Brain MRI – Faster Lesion and Brain Metastasis Detection: Dr. Christian Federau, AI Medical

Dr. Christian Federau, AI Medical: 0:00

The AI is just here to help, but it is not a fully automatized thing.

Alex, X-Health.show: 0:08

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 at ETH Entrepreneurs Club. ETH is the Swiss Federal Institute of Technology in Zurich, a public research university. My guest today is Dr Christian Federau, a physicist and a physician with 15 years of expertise in neuro radiology, brain imaging, and extensive programming experience. He worked days at a hospital and nights programming. In 2021, he founded AI Medical and now is ready to offer the first product, Jazz, an Artificial Intelligence platform that makes follow-up assessments for multiple sclerosis and brain metastasis faster and more accurate. You'll hear today about a difference between CT scan and MRI, what do radiologists spend on, b rain metastasis, multiple sclerosis, scrolling in time, and how our brains differ beautifully. Christian, Google just announced that their AI could describe X-rays. Are you worried?

Dr. Christian Federau, AI Medical: 1:16

Not at all. We are very happy that Google is working on this and it is exciting news. But radiology is very complicated. It is not just like this kind results... it is not going to be applicable on every kind of radiology and every finding. So it is really nice to know that Google is working on this.

Alex, X-Health.show: 1:36

They said it will be based on LLM, Large Language Models. You would need to prompt it, like, " write a report for me". You don't see it on these radiology reports, do you?

Dr. Christian Federau, AI Medical: 1:51

We see. But currently it is one thing to write a nice report and LLM, I am sure, is going to write nice reports but it is another thing to really describe the findings correctly. In radiology you want to be sure that it is correct. If it is not 99% of the time correct, it is maybe not what you want. You want to be sure.

Alex, X-Health.show: 2:10

Let's go back a bit. What does a neuro radiologist's work look like now? What do you have to do? You used to work as a neuro radiologist, you don not work anymore, 15 years of experience there. What is it like?

Dr. Christian Federau, AI Medical: 2:30

A radiologist most of the time is going to spend time in a dark room looking at image. The daily routine is typically 20 to 40 cases per day. But a radiologist not only looks at the image. He always prepares the protocol, so you are not going to have the same exam for every patient. He is also talking to the patient. He is talking to the referring clinicians. So, yeah, that is basically the day of a radiologist: looking at a lot of images.

Alex, X-Health.show: 2:57

What do you like in it? Why have you chosen that?

Dr. Christian Federau, AI Medical: 3:00

It is very interesting. V ery, very interesting. You see, first of all, you look inside the patients. And people are nice outside but the anatomy inside the people is also really nice, and that is one of the things that radiologists actually enjoy every day. They can look at anatomy of the people, inside. It is really nice.

Alex, X-Health.show: 3:18

How does it differ? Because when we spoke before, you said that there are a lot of differences in people's brains.

Dr. Christian Federau, AI Medical: 3:26

Exactly. That is the funny thing that you actually do not maybe think about. But we all look different from outside, but we also all look different from inside, and that is very interesting. It is very interesting to actually see that people might, for example, have arteries at different places than other people have. It is really fascinating to see.

Alex, X-Health.show: 3:45

That does not mean like one part of the brain is bigger or smaller, or..

Dr. Christian Federau, AI Medical: 3:57

S ome arteries might run differently. You are not going to have a much bigger part of the brain but yeah, there is a difference that is astonishing that you might not think [about]. And then it is going to have consequences for pathologies, like, if you have a stroke. People might have, for example, different stroke, just because the arteries are differently placed in the brain.

Alex, X-Health.show: 4:16

Could you give an example of two different strokes. Could you recall two different pictures?

Dr. Christian Federau, AI Medical: 4:22

So you have symmetry – left, right – and then, basically, in the brain you can have your blood flow coming from the base of the brain and then goes up circularly. So there are these arteries bringing the blood. But then you also have blood flow that can come from the other side, and this is very different from patient to patient. And so patients that have very good, we call this collaterals, they might have a occlusion of artery but still have a very small stroke or no stroke at all. And other people who do not have these good collaterals, they might have huge stroke, and it is totally different outcome then.

Alex, X-Health.show: 4:59

How do you learn to recognize is that this artery or not? Or like how to find it? And with these pictures, how do you learn what to see, how to recognize different things?

