A lab at the lab: These Israeli dogs could save your life

A zoologist, a dog trainer and a scientist specializing in smells don’t sound like the typical team behind a technology startup, but SpotItEarly, which is developing technology for the early detection of disease, is no typical startup. Nor is its location – SpotItEarly works out of what used to be Kibbutz Hamaapil’s dental clinic, amid green lawns and winter pools brimming with birds.

On coffee breaks, the workers sit on a low bench beneath a pair of grand trees with serrated leaves and reddish fruits. A few times a day some of them take 12 Labradors for walkies between the orange orchards and cotton fields. Most strangers wouldn’t realize that the dogs have been trained to identify four common types of cancer – breast, prostate, lungs and colon cancer, in their early stages.

The idea that dogs can help detect illness has been known in modern medical circles for years. For instance, dogs have been trained to identify low or high glucose in diabetics, which affects the smell of their breath. Service dogs already accompany epileptics and can warn of imminent attack, by changes in body odor caused by excitation of nerve cells and hormone secretions.

The case of a dog identifying his owner’s melanoma was reported three and a half decades ago, which inspired dozens of methodical clinical trials showing that dogs can distinguish between a cancerous sample and a healthy one based on nothing but smell, before the human develops any symptoms.

The main hurdle, to this day, has been to implement this ability in the field; taking it beyond the labs of academia and turning it into an accessible test method, at a vast scale – suitable for the general public – has been a huge challenge. The people at this Israeli startup believe, and are trying to prove, that the missing link is technology. If things go as they expect, the startup people believe that 20 dogs working in ten lab rooms can test more than a million people annually, which is about the number of cancer tests done in Israel a year.

• Israeli Scientists Developing System to Detect Cancer Remnants During Surgery
• The Israeli Scientist Who Is Sniffing Out Cancer
• Study Reveals How Dogs Hijacked Our Hearts and Wallets

That’s without even getting into pooling, based on testing a large number of samples together, a method that has proven its merit with coronavirus tests, and increases profitability by hundreds of percent. And all the dogs would have to work is three hours a day. The rest of the time they can play, train and go for walks in nature, the startup people stress.

This is how it works: the test subjects, after a short break from smoking, alcohol and food, don a disposable surgical mask and breath for five minutes. The mask is then sealed in a bag and sent to the company’s lab. Pieces of the mask are retained for quality control and the rest is placed in cones in the designated, controlled Olfaction Room, which has sample stations along the wall. Each dog enters the room, walks past the samples and marks the positive samples by sitting next to them. When at least three of five dogs react similarly to a sample, it is declared positive for cancer.

Manual processes involving paper and pen are replaced by computer systems, rendering the work more efficient. The lab is equipped with sensors that enable the system to automatically collect and register the dogs’ rulings, and grade them based on behavioral signals and physiological data attesting to the dogs’ degree of confidence in classifying a sample – for instance, how long they sniff and their intensity of excitement. With enough data accrued over time, the company’s algorithm may manage with fewer dogs at a time to make the determination. The data also help the lab monitor the dogs’ aptitude for the job.

“The dog isn’t the bottleneck. The dog can handle the samples within seconds. The labwork is what creates the delay,” explains Assaf Rabinowicz, who did his doctorate in statistics and serves as the company’s chief development officer.

A lab on fur

It all began when Roi Ophir, a veteran tech entrepreneur and today the company’s active chairperson, watched a BBC news item on a doctor who founded an association to study dogs’ abilities to smell cancer, after her own dog identified her breast cancer and saved her life. In 2020, Ophir brought together three friends – Ohad Sharon, the company’s chief strategy officer; Udi Bobrovsky, chief operating and product officer; and Ariel Ben Dayan, formerly commander of the IDF canine unit (Oketz) and today the company’s CEO.

SpotItEarly started with money from its founders and in the middle of 2022 raised $6.2 million in seed funding from the Hanaco Ventures fund and private investors including Michael Eisenberg and Avishai Abrahami. In September it raised $3 million using the “SAFE” format (quick investment without a company valuation) in a mezzanine round led by the impact fund Menomadin. At present the company has 31 employees.

“Early detection is the holy grail in the realms of cancer,” Ophir says. “About $150 billion a year is invested in screening, but they don’t provide a good enough solution – only 54% of cases are detected at an early stage.”

If melanoma and breast cancer are removed, then the statistics dramatically worsen, he says. “We also realized that about half the people who need screening aren’t doing it because of problematic user experiences – they’re invasive and unpleasant, and only check for one cancer at a time. People simply prefer to ignore them. But then there is this capability in nature that isn’t exploited, except for the purpose of academic papers. The cost of our test is estimated at about a fifth of the present cancer detection tests.”

