Lung cancer screening in the future may be as easy as taking a breath. (Credit: New Africa on Shutterstock)

In A Nutshell

• A handheld breath sensor read eight chemicals and flagged lung cancer risk in a 67-person study.
• Accuracy was 85.71% using the device alone and 92.86% when paired with GC–MS in analysis.
• Runs on a small on-device neural network (~35k parameters) designed for privacy and offline use.
• Promising, but early: larger, diverse trials and disease-specific testing are needed before clinical use.

The prospect of simply walking into a doctor’s office and breathing into a handheld device for a few seconds to screen for lung cancer sounds too convenient to be true. Well, that scenario moved closer to reality after researchers at the University of Texas at Dallas and UT Southwestern Medical Center developed an electrochemical breath analyzer that can flag potential lung cancer and other chest malignancies.

The device works similarly to roadside breathalyzers police use to detect alcohol, but instead of measuring blood alcohol content, it identifies eight specific volatile organic compounds in exhaled breath that may signal the presence of cancer. In a proof-of-concept study with 67 participants including 30 with biopsy-confirmed intrathoracic malignancies, the researchers found their handheld prototype could distinguish between cancer patients and healthy individuals with 85.71% accuracy using breath samples alone.

“Lung cancer remains a leading cause of cancer death globally and affects millions of lives across the globe,” the researchers wrote in their study published in Sensing and Bio-Sensing Research. “Early detection of lung cancer significantly improves survival rates.”

Dr. Shalini Prasad holds a screen-printed electrochemical sensor for lung cancer signals. (Credit: University of Texas at Dallas)

Why Current Screening Methods Fall Short

Current lung cancer screening relies heavily on low-dose CT scans for high-risk individuals, typically those aged 50-80 with a significant smoking history. CT scans catch cancer earlier than waiting for symptoms to appear, but they require expensive equipment, expose patients to radiation, and often produce false positives that lead to unnecessary anxiety and follow-up procedures. When suspicious nodules do appear on imaging, patients must undergo invasive biopsies using techniques like CT-guided needle insertion or bronchoscopy to confirm whether cancer is actually present.

Detailed in Sensing and Bio-Sensing Research, the breath analyzer takes a different approach. Patients simply exhale into a disposable mouthpiece attached to the device, which pulls the breath sample through eight separate electrochemical sensors. Each sensor contains a thin coating of room-temperature ionic liquid specifically chosen to interact with one of the eight target compounds: pentane, heptane, isoprene, methyl pentane, octane, decane, benzene, and toluene.

How Cancer Cells Leave Chemical Fingerprints in Breath

These volatile organic compounds emerge in exhaled breath because cancer cells have altered metabolism compared to healthy cells. As tumors grow and spread, they produce distinctive chemical signatures that eventually make their way into the bloodstream and then into the lungs, where they get exhaled with each breath. The device measures these compounds using chronoamperometry, an electrochemical technique that monitors tiny electrical currents generated when the breath chemicals interact with the ionic liquids.

The measurement process allocates a six-second window per compound, though usable signals often appear sooner. Before analyzing the patient’s breath, the device first measures baseline levels in ambient air to establish a reference point. Then it compares the breath sample to that baseline, looking for elevated concentrations of the eight target compounds. The device accounts for environmental factors like temperature and humidity that could skew readings, applying mathematical corrections to ensure accuracy across different testing conditions.

Machine Learning Powers the Screening Algorithm

Behind the scenes, machine learning algorithms do the heavy lifting. The research team trained a neural network using data from both the electrochemical sensors and validated the results in parallel against gas chromatography-mass spectrometry, which served as the reference standard for analyzing breath compounds. In this 67-person pilot study, the neural network learned to recognize patterns that distinguish cancer patients from healthy controls, achieving its 85.71% accuracy rate when relying solely on the electrochemical data.

When researchers combined both the electrochemical sensor data and the reference lab results in their analysis, accuracy in the small study jumped to 92.86%. The team noted that the goal is eventual deployment of a standalone device that doesn’t require expensive laboratory backup. At 35,000 parameters, their neural network remains small enough to run on the device itself without requiring an internet connection, protecting patient privacy and allowing use in settings without reliable connectivity.