Dr. Christian Federau, AI Medical: 5:16

It is a practice. It is basically a lot of practice. So you just see cases one after the other. You have six years of residency and then you make a fellowship, so you are going to look at cases and then someone that has experience is going to look at them with you and is going to explain to you what you are supposed to see. And yeah, there are really two aspects that are very important in looking at image. One side is the anatomy. So you really have to learn really well the anatomy And this is somewhat surprising, but you might maybe not think about this but lesions can look exactly the same, but depending on where it is anatomically, it is going to be totally different lesion. And so that is why you need to know perfectly your anatomy. And the second thing that is very important is to understand the MRI signal, so the whole physics behind it. But it is very intriguing, the signal is very complex and you need to understand it to be able to interpret the image.

Alex, X-Health.show: 6:13

Right. So let's cover a couple of basics here. You are a specialist in MRI or Magnetic Resonance Imaging. What is the difference with a CT scan, so Computerized Tomography. We can use either/ or right? For some diseases this, for some that.

Dr. Christian Federau, AI Medical: 6:31

Yes, exactly. So, basically, the physics behind is different. So for CT, you use X- rays. It is like an X- ray but three dimensional and so you reconstruct the image three dimensionally, based on X- rays that you send to the brain in three dimensions. MRI is based on a totally different principle, but then it is going to make a three dimensional image of the brain as well. And yeah, for given indications, typically, if you want to see a bone fracture, CT is going to be better. So for given indications it is going to be better to have a CT, for given indications it is better to have an MRI.

Alex, X-Health.show: 7:02

And for the neurological diseases that you specialize in?

Dr. Christian Federau, AI Medical: 7:07

Y ou do both. Typically in emergency rooms, you're going to make a CT because you often have fractures. So you want to see this and you want to see bleeding that you see very well in CT. And CT is also much easier to do. In MRI you have a very strong magnetic field so you have to be very careful who you put in the MRI. You want to be sure that they don't have any metal with them, because it is going to be very dangerous. And so in the emergency room, you don't have time or people might not even be able to answer questions, so you prefer much more having a CT first.

Alex, X-Health.show: 7:43

And why you chose MRI?

Dr. Christian Federau, AI Medical: 7:47

I did both but MRI is, of course, it is very interesting from the physics point of view. You know I'm both a physicist and neuroradiologist.

Alex, X-Health.show: 7:55

Right, you are a physicist and a physician.

Dr. Christian Federau, AI Medical: 7:58

Yes, exactly Exactly. So, maybe because it is easier to explain, I can say I studied both. But yeah, MRI is very interesting from the physics point of view. Alex, X-Health. show And what's interesting there?

Dr. Christian Federau, AI Medical: 8:13

Dr. Christian Federau, AI Medical Y ou know, there is a long history about MRI. There is a long list of Nobel Prize that arise from from MRI. It is very, very interesting. Basically, the contrast arises from the nucleus of the atom. So you basically play with the atom nucleus, you get contrast out of it. And it is just beautiful the way that people set this up and that you can have an image of inside your body, which is a beautiful principle.

Alex, X-Health.show: 8:39

When I look at these pictures, I can't see right. What is there, what is wrong there. So if you could recall, like, when you were a student learning that, so what do you need to focus on? Or, how can you recognize there is a lesion and there is something else, or is it the contrast that does the job?

Dr. Christian Federau, AI Medical: 8:59

Yes, I mean it is true that once you get used to this, you immediately then see, and it is just like looking at any kind of other image. Ultrasound is even worse. I really remember not seeing anything. And people are like, oh, look at this and this – you don't see anything. For two weeks you don't see anything. And then slowly it comes and then once you train, it is just totally normal to see. You are just used to see normal image and then you are used to see lesions. So that is that easy. But yeah, it is a matter of habits.

Alex, X-Health.show: 9:28

Okay, okay. And what made you then think about some other solution or some support for you, for your work, for your watching the images the whole day.

Dr. Christian Federau, AI Medical: 9:47

The story actually starts a long time ago. I was in a first year of residency. I read a book called "omplication that explains how given industries made their homework on complications, on how to be secure. And one example is airline industry and the other example is anesthesia. The two nice examples. You can have a full anesthesia, it is absolutely not dangerous And it is crazy because they put you completely to sleep and there is absolutely no danger. And same for planes. You never hear now [about] a plane crashing. And it is unbelievable. It is very a complicated process. And there is never a plane crashing. And other fields of medicine have not done their homework, and radiology has not done their homework, surgery has not done homework. Alex, X-Health. show What do you mean by not done their homework? There are many complications, is that what you are saying?

