reproduce with sensors.

A famous human example points the same way: a Scottish woman with an unusually keen sense of smell noticed a change in her husband's odor years before his Parkinson's diagnosis, and researchers later traced the smell to compounds in skin oil. Each such case adds to the evidence that the chemistry is there to be read.

Dogs also illustrate the training-data problem that e-noses face: a dog trained on samples from one hospital can learn that hospital's smell rather than the disease, a mistake researchers have had to design studies carefully to avoid.

Lab methods: GC-MS and real-time mass spectrometry

Electronic noses recognize patterns; laboratory instruments identify individual compounds. Gas chromatography with mass spectrometry, GC-MS, is the reference method: it separates the compounds in a breath sample and identifies each by its mass fingerprint. It is precise but slow and requires a lab. Newer techniques such as proton-transfer-reaction and selected-ion flow-tube mass spectrometry analyze breath in real time, at the cost of large, expensive machines.

Researchers often use both approaches together: lab methods to discover which compounds matter for a disease, and cheaper sensors to detect those compounds or their pattern in practical devices. Several companies offer standardized breath collection kits and lab analysis for clinical research, which helps different studies compare results.

Electronic noses outside medicine

The same technology is already used in industry, where conditions are easier to control. Electronic noses check food freshness and spoilage, grade coffee, tea, wine and olive oil, monitor odors from factories and waste sites, and detect leaks of gases. Related sensors in security screening look for traces of explosives. In these settings, the sample, temperature and humidity can be standardized, and a wrong answer is less serious than a missed diagnosis, which is why e-noses succeeded there first.

Experience from industry, such as methods to correct sensor drift and recalibrate devices, feeds back into medical research, helping breath devices become more stable over time.

Infections: a promising area

Infections are one of the more promising targets for breath testing, because bacteria and viruses produce their own distinctive compounds. Researchers are studying breath tests that could help tell bacterial from viral infections, which would guide antibiotic use, and tests for lung infections such as pneumonia and tuberculosis, where a quick, cheap screening tool would be especially valuable in places with few labs. As with other uses, results so far come from studies, and confirmation with standard tests remains essential.

For low-resource settings, the attraction is cost: a reusable sensor that needs no reagents or lab could screen many people in a day, sending only likely cases for confirmatory testing.

Consumer breath devices: what to trust

A few breath devices are sold directly to consumers. Personal breathalyzers estimate alcohol, with accuracy depending heavily on the sensor type and calibration. Breath acetone meters are marketed to people following low-carbohydrate diets to estimate ketosis; they can indicate trends but are not a substitute for medical ketone testing, and they cannot diagnose diabetes or its complications. Hydrogen breath gadgets claim to help people track food intolerances; they measure real gases, but interpreting them without clinical preparation and guidance is unreliable.

Be wary of any consumer device or app claiming to detect cancer, infections or serious diseases from breath. As of this writing, such claims have not been backed by the large validation studies and regulatory approvals that diagnosis requires. A false reassurance can delay care, and a false alarm causes needless worry. If you have symptoms, the right breath test is the one your doctor orders.

The most useful consumer breath device for most households remains a well-calibrated fuel-cell breathalyzer, used as a rough guide, never as permission to drive.

Breath tests for drugs other than alcohol

Companies and researchers have worked on breath tests for cannabis and other drugs, aiming to give police a roadside tool like the alcohol breathalyzer. The chemistry is much harder: drug levels in breath are far lower, and, for cannabis in particular, the link between the level in breath and actual impairment is much less clear than for alcohol. Some devices are in trials, but as of this writing none is in broad use in the way alcohol breathalyzers are. Treat claims about drug breathalyzers with the same caution as disease breathalyzers.

Breath data and privacy

Breath tests generate health data, and consumer devices often send readings to apps and cloud services. In the United States, health privacy rules such as HIPAA cover doctors, hospitals and insurers, but generally not consumer gadgets and apps you buy yourself. Before using a connected breath device, read what data it collects, where it is stored, whether it is shared or sold, and whether you can delete it. The same caution applies to any AI-powered health trend; our guide to AI trends and your personal data covers the wider risks.

In research studies, consent forms explain how samples and data are used and stored. Read them, and ask whether your data may be shared with companies developing commercial devices.

Where breath testing is heading

Research is moving on several fronts at once. Sensor makers are improving stability and reducing drift; researchers are standardizing how breath is collected so studies can be compared; and larger multicenter trials are testing e-noses for specific, practical questions, such as whether a lung nodule is likely to be cancer or whether an infection is bacterial. Regulators have begun to engage too: in 2026, a breath-based test for volatile compounds received the U.S. FDA's Breakthrough Device designation, a status that speeds review for promising devices but is not an approval.

The likely path is not one breathalyzer for every disease, but targeted breath tests that answer narrow clinical questions well, used alongside existing tests. That is how the urea breath test earned its place, and it is probably how the next generation will too.

Could breath sensors go into wearables?

Engineers are exploring small gas sensors in face masks, phones and wearables that could track breath markers over time, much as smartwatches track heart rate. A personal baseline could make changes easier to spot than a single test compared with strangers. The obstacles are the same as for e-noses, drift, humidity and confounders, plus battery life and cost. Expect to see trend-tracking features before diagnostic ones, and expect any diagnostic claim to need medical approval.

If such devices arrive, the questions on this page, what is measured, how accurate it is in people like you, and who sees the data, will matter just as much.

Battery-powered gas sensors also need regular calibration against known gases, something manufacturers will have to automate before breath tracking can become as effortless as counting steps.

Questions to ask about any breath test

• What exactly does this test measure, and what condition is it approved or validated for?
• How should I prepare: fasting, diet, stopping medicines, avoiding smoking or exercise?
• How accurate is it in people like me, and what happens if it is positive or negative?
• Will a positive result need confirming with another test?
• Is this a routine clinical test, or part of a research study?
• Who sees my results and data?

A good clinician will answer these readily. If a seller of a breath device cannot, that is your answer about the device.

A checklist for reading health-tech news

1. Find out whether the result comes from a study, a company press release or a regulator's decision.
2. Check how many people were tested and whether the test was checked on a separate group.
3. Look for who the comparison group was: healthy volunteers or people with similar symptoms.
4. Note whether the device is approved for use, in trials or only a prototype.
5. Look for independent experts' comments, not only the developers'.
6. Ask what happens after a positive result, and how many would be false alarms.

Ethan now runs every health gadget headline through this list. Most turn out to be early research, which is interesting news, just not a product.

What this page said in 2016

The original version of this page, from December 2016, was a short news note about the Technion study and its 17 diseases. It captured the excitement of the moment but not the context: which breath tests were already in daily use, how the e-nose worked, and why a promising study is only the start of a long road. This rewrite keeps the story and adds that context, so the headline can be read for what it was: a genuine scientific milestone, not a device you could buy.

It also reflects a broader shift on this site: older news posts are being turned into explainers that stay useful after the headline fades.

Breath tests and electronic noses: frequently asked questions

Ethan's father finished his antibiotics, and a follow-up urea breath test confirmed the infection was gone. Ethan never did buy a disease-detecting breathalyzer, because, as Jake explained, there was none worth buying. But he now follows the research with real interest. The idea that a breath could one day flag disease early is not hype; it is a hard engineering and medical problem being solved one careful study at a time.

Revision note. Written October 2, 2026, rewriting our 2016 news note on the 17-disease breath study into a complete explainer on breath tests and electronic noses. May every test you take be the right one, read the right way.