Disease detection¶

In a nutshell¶

The VitalSensor, placed between the frames, continuously "smells" the air inside the hive. Every colony has its own natural "scent": a chemical fingerprint made up of many compounds. When a disease develops, this fingerprint changes in a characteristic way, usually long before you can see any symptoms with the naked eye.

Our system has learned to recognise these patterns and, based on them, assigns the colony a health status and an infection level. You see the result in the Apisense app.

A complement, not a replacement

Detecting disease from hive air supports your observations and inspections. It does not replace them. Treat alerts as an early "take a closer look at this colony" signal, not as a final diagnosis.

What the sensor "senses"¶

The VitalSensor measures more than a dozen parameters inside the hive. The most important are:

The chemical composition of the air: volatile organic compounds (VOCs), i.e. the colony's chemical "scent",
Temperature and humidity: the microclimate of the nest,
Other environmental parameters, including acoustic activity (the sounds of the colony).

How our system knows what is sick¶

A sensor reading on its own would tell us nothing without a reference to reality. That is why our models learn from labelled samples, in which sensor readings are paired with the actual examination result for a given colony, e.g. the number of Nosema spores under a microscope or the Varroa infestation level.

We collect this data in the laboratory (bees in cages with sensors, infected and healthy groups) and in experimental apiaries, cared for by beekeepers and scientists. Each disease leaves its own recognisable "signature" in the data, and the model learns to tell it apart from a healthy state.

American foulbrood: data from vets¶

American foulbrood (AFB) is a notifiable disease, so we never induce it in apiaries for research. We collected foulbrood data differently:

the sensors were installed in hives by befriended beekeepers and vets when they had a well-founded suspicion of foulbrood,
we recorded the data until laboratory confirmation,
once the disease was confirmed, the equipment was disposed of, in line with biosecurity rules.

Infection levels: what they mean¶

For some diseases the app shows not just "present / absent" but also a level: low, moderate or high. The values below are approximate thresholds we use to label our data. They are aligned with the standard examinations beekeepers perform every day, so they are easy to relate to your own apiary.

Varroosis (Varroa destructor)¶

Level	Alcohol wash or sugar roll	Daily natural drop	What it means
Low	up to ~2% infestation	2-3 mites per day	Acceptable and common: Varroa is present but does not yet threaten the colony.
Moderate	~3-4% infestation	4-10 mites per day	A disease: plan a treatment.
High	above ~5% infestation	above 11 mites per day	Advanced varroosis: the colony needs intervention.

Alcohol wash vs. sugar roll

The thresholds above apply equally to the alcohol wash (the reference method, the most accurate, but it destroys the sample) and to the gentler, non-invasive sugar roll, after which the bees go back to the hive. Both methods report infestation as a percentage, but the sugar roll usually recovers slightly fewer mites. So read a sugar roll result conservatively: if it is close to a threshold, treat it as the next level up.

Nosemosis (Nosema / Vairimorpha spp.)¶

Level	Spore count (per 10 bees)	What it means
Low	1-3 spores	Acceptable and common: nosemosis is present but does not yet pose a threat.
Moderate	4-40 spores	A disease.
High	above 40 spores	Advanced disease.

Where this spore count comes from

The spore count is determined under a microscope. We describe it in the Nosema/Vairimorpha microscopy procedure. The values in the table are the average number of spores per 10 bees.

American foulbrood and chalkbrood¶

For these diseases the model detects presence, not a level:

American foulbrood (AFB): a "foulbrood risk has appeared" signal. This is a notifiable disease, which is why we collect its data in a special way, described in American foulbrood: data from vets.
Chalkbrood: we distinguish a healthy state, mere exposure (contact without visible symptoms) and clear disease symptoms.