Dr. Christian Federau, AI Medical: 10:35

Dr. Christian Federau, AI Medical No but there are just too many errors, unnecessary errors happening all the time. And at that time like I thought maybe I could do something about this.

Alex, X-Health.show: 10:45

So that was a seed? Dr. Christian Federau: Exactly, right. Alex, X-Health. show: When, when did you add the next layer?

Dr. Christian Federau, AI Medical: 10:53

I worked on this, learned everything that I had to learn to do this at some point in my life, I think. And then, like when I saw AI coming, I realized, wow, now, now there's all the things that you can do that you couldn't do before. And I think then I decided, oh, I should do this now.

Alex, X-Health.show: 11:09

Did you share the idea with anyone?

Dr. Christian Federau, AI Medical: 11:12

I did. I mean, but you know it's not really an idea, it's a decision. I decided to do this. And then I had many ideas. Then we started working. We looked at what is the problem and then we got ideas. And we get ideas all the time, so we continue developing. So it's a decision, we decided to do something with this and then we started.

Alex, X-Health.show: 11:32

What were the complications, or "errors as you called them?

Dr. Christian Federau, AI Medical: 11:36

Y ou know, you miss lesions, you miss stuff so constantly and one of the reasons is that you don't work in an environment that, for example, multiple sclerosis that we work on, there are a lot of lesions, there are hundreds of lesions that you have to describe. The anatomy is very complicated. It is impossible to do this work correctly. It is impossible with the current tools.

Alex, X-Health.show: 11:57

Because there are so many of them? They are so close to each other? They are unrecognizable? What's the difficulty?

Dr. Christian Federau, AI Medical: 12:03

You have to imagine that the anatomy of the brain is very complicated. And then some patients might have hundreds of lesions, and for multiple sclerosis what is really important is to find one new lesion. So you have to find one new lesion. But maybe the patient has hundreds of them. You have images but they are not alligned. You might even make an error looking for the previous exam, you might take the wrong previous exam. These kind of errors that are totally unnecessary. That is why we made software that cannot make these errors. It finds the previous exam for you And it's a computer and it is never going to make a mistake for this. That's a very simple computer task a nd a human might make a mistake doing this.

Alex, X-Health.show: 12:47

What are the consequences of the mistakes?

Dr. Christian Federau, AI Medical: 12:50

One new lesion might mean that you should re-evaluate the case. Maybe you might need to change the therapy, maybe you might not, but you should think about this. And so if you do not see it, then the clinicians are not going to know. So nobody is going to know and that is it. So it is a missed opportunity to change the therapy. But, for example, for metastasis it is much worse. If you miss a small metastasis you might have treated it. Then it is too late later.

Alex, X-Health.show: 13:17

So we are speaking about cancer spreading now, into the brain.

Dr. Christian Federau, AI Medical: 13:21

Yes, exactly.

Alex, X-Health.show: 13:23

And I can imagine that these kinds of lesions have an immediate effect on how people behave, or even if they can see or not, if they have troubles moving or not. Why do you need the picture?

Dr. Christian Federau, AI Medical: 13:39

Yeah, well, for different reasons. Some of them might not make any symptoms. And then some of them – you might have so many symptoms that you might not realize that you have one more, or so you know it can be very complicated. So if you have already plenty of lesions, then one more, you might not always see it. So it is much more exact if a patient has a new symptom, you are immediately going to look in the portion of the brain that is responsible for the symptoms. So that is how you can find lesions too.

Alex, X-Health.show: 14:03

What do you do with the knowledge? If you know there is a new lesion or there is another lesion?

Dr. Christian Federau, AI Medical: 14:11

Currently, one new lesion for multiple sclerosis is the most important parameter to reassess the therapy. And so you know, you might increase the therapy, you might change the therapy. Maybe it is not the ideal one. So it is the best predictor of progression of the disease.

Alex, X-Health.show: 14:27

Let's also explain a bit what are the lesions and why do they form? Ah, no, we don not know why do they form, right, but they form on the white matter of the brain.

Dr. Christian Federau, AI Medical: 14:38

Yes, so it is autoimmune disease. So, basically, what happens is that your immune system is like a warrior, defending you constantly against outside, and if you did not have an immune system, you would die within, I do not know, a few minutes. So they are constantly fighting for you to survive. But often what happens is that the immune system starts to misunderstand a protein inside your body and is going to start attacking this instead of an external pathogen And that is what we think happens in in multiple sclerosis, for example. So basically this autoimmune reaction inside your brain is attacking the mailing in of your white matter, so it's a specific protein, and then and then this, this autoimmune process that happens.