The study validated results using breath samples collected immediately before patients underwent diagnostic bronchoscopy procedures. Researchers filled three-liter bags with exhaled breath, then tested portions both on-site with the portable device and later in the laboratory using gas chromatography-mass spectrometry as the reference method.

Dr. Shalini Prasad (left) and biomedical engineering doctoral student Nikini Subawickrama show off the electrochemical sensor technology that could lead to a new tool for detecting various thoracic cancers. (Credit: University of Texas at Dallas)

Expanding Screening Beyond High-Risk Smokers

Smoking remains responsible for more than 80% of lung cancer cases, but air pollution, occupational exposures, genetic factors, and chronic lung diseases also contribute significantly. Crucially, nonsmokers can develop lung cancer too, yet current screening guidelines focus primarily on people with extensive smoking histories. A quick, affordable breath test could potentially expand screening to broader populations.

The researchers noted their study represents an early step requiring further clinical validation with larger patient groups. The 67-person cohort, while sufficient to demonstrate proof of concept and train initial machine learning models, falls short of the thousands of patients typically needed to establish a new diagnostic test for widespread clinical use. The device has not been tested on patients with other respiratory conditions such as chronic obstructive pulmonary disease, asthma, or pneumonia that might produce similar breath profiles, meaning the false-positive rate in real-world settings remains unknown. The team also pointed out that the study population came from a single medical center, potentially limiting how well the results apply to different geographic regions and patient demographics.

The Economics of Cancer Detection

Economic factors loom large in cancer care. The annual economic burden of lung cancer in the United States alone exceeds $20 billion when accounting for direct medical costs and indirect costs like lost productivity. Current diagnostic pathways involve multiple imaging studies, specialist consultations, and invasive procedures that drive up expenses while delaying definitive answers. A breath test that could be administered during routine primary care visits might catch cancers earlier and reduce the cascade of expensive follow-up testing triggered by ambiguous imaging findings.

The technology builds on previous work by the research team using room-temperature ionic liquids for gas sensing applications. These ionic liquids, which are salts that remain liquid at room temperature, have advantages over traditional gas sensors that require high operating temperatures and consume significant power. The researchers used computational simulations to select which ionic liquids would interact most effectively with each target compound, then validated those predictions through laboratory testing.

The team envisions a screening tool that could be deployed in outpatient clinics, occupational health settings, or even at home for high-risk individuals. The device would need to demonstrate consistent performance across diverse patient populations, different geographic regions, and various environmental conditions before regulatory agencies would approve it for clinical use. Large-scale clinical trials would be required to determine how well the 85.71% accuracy observed in this 67-person study holds up in broader populations and to establish how often the device produces false positives when testing people with other lung conditions.

For now, this breath analyzer exists as a research prototype demonstrating that electrochemical sensors combined with machine learning can flag cancer-associated volatile compounds in a proof-of-concept setting. Whether it eventually becomes a routine screening tool depends on results from larger clinical trials, regulatory approval processes, and the practical challenges of manufacturing and distributing medical devices at scale.

---

Paper Summary

Methodology

Researchers developed a handheld electrochemical device containing eight sensors, each coated with room-temperature ionic liquid selected to detect specific volatile organic compounds associated with lung cancer. The team collected breath samples from 67 participants (30 with biopsy-confirmed intrathoracic malignancy and 37 healthy controls) at UT Southwestern Medical Center immediately before diagnostic procedures. Patients exhaled into three-liter Tedlar bags, and researchers analyzed approximately 0.5 liters on-site with the prototype device and another 0.5 liters using thermal desorption gas chromatography-mass spectrometry for validation. The device used chronoamperometry to measure diffusion currents when breath compounds interacted with the ionic liquids, with measurements allocating a six-second window per compound. Researchers applied mathematical corrections for temperature and humidity variations, then used these measurements to train neural network models using five-fold cross-validation.

