A Second Opinion from an Algorithm

For many women, vaginal discharges are an embarrassing and, mainly, needless irritant. But for Dr. Shlomit Yehudai-Reshef, a trained biologist who leads cancer research at the Clinical Research Institute at Rambam (CRIR) and Dr. Inbal Zafir-Lavie, a doctor of immunology and an entrepreneur, this is no less than a medical treasure that can generate a revolution in the field of women’s health.

The initial idea was conceived by Dr. Yehudai-Reshef and in 2015 she founded Nevia bio together with the co-founders Prof. Roni Michaely and Dr. Inbal Zafir-Lavie. The company develops a platform for early diagnosis of diseases in the field of women’s health with the key being machine learning-based analyses of vaginal discharges.

Dr. Zafir-Lavie, the company’s CEO and co-founder, describes how the developmental environment in the medical field suffers from a form of blindness with regard to women’s health: “Over the years, there have been numerous medical developments – but significantly, relatively few have been in the field of women’s health. A look at the ratio of the total budget that pharma companies invest in R&D in the field of women’s health reveals a figure of just 4%.

“This is also the reason that many things which women experience throughout their lives – including menopausal symptoms – have no proper medical solution. Endometriosis, for example, which doctors are frequently unaware of and where women suffer excruciating pain and fail to get pregnant. “One option is simply to give up”, says Dr. Zafir-Lavie. “On the other hand, the lack of solutions creates opportunities for innovation throughout the field of women’s health.

“Entrepreneurs and investors are needed for the femtech market to expand and grow, thankfully more female entrepreneurs have recently entered this field. When large companies and senior female executives in the finance world perform market surveys and generate data on the women’s health market, it helps investors understand its inherent economic power. The femtech market expanding – especially in the wake of the US court ruling against abortions. Women understand that no-one will do the job for us, and we need to take care of ourselves”.

Nevia bio operates in the Innovation Authority’s MindUP incubator. “Greater awareness is needed about what the Innovation Authority is doing in Israel. This is an entity that invests money in innovation which it is extremely difficult to advance in its initial stages”, says Dr. Zafir-Lavie, who emphasizes that it is important for any of our readers with a good idea to contact the Authority for support. “We can achieve amazing things – all we need to do is dare”.

A Biobank of a Thousand Samples

In the case of the entrepreneurs behind Nevia bio, the daring emanated from Dr. Yehudai-Reshef’s idea: “I get asked a lot if I had a revelation about this project. The answer is no – it didn’t just pop into my head”, she says. “But from the first moment I was sure that discharges could be a key to early detection of pathologies among women, I knew it would work. I knew that the test would work and that we would succeed”.

Dr. Zafir-Lavie explains: “Today I can gather discharges of 40 microliters – the size of a waterdrop – and perform numerous analyses on them that provide us with data on thousands of proteins and hundreds of thousands of genes”. The next stage was to establish a “biobank”. Nevia bio is probably the only company today that has almost a thousand samples of women’s discharges with each sample representing a specific woman, and there is a database of information about the proteins and the medical background of the woman herself.

The magic happens in the interface between humans and machine – or, in the case of Nevia bio, in the interaction between a life sciences expert and a data scientist. The platform created by the company stores all the data in the cloud and runs AI and machine learning algorithms on it. In practical terms, the platform learns how to identify the discharges of healthy women and ovarian cancer patients.

The use of learning algorithms to develop a test – such as that being developed by Nevia bio – is not trivial. In-depth knowledge is required of the algorithms that presently exist in Artificial Intelligence and machine learning and of what they are not capable of. This also calls for an in-depth understanding of the biological processes in the disease being examined and of the clinical data that has been gathered.

The objective is to let the machine learn but also to ensure that it learns correctly, in order to achieve results that are both logical and relevant for ovarian cancer. The platform currently achieves high accuracy when answering the question as to whether or not a specific subject has ovarian cancer.

Nevia bio is currently concluding a clinical trial at eight medical centers around Israel. The company completed the seed fundraising stage and proceeded to the series A funding stage to facilitate a pivotal trial after which the product can be submitted and released onto the market. “We are just beginning”, says Dr. Zafir-Lavie. The next stage is a large-scale clinical trial in the US that will encompass thousands of women. The trials in Israel are already showing very promising results.

ginalife
From right to left: Dr. Inbal Zafir-Lavie, Dr. Shlomit Yehudai-Reshef

The product currently focuses on the early diagnosis of ovarian cancer, but the vision is to diagnose other women’s diseases such as endometriosis, uterine cancer, etc. And the vision does not end with the list of diseases. Nevia bio’s vision is to make the medical world more accessible to women so that they can make decisions for themselves – similarly to what happened with the revolution of home pregnancy tests. “This gives women independence, precisely during a time when women’s independence is in doubt. We view empowerment as part of our mission”, Dr. Zafir-Lavie concludes.

