Medical

Exhaled air contains several hundred volatile organic compounds. That some of them relate to metabolism has been known and well documented for decades. Yet the step from that insight to a usable device rarely fails on the sensitivity of the sensing — and that is what this page is about.

Acetone: the best-studied case

Acetone forms during the breakdown of fatty acids and is exhaled through the lungs. It is the marker with the broadest data basis, and the figures are public.

Group	Breath acetone	Relative to the 25 ppb limit of detection
Healthy subjects	0.2–1.8 ppm, mean 1.1 ± 0.5 ppm	roughly 8 to 70× above
Type 2 diabetes	1.5 ± 1.3 ppm	roughly 60× above
Type 1 diabetes	4.9 ± 16 ppm	roughly 200× above
Diabetic ketoacidosis	up to 1,250 ppm	far above

Values from the published breath-analysis literature; they vary considerably between studies, measurement methods and sampling protocols. The 25 ppb limit of detection for acetone comes from datasheet rev. 3.3.0 and applies to laboratory conditions, not to breath.

For acetone, NZGS 2 has a warranted limit of detection of 25 ppb, that is 0.025 ppm. That sits roughly fifty times below what a healthy person exhales anyway.

What this table actually says Sensitivity is not the bottleneck in breath analysis. Any usable sensor sees acetone in breath without difficulty. The bottleneck is deriving something from the reading.

Why a threshold is of little use

Place the ranges in the table side by side and the problem is immediate: they overlap. Healthy subjects reach up to 1.8 ppm; people with type 2 diabetes average 1.5 ppm. A single cut-off cannot cleanly separate those distributions.

The literature proposes a cut-off around 1.19 ppm for detecting diabetic ketosis. It achieves a sensitivity of about 91 per cent — but a specificity of only around 77 per cent. In plain terms: roughly one in four healthy people would be flagged.

Added to that are influences that have nothing to do with disease. Breath acetone rises during fasting, on a low-carbohydrate diet, after physical exertion, and fluctuates across the day. Testing a single reading against a fixed threshold measures all of that too.

What the actual bottleneck is

Three things make breath difficult for a gas sensor, and none of them is concentration:

Humidity. Exhaled air is close to saturated at body temperature. The permitted operating range of NZGS 2 ends at 90 per cent relative humidity, non-condensing. Sampling without conditioning — drying, dilution or temperature control — does not work. This is not a nicety but the first design decision.
Matrix. Alongside the target substance, hundreds of other compounds are present, many in a similar order of magnitude. Ethanol from a mouthwash or acetone from a cosmetic product sits in the same range as the metabolic signal.
Between-subject variation. One person's normal range is narrower than the population's. A trend over time in the same person is therefore more informative than a single value against a population mean.

Where a pattern approach helps

NZGS 2 carries four sensing elements on a die of 1.15 × 1.15 mm with two different layers and different operating temperatures. One gas event produces four signal traces, and classification works on the ratio between them rather than the height of a single deflection. The method is set out on the technology page.

For breath analysis that means: not "how much acetone" but "what does this pattern look like". Whether that overcomes the overlap of the ranges is not a question of sensing alone — it depends on whether the cohort is large enough and the sampling uniform enough to train a model that holds. That work sits with the user, not with the component.

Applications outside medicine

Part of the breath-analysis field touches no diagnosis at all and is therefore considerably simpler in regulatory terms:

Fat metabolism and nutrition. Breath acetone relates to the breakdown of fatty acids and is studied as a way of following dietary change. That is a trend observation in the same person — precisely the task for which single-value-against-threshold fails and a trend does not.
Workplace exposure. The question of which substances someone was exposed to is not a diagnosis but an exposure measurement.

What is being investigated in research

Beyond acetone, the scientific literature investigates further compound groups as possible breath markers, among them aldehydes such as hexanal, heptanal, octanal and nonanal, benzene derivatives, alkanes and sulphur-containing compounds.

Explicitly not a product claim That a substance is studied as a marker does not mean a sensor detects a disease. These research fields are named here to place the state of the literature — not as a description of what NZGS 2 does. We make no claim that any disease can be identified with it.

Regulatory position

This is the most important section on this page.

NZGS 2 is an electronic component and not a medical device. It carries no CE marking under Regulation (EU) 2017/745 on medical devices, is not designed for any medical purpose, and replaces neither a diagnosis nor a clinical judgement.

Anyone building a device with a medical purpose from this component thereby becomes the manufacturer of a medical device under that Regulation and assumes the full obligations: classification, risk management, clinical evaluation, conformity assessment, technical documentation and post-market surveillance. Those obligations sit with the device manufacturer and cannot be passed to a component supplier.

This is not an evasion but the division of labour. We supply a component with documented specifications and support characterisation. The intended purpose, and everything that follows from it, is defined by the user.

What the sensor does not do in this application

It makes no diagnosis and is not a medical device.
It does not work with unconditioned breath. Humidity has to be controlled before the measurement.
It brings no model with it. Classification emerges from your data, your cohort and your sampling.
It is vulnerable to poisoning. Silicone compounds destroy the tin dioxide layer irreversibly — with breath devices involving tubing, mouthpieces and seals a very realistic risk that must be considered in material selection.

How a project begins

At the outset stands the question of whether your target substances separate under your sampling. That cannot be derived from specifications, only measured. The evaluation kit with read-out electronics and software exists for this: it records the four signal traces, extracts the features and shows whether the target substances form separate regions in feature space.

For breath applications, sampling is a work item in its own right — humidity conditioning, a defined breath manoeuvre, exclusion of interfering sources such as mouthwash or cosmetics. All specifications of the component are in the datasheet extract.

Assess feasibility for your question

Tell us which substances you want to measure in which matrix. We will tell you where the technical limits are — and what you will need on your side in terms of sampling and data.

Send an enquiry

Concentration ranges and cut-offs after published breath-analysis literature; sensor specifications after NANOZ NZGS 2, datasheet rev. 3.3.0. This page describes a component, not a medical application. It constitutes no health claim and does not replace medical advice.