Olfactory AI Arrives in Medical Diagnosis: Smell Language Models

What happened?

Biotech and AI company Ainos, in collaboration with National Taiwan University (NTU), has announced a one-year research program to evaluate whether its artificial olfaction platform can diagnose diseases by analyzing volatile organic compounds (VOCs) in breath. The study, starting in July 2026, will focus on patients with dyspnea (shortness of breath), a common symptom in emergency rooms that can be due to acute exacerbations of chronic obstructive pulmonary disease (AECOPD) or acute decompensated heart failure (ADHF), two conditions requiring very different treatments.

The technology is based on the 'AI Nose' module, which integrates multiple micro-electro-mechanical sensors (MEMS) and a digital processor. When sensors detect gases, their resistance changes, generating a digital signal interpreted by a proprietary Smell Language Model capable of learning, classifying, and contextualizing complex odor patterns.

Why is it important?

Rapid and accurate diagnosis of the cause of dyspnea is critical in emergencies, but often requires invasive or slow tests. If olfactory AI proves effective, it could provide a non-invasive, instant, and low-cost method to differentiate between AECOPD and ADHF, improving treatment times and reducing mortality. Moreover, this approach represents a significant advance in applying language models to non-textual data, in this case olfactory, expanding AI's reach beyond text, image, and audio.

Ainos CEO Eddy Tsai noted that the system was originally conceived for medical diagnosis, and this research brings it back to a high-value clinical setting. The project also aims to create a database of 'breathprints' for dyspnea, which could lay the groundwork for future studies in emergency rooms, outpatient clinics, and even home monitoring.

Consequences and context

If the study succeeds, we could see early adoption in hospital emergency services, where the ability to quickly differentiate between causes of dyspnea would save lives and optimize resources. In the long term, the technology could extend to other diseases with known VOC signatures, such as certain cancers, infections, or metabolic disorders. However, challenges remain: rigorous clinical validation, standardization of breathprints, sensitivity and specificity in diverse populations, and integration with hospital workflows.

This announcement follows another Ainos program where its AI Nose was deployed in the emergency department of National Taiwan University Hospital to monitor respiratory infections and overcrowding. The company also mentions industrial applications and physical AI settings, suggesting a broader vision of creating a universal 'Smell ID'.

What readers should know

• Not a commercial product yet: This is a research study; the technology is not approved for clinical use.
• Limitations: VOCs can be affected by diet, smoking, or medications, potentially leading to false positives/negatives.
• Competition: Other companies like Owlstone Medical or Breathomix also work on breath analysis, but with different approaches (mass spectrometry vs. MEMS sensors).
• Ethical implications: Collecting breath data raises questions about privacy and consent, though being non-invasive makes it less intrusive than blood tests.

"AI Nose was originally developed with medical diagnostic applications in mind, where non-invasive detection, accuracy, and real-world validation are essential." — Eddy Tsai, CEO of Ainos

In summary, olfactory AI represents a promising frontier in medical diagnosis, but its success will depend on the solid clinical evidence this study aims to generate. Healthcare professionals and investors should closely follow the results, expected by mid-2027.