Prof. Raphael Sznitman: 0:00

And that allowed us to go from an exam that was about five and a half to six minutes to something that was roughly two minutes long.

Alex Jani, X-Health.show: 0:12

Hi, I'm Alex and welcome to the X-Health.show, where I talk to visionaries behind the latest innovations in healthcare. For the eXtra health of the future. We're in Lausanne, Switzerland, at the Biopole campus costing 120 life sciences companies, covered in thick fog today. Professor Raphael Sznitman is the co-founder of Perivision. A computer scientist and biomedical engineer that works closely with clinicians on AI solutions in ophthalmology. Perivision is a Swiss startup, a spin off from the University of Bern. Their eye test headset uses Virtual Reality, cloud computing, and AI to test for glaucoma and visual field. It reduces the test time up to 70%. You'll hear why many people don't even know they have glaucoma, how glaucoma is diagnosed now, why it is so frustrating, and how to move the whole process home to the patients. Raphael, three days ago you got this news that you might be able to help pretty soon – how many people in the world?

Prof. Raphael Sznitman: 1:21

We are talking about around 80 to 100 million people who suffer from glaucoma. Indeed, we got the good news just a few days ago that we will be selling our system in the US. We are really excited about that potential, finally seeing the fruits of the work that was involved into bringing something that came from a very early idea into what is now a commercial entity with a product.

Alex Jani, X-Health.show: 1:55

Let's maybe just start with what is glaucoma? It leads to blindness, right?

Prof. Raphael Sznitman: 2:02

Exactly. Glaucoma is one of those those chronic eye conditions. Once you have it, you have it for life and there's no cure. It is a disease that attacks the optic nerve, the material that connects your eyes to your brain and sends all that sensory information. The cause of the disease itself is not really known. The way we manage the disease is really what happens on a clinical level. So how do you treat and manage glaucoma to ensure that the progression isn't aggressive? Or that we try to slow down that progression. In a sense, it's kind of a lose-lose for the patient and from a health care point of view, we are really just trying to manage that disease.

Alex Jani, X-Health.show: 3:02

Could you go into more details about the eye anatomy and how it starts to dysfunction, please?

Prof. Raphael Sznitman: 3:11

back of the eye called the retina. This is where you have photoreceptors that pick up light. These photoreceptors carry the information through nerves and these nerves come together at what's called the optic nerve. That is at one location in the eye. This optic nerve then goes all the way to the brain and that is how we have connection to the brain. In the case of glaucoma, it is often associated with some eye pressure that is built up inside the eyes, which causes tension and presses down, and affects these connections. A lot of glaucoma is linked to eye pressure. Not always but often. A consequence of that is that a lot of the treatment approaches is about relieving eye pressure either through drops or through surgery.

Alex Jani, X-Health.show: 4:14

So guys, just to note here, we are not medical doctors, neither me nor Professor Sznitman. So please do not consider anything we say here as medical advice. How is glaucoma diagnosed nowadays?

Prof. Raphael Sznitman: 4:33

Typically with a multitude of instruments. There is no one single instrument that says you have glaucoma, it's a mixture of things because the disease typically is a perceptual disease. It is a slow moving disease that patients start to realize, Hold on a second, I am walking into chairs, I am not seeing things on the sides. Glaucoma typically reduces your field of view from the outside in. So initially, you do not notice it until you start banging more often into objects or people, or you cannot see a whole side on the right side or the left side.

Alex Jani, X-Health.show: 5:18

You get used to it, basically.

Prof. Raphael Sznitman: 5:22

Exactly, the vision narrows. The nickname for the diseases is the silent thief of vision. It steals the vision quietly, which is scary. And so the way that it is diagnosed is a mixture of using pressure measurements of the eye, also some imaging of the eye to see what is happening on the retina, specifically on the optic nerve area, and using an instrument called a perimeter, which is basically an extension to what is called visual acuity. So everyone who has glasses and those who have done driving exams with the vision exams know, there is this chart where you read the letters at a set distance. Depending on the size of the letters, you calibrate your glasses. Perimetry is just an extension of that, that does not just focus on the central field of view, so where you really can read well, but everything outside of that, so on the outside. That is exactly where glaucoma typically affects vision. So we do what is typically called a perimetry exam to measure what is called the visual field – how well you see light, not just in the center of your field of view, but outside of it.

Alex Jani, X-Health.show: 6:43

What does this perimeter look like?