Alex, X-Health.show: 15:23

And then in tumors, so in metastasis, this looks different on pictures or not?

Dr. Christian Federau, AI Medical: 15:31

Yes, yes, it looks different. So so you look at different MR sequence for this and yeah, they look totally different.

Alex, X-Health.show: 15:40

So what is in common, what I understand, with metastasis you also want to follow up on what was before. So basically, you say you have a picture from before and you have a picture now. You have two pictures.

Dr. Christian Federau, AI Medical: 15:52

Exactly. Interestingly, most of the time the neurologist spends looking at the image from follow ups. So maybe 90 percent of the time, everything is already known about the patient, and maybe 10 percent of the time you are looking at new diagnosis and you wonder what it could be.

Alex, X-Health.show: 16:06

Now, what you are saying, if we do the pictures, they might have been done from a different angle? Dr. Christian Federau, AI Medical: Yes, exactly. Alex, X-Health. show: And then you need to match what is the same on both and what's new?

Dr. Christian Federau, AI Medical: 16:20

Exactly.

Alex, X-Health.show: 16:22

Okay, and this is where Jazz comes in.

Dr. Christian Federau, AI Medical: 16:25

Yes.

Alex, X-Health.show: 16:25

And how does it work?

Dr. Christian Federau, AI Medical: 16:27

That's one of the components. So there are two main things. One is this, so basically, we made a user interface around the work of the radiologist, automating everything that can be automated. So the radiologist can concentrate on his work, so he doe s not have to look at the previous image, he does not have to coregister the image, which means putting them one top of the other. He does not have to count lesions, he does not have to write a report, everything is automated. Except what he really has to do is detecting the lesions and deciding is it the lesion or not. So that is one thing, and we already showed that just doing this, permits to reduce the number of errors a lot, a lot. And the other thing that we do is basically we extract the information that the radiologist finds from the image and we extract the data, and this data, we extract it, and extract it in the patient space, so we know where it was in the patient space.

Alex, X-Health.show: 17:22

What is the patient space?

Dr. Christian Federau, AI Medical: 17:23

So basically, we know where it was in the brain, like, let's say, in the frontal left, or the parietal right, the position in the brain, and this data we can then, reuse next time that the patient comes. We can reuse it. And so we know exactly where the lesions were last time and we give this to the doctor. The radiologist then just needs to look what is different. And we can do this forever, which means that if the patient has 20 follow-ups, let's say, then we know all this information. We're capable of showing statistics, currently it is impossible to understand what happens over 20. Like you know, there are text reports, nobody is going to understand this. But we are capable now of making statistics and presenting this data in a way that you can understand what happens.

Dr. Christian Federau, AI Medical: 18:09

And so, for example, like a slowly progressing disease, we can represent this and we can see that it is slowly progressing. We can make graphs showing that this is slowly progressing. And that is something that currently you do not even know. So we expect a lot of change in therapy. And once we apply this, we expect a lot of new knowledge, because this is currently not known. If you might have patients with multiple sclerosis that slowly progresses and then maybe a little bit more, and this might be two different categories that you might need to treat differently. Currently you d o not know, because this information is not available.

Alex, X-Health.show: 18:40

How long does it take for the platform, the AI platform, to find all the lesions or all the new metastases?

Dr. Christian Federau, AI Medical: 18:49

It's very fast. I mean, you know, because everything is automated, so it's very fast. Alex, X-Health. show: But we are speaking, like, I don't know, eight minutes or Dr. Christian Federau, AI Medical: No. I mean it depends a little bit which parameter you want, but it is between one minute or two minutes. And it depends on the computer. So if you have a good computer you can do this in 20 seconds. Alex, X-Health. show: How long does it take for a radiologist? Dr. Christian Federau, AI Medical: To look at the image? Alex, X-Health. show: To count the lesions.

Dr. Christian Federau, AI Medical: 19:17

Dr. Christian Federau, AI Medical: It depends how well he is going to do his work, it should take, depending how many lesions are there going to be, but it could take 20 minutes. So it speed ups the work a lot because it automates all these things that usually you have to do.

