A £100,000 donation from the perfume entrepreneur Jo Malone will accelerate work on a cancer-detecting electronic nose, the charity receiving it says.
Medical Detection Dogs (MDD), based in Buckinghamshire, has worked with the quantum physicist Dr Andreas Mershin for more than ten years on the device, which is modelled on the scent-detection abilities of dogs. The donation followed a demonstration to Queen Camilla, the charity’s patron, in which a prototype successfully detected prostate cancer from a urine sample.
What the money buys
Not the device. The funding will install a SniffSpace — a secure room where the charity’s dogs can sniff more samples and generate more “robust data” to develop the electronic nose.
That is worth dwelling on, because it is the real bottleneck. At present the dogs sniff samples arranged in up to eight stands; after each row, trainers have to manually record the data and swap the samples around. SniffSpace will rotate the samples automatically, so more can be sniffed in less time.
In other words, the limiting factor on building an artificial nose is not the sensor or the model. It is how fast humans with clipboards can run the dogs that produce the training labels.
Why a dog, and why that is hard to copy
MDD says every disease has an odour, and that the biochemical changes involved can be smelt by its dogs. A dog’s nose is an extremely sensitive chemical detector, with around 300 million sensory receptors against roughly five million in a human.
The charity offers a comparison that makes the gap concrete: if a human can smell a teaspoon of sugar in a cup of tea, a dog can smell that same teaspoon in a volume of water equivalent to two Olympic-sized swimming pools.
The electronic version, created by Mershin at RealNose.ai, works on a similar basis. It uses protein odorant receptors to capture scent molecules from a urine sample, then uses AI to find recognisable patterns in which receptors activate. MDD has been assisting in training that AI element.
Where it has reached
The prototype is currently sniffing 500 urine samples provided by patients from Milton Keynes Hospital. MDD-trained dogs have already identified which of those samples were cancerous. Once that dataset is complete, researchers can compare the machine’s results against the dogs’ — which is the right way round: the dogs are the benchmark, not the marketing.
Chief executive Dr Claire Guest called the donation “game-changing”. “This technology gives us an unprecedented opportunity to decode those scent signatures and move closer to developing a reliable electronic nose for cancer detection,” she said.
Mershin said the charity’s facilities and protocols were “cutting edge” and were “inspiring others around the planet”. He also set out what the partnership is actually aiming at: “RealNose.ai and MDD are focused on establishing a formal partnership aimed at creating in vitro diagnostic medical devices specifically for the NHS to look at as potential solutions to the prostate cancer screening and early diagnosis problems before also trying out the same olfactory approach for other cancers.”
Malone said: “My sense of smell has given me so much throughout my life. So I have always wanted to find a way of using that world of smell to give something back.”
A team of dogs will now be prepared for training on the equipment. Data collected in the SniffSpace could eventually help the e-nose detect other diseases. Other electronic noses are already used in food and beverage quality control, environmental monitoring, and security and crime prevention.
Why this matters more here than in Buckinghamshire
Prostate cancer screening is a problem of cost and access, not only of accuracy. A test that needs a urine sample, a sensor and no radiologist is exactly the shape of test that could work in a district hospital — which is the opposite of most new diagnostic technology, where the equipment and the specialist have to arrive together.
Two cautions, though. The timeline here is long: more than ten years of work, and the stated goal is still to produce a device for the NHS to look at as a potential solution. And an AI reading sensor patterns has the same weakness as every other diagnostic model — it is only as good as the labelled data behind it, which is why the charity is spending a six-figure donation on a room that lets dogs work faster.
We have seen the version of this that has already arrived: an NHS clinic using AI to rule out melanoma and see a third more patients. The pattern is the same — the technology earns its place by removing a bottleneck, not by replacing the judgement at the end.