Prof. Raphael Sznitman: 6:45

You can think of it as a pretty large machine that can sit on a table. It is roughly maybe 50 by 50 by 80 centimeters big. It is a big cube. And it has a characteristic big bowl. Think of it as a giant bowl. You put your head in this bowl and you set your chin on a stand, and you just fixate a target at the back of this bowl. What happens is that inside this bowl, the machine shines light of different intensities, while you are focusing at one location. Every time you see the light, you click a button. After about eight minutes, or depending on the exam type, five to eight minutes, we have a map of how well you can see light all around you. You typically do this per eye. So that is traditional 1.0.1 perimetry.

Alex Jani, X-Health.show: 7:43

So there are these different diagnostic approaches and then there is the diagnose. What what can be done? Because it's not curable, is it?

Prof. Raphael Sznitman: 7:53

Exactly. Once the diagnosis is set by looking at perimetry imaging and pressure, typically what happens then is that depending on the severity of the stage of the disease, there is a treatment recommendation. So depending on where you and your ophthalmologist or healthcare provider are in the world, there are different ways glaucoma is managed. There are typically three types of treatment options. One of them is eyedrops that you take home and you apply them yourself. Then there is a slightly more invasive approach using laser therapies. And the last one is surgeries. All three of these are effectively trying to manage eye pressure. Depending on the severity, you opt for one early kind of approach before the other. That is really the treating ophthalmologist will dictate what is the best approach for that specific patient.

Alex Jani, X-Health.show: 8:58

So it's not that you actually treat the nerve, you do treat the eye– the eyeball.

Prof. Raphael Sznitman: 9:03

What you are really trying to do is to remove the pressure of the optic nerve. So that's the long and the short of it.

Alex Jani, X-Health.show: 9:14

Does it work well? I would not be here probably if it worked. We would not be speaking.

Prof. Raphael Sznitman: 9:22

Yeah. It is important to know that glaucoma can onset at about age 50, so relatively young. A decent part of the glaucoma population are in working age. So does it work well? Yes and no. The issue with glaucoma is that part of it will be stable for long periods of time or relatively stable and then will drop. And you have other patients that drop continuously but very slowly. So the reality is, you have different kinds of patient profiles. These therapeutic approaches – yes, they can be very effective for given parts of time but it is still something that you have to monitor every year. So typically patients will go to their treating healthcare practitioner twice a year. Every six months they will go for a routine checkup of what is happening with the disease.

Alex Jani, X-Health.show: 10:26

A question here, what is the stage when you start actually banging on furniture? Or you don't don't see it? Is there any earlier stage that you can notice there is something going wrong.[MID-ROLL starts] We'll be right back. This episode is brought to you by the X-Health.show. And me, Alex. If you still have not, please hit follow at the top. That will help me bring more of these visionaries to you. Thanks a lot. Stay awesome. Now, back to the episode. [Mid-roll ends]

Prof. Raphael Sznitman: 11:19

That's a hard question. Most people who will find out that they have glaucoma at a very early stage do so at a stage where diagnosing glaucoma is actually very tricky. Typically, when people realize that they are banging into things, they are already at, they call it, early glaucoma stage. But effectively, there is already damage.

Alex Jani, X-Health.show: 11:45

Is there any way they could see it before?

Prof. Raphael Sznitman: 11:48

As far as I know, not in routine clinic but I am sure people are working on this, from a research point of view.

Alex Jani, X-Health.show: 11:59

I am just thinking whether to ask you now about what you offer. Because you do offer a different diagnostic solution. But maybe before we jump into this, we can speak about, how did you really invented it? Or where did the idea come from? Was there any goal of this research?

Prof. Raphael Sznitman: 12:15

For us, the story started in 2015 when I hired Serife Kucur, who is actually the CTO now at Perivision. I hired her as a PhD student to work on a project It was a pretty basic research question at the time. It was to which was to investigate, using Machine Learning techniques and AI, the relation between the structure of the retina and the function, so the visual field, how, what we perceive. There is always this relation in ophthalmology between structure and function. understand if imaging could help in giving better indications on visual function. So it was a basic question and we started together working on it. And we were just getting nowhere, we were absolutely getting nowhere.

Alex Jani, X-Health.show: 13:15

How long? For how long?

Prof. Raphael Sznitman: 13:16

I think we were swimming in the nowhere ocean for a good six to nine months. At that point we went back to clinicians and said, Look, let us try to understand a little bit more on these visual fields. Because we are looking at this data and we just do not get it, we do not get it, there seems to be a huge amount of noise, what is happening. So we started really spending time in the clinic, looking at how these exams were done, how the study nurses were performing them, how the assistants were doing them, what was happening with the patients, and then it became super apparent that perimetry as an examination – just no one liked it. It was just one of those things. Patients did not like it – it is long, they get tired, you are moderately comfortable sitting with your head inside this bowl. So they are not happy. And because it is a functional exam, the quality of the results you get depends on the psychology of the patient in that moment.