“Why does a woman in the US have to travel 300 kilometers to visit a gynecologist?”, asks Dr. Zafir-Lavie who wants women to be able to test themselves, to use medical apps and to monitor themselves using a test kit delivered to their home. “This way, we will have her medical profile and, if there are any changes that may indicate a risk, we can instruct her to go for a checkup”, she explains.

Dr. Zafir-Lavie, Dr. Yehudai-Reshef and the company’s team see their mission as saving lives and far beyond simply receiving a salary at the end of the month. “Our team consists only of people who understand our passion”, the two say. “Maybe this starts with the understanding that questions about women’s discharges can lead to questions about the women themselves but continues with the question of how women’s health is perceived in different places around the world. We will have a lot of medical information that will be connected to geographical data. This added value has a commercial aspect but also a facet of public health and that is part of our agenda”.

When Competitors Collaborate
Dr. Zafir-Lavie says that as far as competitors are concerned, some companies active in this field try and take a liquid biopsy – but they don’t have data on ovarian cancer. The competitors’ approach in the field of femtech seems to be different however because, as she says, they collaborate with each other. “There is room for everyone. They helped me in finding someone to conduct the trial in the US and I helped them with something else – we have a common objective.

At the same time, Dr. Yehudai-Reshef explains that “Nevia bio’s uniqueness is in the fact that it develops a platform and not a specific test. Once a platform exists, a change in the composition of the biomarkers also allows us to alter the medical indication we are looking at. Another unique feature is the fact that this is a non-invasive test which will be performed at home by the woman herself”.

In practice, Nevia bio’s solution addresses four challenges that exist when performing tests today: a non-invasive test; a combination of several biomarkers; performing the test in the comfort of the home; the algorithms’ advanced technology that can analyze and provide each woman with an accurate result about herself.

As Dr. Yehudai-Reshef explains: “Our vision is to reach every woman on the planet – even those in remote places have a smartphone today and that’s all you need. Doctors can analyze results from anywhere. I envision a Nevia bio shelf in every pharmacy with all sorts of tests, alongside pregnancy and ovulation test kits. This is a very broad platform that meets a lot of needs and will generate a revolution in the field of women’s health”.

“When we established Nevia bio, people told us that we had no chance of success because others had already failed to do what we were trying to do”, says Dr. Zafir-Lavie. “But not everyone thinks differently or innovatively – otherwise they would have achieved what we have. When you come up with a different or innovative idea, you should consider that you will encounter objections, but that doesn’t mean it’s not a good idea. Don’t stop aiming for what you believe in – and that’s what we did. Accept the negative reactions with understanding and carry on”.

Pathologists’ Digital Assistant

Ibex Medical Analytics was established in 2016 by two founders with a background in computer science: Joseph Mossel and Dr. Chaim Linhart. The two wanted to create a company that would provide significant added value in the healthcare field and looked for an area where there would be an advantage to using Artificial Intelligence (AI) and machine learning.

Mossel and Linhart identified cancer diagnosis – an extremely critical element in today’s medical world – as a field that suffered from non-optimal performance. Despite the state-of-the-art microscopes and the sophisticated staining methods, diagnosis still relies on the doctor’s pair of eyes examining one biopsy after another throughout a busy workday.

When they started looking deeper, the Ibex founders realized that the problem was even bigger than they originally thought – on the one hand, the number of cancer patients is constantly increasing while on the other hand, the number of pathologists worldwide is declining.

Pathologists in most countries work under tremendous pressure, a phenomenon that has two main ramifications: first, working faster invariably means making more mistakes. In certain types of cancer, a mistake could result in incorrect treatment or, even, a decision not to prescribe any treatment at all. Second, the heavy workload leads to long waiting lists and delays in receiving test results. In many countries, people presently wait weeks to receive test results, a reality that naturally causes unnecessary distress.

The insight they gained was that just as pathologists are trained to identify cancerous cells according to clear visual characteristics, AI can be used to teach algorithms to do the same thing. Mossel and Linhart established Ibex, and after initial fundraising, the company purchased a special and expensive pathological scanner that is able to scan a biopsy and generate extremely high-resolution digital images.