Asked how a canine olfactory screening service could obtain certification from the U.S. regulator CLIA (Clinical Laboratory Improvement Amendments), Ophir explains that the answer lies in the statistics. A great deal of research has been done on cancer detection by dogs and most studies show the method produces impressive and repeatable results over time. “Most [of the studies] reach the same percentages of detection, so there seems to be a common denominator. Medical devices also often malfunction,” he points out. Their solution is to work with multiple dogs and weigh their output, then feed the result to the technological system – the AI, for continuous validation of the result.

The company will also need certification from the U.S. Food and Drug Administration for its sample collection kit, which as far as the regulator is concerned, is equivalent to a medical device. But the company’s advisers believe its reliance on existing products such as surgical masks could facilitate this process.

The part of the canine brain responsible for olfactory memory is 40 times more ‘developed’ than the human counterpart.

Odor of acetone and smell of sulfur

Smell perception is the result of volatile compounds in the air interacting with olfactory neuron receptors in the nose. Evolutionarily speaking, smell is the oldest sense, and the only sense whose signal goes to the brain directly (to the olfactory bulb), while signals from all the body’s other sensory systems are routed through the thalamus. Over the years, artificial sensing technology has developed based on odors, for instance in food quality control, detecting toxic gas leaks, perfume development and various uses in medicine. Science can say for example that change in the glucose balance in diabetes patients makes their breath smell like acetone; lung diseases creating bacterial accrual smells like sulfur; and patients with kidney disease emit a sweet fruity odor because of accruing urea in their blood.

But in contrast to colors, sounds and flavors, smell is challenging. There is good reason why most societies haven’t developed a unique language for smells. A 2014 study found that the difficulty is reflected, among other things, in length descriptions, in reliance on other smells (“like banana”) and in general disagreement over the description among testees speaking the same language.

“You call it smell, I call it volatile organic compounds,” says Dr. Reef Einoch Amor, SpotItEarly’s expert on smells. “These are molecules that I identify, not something amorphous that can’t be seen – I see their structure and chemical formula.” Analysis of volatile organic compounds produced by cancer cells that can be detected in various body fluids – blood, urine and sweat, and in exhaled gases is, she claims, a promising development in the field of the biological signals of cancer.

Einoch Amor probably couldn’t have found a more appropriate work environment. She had just completed her doctoral studies at the Technion University, in the laboratory of Prof. Hossam Haick, a renowned researcher working on developing an “artificial nose” to identify diseases. The device is built of gold particles and carbon nano-tubules underlying a layer of sensors that react to volatile compounds in breath.

His mission is especially difficult because of the sheer multitude of parameters that can affect breath samples, such as what the testee ate the day before, and even the testee’s mood; in addition, the concentration of molecules characterizing the disease is low, one per trillion other particles. Moreover, different diseases may produce similar or even identical volatile compounds, and distinction will require levels of sensitivity that haven’t been commercially developed yet.

The best-known biotechnology company in the field of “electronic noses” is apparently Owlstone Medical, which was established in Cambridge University in 2016 and has already raised more than $150 million, for the early detection of diseases such as cancer, asthma, cirrhosis of the liver, and chronic pulmonary blockage diseases, by means of what he calls “breath biopsy”.

“At SpotItEarly, we sort of skip over the obstacles of the electronic nose, because the means aren’t analytical and there’s no need for sensors,” says Einoch Amor. “A dog can smell a teaspoon of sugar spilled onto an area the size of two football fields. The measurements we make are really just to validate the method – to make sure that the dog was given a smell from the testee’s breath, not the environment.”

At the startup, she is presently looking into processes of development and improving the sampling using surgical masks, for instance by adding materials to the masks that would increase the amount of scent molecules and improve the accuracy of the dogs’ output, not dissimilar to DNA amplification techniques.

Intellectual challenge for the dogs

Dr. Irit Gazit, a zoologist who ran the RD team of the Israeli army dog unit Oketz for 25 years, joined the company a month and a half ago.

At SpotItEarly, Gazit is responsible for training the dogs and verifying that the experiments with them are not biased. To illustrate the challenge she recounts the story of Clever Hans, a German horse who made headlines in 1907 for answering mathematical questions. Hans would tap his hoof until reaching the correct number and German academia was most impressed, until the psychologist Oskar Pfungst showed that Hans could only succeed if the questioner knew the answer. Clever Hans couldn’t count. He was responding to unwitting body-language cues.

“With us, the trainers go into a room with the dogs but the experiment is double-blind. The trainer doesn’t know where the target scent is, so the dog doesn’t get unwitting cues,” Gazit says. The main obstacle in training dogs to sniff our cancer is that we don’t know how it smells, she says.