Alex Jani, X-Health.show: 14:32

What does it mean? Like, how patient they are?

Prof. Raphael Sznitman: 14:35

Yes! It is an exam where you see a light and you click a button. If your patient is anxious, your results will get skewed. So patient collaboration in the exam is really important for good quality data. On the other side, assistants are forced to look at a screen to make sure the patient is always fixating on one location. So that is intensive for them. And the physicians, they look at this data and they know that a lot of the data is very noisy, so it is hard for them to make diagnostics information,

Alex Jani, X-Health.show: 15:16

What is the noise?

Prof. Raphael Sznitman: 15:19

When you shine this light, you have to be

Alex Jani, X-Health.show: 15:20

I can also imagine that the line, when sure that the patient is looking at the right thing at the right location in the field of view. And also, having a coffee potentially changes your state. And you can see impacts. In fact, there are studies that have shown the effects of coffee on on visual field measurements. It is just something that the clinical community has gotten used to, that visual field examinations are noisy and you have to interpret them with a decent amount of experience. Or with a grain of salt, depending on how you look at it. So we saw this and we said, from a data science point of view, this is both a disaster and a goldmine at the same time. It was really fascinating to see. And then with with Serife, we said, okay, let's take this opportunity and look just at visual fields to understand things a little bit better. And the first thing we focused on was what we saw from the patients is that these exams were way too long. You know, you the patient is focused, is actually dropping, right, with are putting your head in a bowl and that is the only thing you are supposed to see. If you want to try this at home, take a spoon and put it on your eye – a large spoon – and just look at it. You will see that after a few minutes, you first of all get disoriented and then over time, it just gets tiring being in that environment. And if you do that for a few minutes, you sense it. Typically patients are are not young patients, they are not healthy patients. They either have glaucoma, they might have other diseases at the same time, they have other comorbidities. So we said, Okay, we gotta somehow find a way to have a really fast exam, to do this really quickly. time?

Prof. Raphael Sznitman: 17:19

Exactly. The first minute, people are there in it, they still have energy to do it. Then there is just a tail off and the longer the exam goes, a tail off gets worse and worse, and worse.

Alex Jani, X-Health.show: 17:34

Why to keep going with these exams for that long?

Prof. Raphael Sznitman: 17:39

That is because what you are doing in this examination is you are shining this light at different locations. These locations correspond to locations under the retina. So you need to test it at different locations. And you want to get a dense image, dense grid of all the regions that see well, and that is the fundamental problem of why it is slow. So what we did with Serife is come up with a method that allowed us to only select a few locations to examine, and predict what the other neighboring locations were without ever looking at them. And that allowed us to go from an exam that was about five and a half to six minutes to something that was roughly two minutes long. We were fortunate enough in the lab to

Alex Jani, X-Health.show: 18:28

Okay, let's just stop here for a be across the street from the eye hospital. So it is really moment. When you were speaking with clinicians, was there any, about walking over and talking to people. After seeing 30-40 like, an aha! moment? Or one clinician that you spoke to and he or she was explaining something. Or a patient that you saw? Is there an image like that? patients doing this exam, you get a sense of the frustration that is happening, especially, this is not their first time they are doing it. They come back and do it every six months. Then from the clinical side, yeah, we had great partners with whom we worked from the assistants to the physicians. And the aha moment maybe came when we had put together the realization of how many locations really need to be tested to get a good sense of what the visual field was, and whether or not every single location was critical for a clinical decision to happen. Could you give a number here?

Prof. Raphael Sznitman: 19:54

In practice, what we have seen, the typical exam is about 60 locations that are imaged and we can bring it down to about 16 to 24 locations.

Alex Jani, X-Health.show: 20:08

How did you select them?

Prof. Raphael Sznitman: 20:09

The core of what Serife did during her PhD is, from data science point of view and numbers crunching perspective, she has come up with an algorithm that automatically selected what locations to test for a given patient. That work was then published, patented, and then subsequently validated in a clinical trial at the hospital on real patients. We showed the efficcacy of have that kind of approach from a time quality point of view. So despite being much faster, you can maintain the same kind of noise level as those five to six minute long exam, so there is no point doing those anymore. They are just longer. We were very excited about that and at that point, with Serife we said, Okay, well, now that we have a better way of capturing data and a better understanding of those visual fields, let us go back to just trying to understand glaucoma a bit more. [The conversation continues on the podcast]