Ibex began developing its technology in collaboration with the Maccabi HMO (Health Maintenance Organization) while relying on the huge pathological archive of Maccabi that provides health insurance for 25% of Israelis. Because medical information in Israel has been maintained digitally for decades, the new company can rapidly extract a large and diverse number of physical samples (biopsies) and pathological reports.

Within just a few months, that company had an initial version of an algorithm that knows how to identify cancer as part of an overall solution they called “Galen” after the Roman Greek physician considered one of the founders of medicine in Ancient Greece.

The initial goal was to use Galen as a quality control tool i.e., a test performed after the doctor’s diagnosis and aimed at detecting errors. The first tissue selected was the prostate and when the first results came in, it was apparent that the algorithm successfully detects errors in the pathologists’ diagnoses.

“This opened our eyes immediately to this tool’s capability to save lives”, said Tal Frieman, Ibex Chief Marketing Officer. “There is no need to explain that a wrong diagnosis, especially in a case that involves missing a patient with cancer, can potentially be the difference between life and death”.

AI Assists, The Doctor Decides

The algorithms, that are distinguished by the type of tissue tested, are taught using a technology called machine learning. As Frieman explains, pathologists mark certain cell types on images of tissue samples e.g., a healthy or a cancerous cell, and this is entered into an algorithm that learns how to identify them itself. This process is performed on dozens of cell types and hundreds or thousands of images.

“As we feed Galen with more and more data of increasing detail, it will learn to identify more types of cells, for example, how to distinguish between different types of breast cancer”, Frieman explains. As mentioned, the company started out with the prostate and subsequently developed algorithms to identify various types of breast cancer, including diagnosis of biomarkers that assist in matching the right treatment.

The third solution developed by the company is capable of identifying stomach cancer and other pathologies typical of the digestive system that may cause a disease. In the future, the company’s team plans to develop systems for other tissues beyond the three current types, such as the large intestine, lungs, and skin.

The physicians who use Galen are pathologists, primarily at hospitals, large cancer centers, HMOs, and private laboratories. “The pathologists call Galen a Digital Assistant – a digital assistant who never gets tired and who always provides consistent answers, but it is important to understand that with all its advantages, it is still just a tool that supports the decision-making process and that the final word is always that of the physician who holds ultimate responsibility for the diagnosis”, Frieman says.

“We have very fruitful collaboration with the Innovation Authority and have completed three successful projects with them. Today we are talking about a system that a pathologist works and consults with while making the initial diagnosis. In other words, when he arrives at work in the morning, he doesn’t just receive a pile of random cases from the lab, but rather an organized list of cases that our system has performed an analysis on. When he sees the list of cases, he also sees where the system located a tumor and identified its type. This saves pathologists a lot of time and enables them to complete treatment in considerably shorter time. Our system is now deployed all round the world”.

Ibex has reported extremely rapid growth over the last year. Frieman says that the company has the highest number of installations in the field of AI pathology solutions: “Our system is also the most elaborate product in terms of the features it provides doctors and its detection capability. Its high level of precision, including identification of pre-cancerous situations and diagnosis of diseases other than cancer, has been proven in a series of medical studies in Europe and the US. We employ a large number of pathologists who collaborate with our development personnel and also receive assistance from senior pathologists from leading cancer centers in the US and Europe. This process ensures that in addition to precise identification, the product will be as useful as possible, and will operate similarly to the way in which normal pathologists observe tissue and write their diagnoses. This allows them to accustom easily to working with it.

“Trust is a critical element without which no medical innovation can be achieved. AI is a new technology that is sometimes perceived as a threat. Our message is beginning to filter down to the market i.e., that the intention is not to replace the pathologists but rather to give them a tool that will enable them to be even better diagnosticians and to work much quicker. From the moment that they experience working with the system, our clients are not prepared to check biopsies without AI anymore”.

Ibex’s long-term vision is to supply each patient with an accurate, rapid, and personalized diagnosis of their disease. As Frieman explains: “Today we are only scratching the surface of AI’s capability in pathology and cancer diagnosis. We are convinced that the revolution in this field will be dramatic and estimate that there are capabilities that we don’t even know exist and that we will discover as we continue researching.

“The system allows us to generate numerous insights in the data that until today we couldn’t identify just with the naked human eye, for example, about the disease’s future development or a patient’s reaction to a certain drug. We believe that AI will be an important tool in providing answers to these questions. The experience we are acquiring with the data will lead us there.

“One of the special features of working in a medical technology company is that we are very driven by the impact that our products create. Each time we receive feedback from the pathologists who work with us, for example that they prevented an error by using the system, or that they succeeded in shortening a patient’s waiting time for results, it gives tremendous satisfaction – an amazing feeling of doing something positive for the world”.