It’s trivial to train a dog to sniff out TNT: you provide the target smell and it gets a reward, then associates between the explosive’s odor and the reward. It’s classical conditioning, she says. Then the dog is taught to distinguish things we don’t want it to signal. “In the case of cancer odors, it’s much more complicated. We can’t characterize the scent molecules. If we could, we could create technology to do it.”

Gazit compares the dogs’ process with artificial neural networks used for machine learning, which are in many ways a “black box”. The programmers working with them can understand the input the networks receive, and can analyze the output they produce, but how they process the information inside remains a mystery. The parallel with the dog is that it gets numerous samples of cancer and is asked to sit by them in exchange for a reward, and by itself understands what they have in common without any methodical training.

“The dog’s sense of smell is one of the most developed in the animal world. Anatomically speaking, the dog has 180 million smell receptors in their nasal folds on average,” she explains. The part of the canine brain responsible for olfactory memory is 40 times more “developed” than the human counterpart, enabling them to remember and distinguish many more smells than we can, including ones for which humans don’t have smell receptors.

Regarding exploitation of dogs for human purposes, she insists that the dog is ultimately an intelligent, curious animal, that needs intellectual stimulation, physical activity, and love and they get all that on the job. “The challenges we give them enable them to utilize their full mental potential,” she says.

Who’s a good dog

The company is presently conducting a clinical trial, one of the biggest ever on identification of disease by animal, with about 2,000 test subjects. The trial is being led by Prof. Nadir Arber, head of the cancer prevention center at the Sourasky Medical Center in Tel Aviv, The Hadassah medical group is also participating. The data so far are encouraging. Intermediate results published in early February were based on 575 samples collected at the hospitals from verified cancer patients and a control group negative for cancer. The samples were sent for testing at the company’s lab in double blind conditions.

SpotItEarly’s screening method correctly identified 92.8% of the sick participants, which is a high sensitivity rate compared with other screening methods. The false positives rate was less than 7%. The test proved stable over four types of cancers “participating” in the study – colon (95%), breast (93%), prostate (93%) and lung (91%). The results are encouraging given that the sensitivity of the test remained high even at early stages of disease – stages 1 and 2, though at this stage in the experiment, in a breakdown by stages and cancer types, the sample sizes produced are too small.

After the clinical stage, the company will have to tackle the regulation. Yifat Zivony, an expert on human genetics and manager of the SpotItEarly labs, had been a public health official until recently, responsible for a team of 70 people and manager of the labs at the Sieff hospital in Zefat. She had tenure, good pay, and a pension track, but last August she went on unpaid leave to look into joining a small startup, where her job would be to build processes from scratch to bring its lab to medical-laboratory standards. That included, for instance, deciding how training dogs to carry to clinical screening would look – what the criteria for accepting or rejecting a dog might be, and what the dog’s “internship” would constitute.

On the choice to take the risk and join the startup, Zivony says, “At the hospital, mainly at the blood bank and hematology department, we understand we’re putting out fires. A blood sample arrives, one sees irregularities, and behind the scenes are human stories that you never encounter early enough, but only after the disease has set in. A young woman just taking finals at university, whose father forces her to be tested because she isn’t eating and is losing weight, and suffered from night sweats – the answers they get are the moment their world is ruined. I did that for years. Now, imagine how moving it is to work on a test that detects the disease in very early stages, which would prevent these situations.”

Traditional methods for diagnosing cancer include a physical examination, blood, stool and urine samples, imaging and biopsy. They are in widespread clinical use, but few have the characteristics necessary to enable cancer diagnosis before symptoms emerge, and they have various complexities – they are invasive, expensive, may involve radiation and are difficult to perform. In the case of breast cancer, screening tests such as mammograms may be important but they are criticized. Their sensitivity is low, leading to false positives and unnecessary biopsies. The statistics show that 50 women undergo biopsy to save one woman’s life.

“A breathing sample inside a mask is more durable than blood, urine or stool. It’s less vulnerable to temperature changes, is small in volume and weightless, so it can be easily transported,” Zivony lists. “We need to think ahead and also prepare for situations in which people do the test at home, under less controlled conditions than collection at a hospital.”

Newer diagnostic methods are under development to bridge the conundrum of early detection by means of biological markers such as “free” DNA segments in the blood, blood tests for genetic changes that signal the risk of developing cancer, and so forth.

“The next-generation screening test market is in its infancy and is emerging these very days,” says Ben Dayan. “It takes a long time and is expensive to develop tests, but we are highly encouraged by the results we have achieved so far.”

If everything pans out as they expect, SpotItEarly will have the opportunity to take its place as a unique test for cancer, different from the other solutions expected to reach market in the coming years.