Results

In this 67-person proof-of-concept study, the electrochemical device alone achieved 85.71% accuracy (±4.60%) in distinguishing cancer patients from healthy controls, with five true positives, two false positives, and seven true negatives in the test set. When combining electrochemical and gas chromatography-mass spectrometry data in the analysis, accuracy increased to 92.86% (±7.14%), with five true positives, one false positive, and eight true negatives. The neural network demonstrated the best performance among six evaluated machine learning models (including support vector machines, XGBoost, convolutional neural networks, transformers, and ensemble methods). Evidential deep learning analysis revealed 21.43% uncertainty in predictions, meaning the model showed at least 78.57% confidence in its classifications. Sensor characterization studies showed good repeatability between devices and within devices across multiple concentration levels of target compounds. The device successfully detected eight volatile organic compounds (pentane, heptane, isoprene, methyl pentane, octane, decane, benzene, and toluene) with detection limits between 10-120 parts per billion depending on the compound.

Limitations

The study’s primary limitation was its small sample size of 67 total participants, with only 30 confirmed cancer cases. This limited population makes it difficult to generalize results to broader patient groups and increases the risk that the machine learning model may not perform as well on future patients. The researchers did not test the device on patients with other respiratory conditions that might produce similar breath profiles, raising questions about specificity and leaving the false-positive rate in real-world clinical settings unknown. Two target compounds (octane and methyl-pentane) were inferred analytically rather than directly validated due to difficulty obtaining sufficiently pure samples for testing. The study population came from a single medical center, potentially limiting geographic and demographic diversity. The device has not been tested under varying real-world environmental conditions beyond the controlled clinical setting. The research represents proof of concept rather than a validated clinical diagnostic tool, and the neural network’s performance may change with larger datasets. The correction equations for temperature and humidity were derived from laboratory experiments and may not apply under extreme environmental conditions.

Funding and Disclosures

The paper does not explicitly state funding sources in the main text. The authors declared no conflicts of interest. The study was conducted under two Institutional Review Board protocols: IRB 23-362 at UT Southwestern Medical Center for the 30 biopsy-positive patients, and IRB 21-442 at UT Dallas for the 37 control participants. The acknowledgments section thanks Dr. Ivneet Banga for assistance with patient breath sample collection, device operation, IRB approval, and data collection. The authors also acknowledged Ms. Maddison Mills for clinical sample collection and on-site testing, and John Machado for contributions to machine learning visuals, cloud development, and data accumulation.

Publication Details

Paul, A., France, K., Bhatia, A., Abu-Hijleh, M., Daescu, O., Thapa, R., Gordon, R.A., & Prasad, S. (2025). “Electrochemical breath profiling for early thoracic malignancy screening,” published in the August 2025 edition of Sensing and Bio-Sensing Research, 49, 100815. doi:10.1016/j.sbsr.2025.100815

About StudyFinds Analysis

Called “brilliant,” “fantastic,” and “spot on” by scientists and researchers, our acclaimed StudyFinds Analysis articles are created using an exclusive AI-based model with complete human oversight by the StudyFinds Editorial Team. For these articles, we use an unparalleled LLM process across multiple systems to analyze entire journal papers, extract data, and create accurate, accessible content. Our writing and editing team proofreads and polishes each and every article before publishing. With recent studies showing that artificial intelligence can interpret scientific research as well as (or even better) than field experts and specialists, StudyFinds was among the earliest to adopt and test this technology before approving its widespread use on our site. We stand by our practice and continuously update our processes to ensure the very highest level of accuracy. Read our AI Policy (link below) for more information.

Our Editorial Process

StudyFinds publishes digestible, agenda-free, transparent research summaries that are intended to inform the reader as well as stir civil, educated debate. We do not agree nor disagree with any of the studies we post, rather, we encourage our readers to debate the veracity of the findings themselves. All articles published on StudyFinds are vetted by our editors prior to publication and include links back to the source or corresponding journal article, if possible.

Our Editorial Team

Steve Fink

Editor-in-Chief

Sophia Naughton

Associate Editor