Alex, X-Health.show: 19:30

[MID-ROLL BREAK] This episode is brought to you by The X-Health. s how and me. So, if you like this podcast, be generous, hit "Follow, leave a review. That'll help me invite more such amazing guests for you, thank you. It's not about trust in AI, right? You tested the device and that's why you are that sure that it will count each and every legion, each and every tumor.

Dr. Christian Federau, AI Medical: 19:54

So the thing is, our software is not fully automatized. So you use AI, but the radiologist does the work. So that is one of the very special things about our software. So we do not try to replace the radiologist. The AI is just there to help him. But the radiologist can and should correct everything that he disagrees. So what comes out is the radiologist opinion. It is not the AI's. The AI is just here to help, but it's not a fully automatized thing.

Alex, X-Health.show: 20:21

How did you test it?

Dr. Christian Federau, AI Medical: 20:24

We have studies. Again, the user interface, which is not even AI based – we tested just the user interface, what I told you before, just simply automatizing everything around the radiologist's work, and we tested this. So we did one study,3-centric study in Stanford, in Paris, and the University Hospital in Zurich. We looked at previous exams and then we looked at them again with Jazz, and we looked how many lesions we described. Just with the user interface we found almost three times more new lesions for multiple sclerosis compared to the standard repor. Which means that all these patients, all these lesions were missed. I mean, that's the reason why they get the MRI. So, basically, you should use the software.

Alex, X-Health.show: 21:09

And if you miss them, or if you find these additional ones with Jazz, what does it change for the patients?

Dr. Christian Federau, AI Medical: 21:19

In one center in, Paris, they decided to reanalyze this. And in 40 cases, two per week, they put them back in the decision tree and anonymized, nobody knew and look what happened. And then they looked and in one fourth of the patients there was a change in therapy. So they decided to make it. So one fourth, which is a lot.

Alex, X-Health.show: 21:43

It is a lot. And it is also, I can imagine, you know, that's the experience but this is also what statistics say, usually doctors, are trained to make a decision and this decision is, final, because, well, that is their experience speaking. And now if you say in one fourth of the cases the treatment could be changed, that is kind of serious.

Dr. Christian Federau, AI Medical: 22:05

Yeah, I mean, of course they decide something different, it still does not mean that it is going to be every time a good decision, but still, you should know this information, at least, so

Alex, X-Health.show: 22:15

So you can base a decision on metrics.

Dr. Christian Federau, AI Medical: 22:18

Exactly. And on better information. So you should know this, you want to know this.

Alex, X-Health.show: 22:23

Why the platform is called Jazz? Dr. Christian Federau, AI Medical: First, of all, it's a cool name, do not you think?

Dr. Christian Federau, AI Medical: 22:27

Alex, X-Health. show: Okay, fair enough.

Alex, X-Health.show: 22:29

Dr. Christian Federau, AI Medical: That's one of the reasons.

Dr. Christian Federau, AI Medical: 22:31

And the second is that I am a jazz dancer. That is why, this is my passion.

Alex, X-Health.show: 22:34

You're a jazz dancer? Okay! H ow long have you been dancing?

Dr. Christian Federau, AI Medical: 22:40

I started in medical school, 15 years or something.

Alex, X-Health.show: 22:59

Okay. And where do you dance? Is it like a special contest or something?

Dr. Christian Federau, AI Medical: 23:04

No, not, at ETH, there are courses there.

Alex, X-Health.show: 23:05

You must be very good at it then? Dr. Christian Federau, AI Medical: No, nor at all? Alex, X-Health. show: How come? Dr. Christian Federau, AI Medical: I enjoy learning.

Dr. Christian Federau, AI Medical: 23:11

So I make some progress maybe, but yeah

Alex, X-Health.show: 23:17

So why do you do it? Why do you dance if you say you are not very good at it?

Dr. Christian Federau, AI Medical: 23:21

I think I really enjoy learning, so I think that is a constant in my life, I enjoy learning stuff.

Alex, X-Health.show: 23:28

You mentioned Stanford University. How did you start the partnership with Stanford?

Dr. Christian Federau, AI Medical: 23:34

I was in Stanford in 2015, I got the Swiss National Science Foundation Grand to go there, So I was visiting a scholar there.

Alex, X-Health.show: 23:42

And then, when you came back, you had this Jazz platform and you offered them for clinical trials.

Dr. Christian Federau, AI Medical: 23:50

Yes, I mean we did a lot, I did a lot of studies with them. I keep collaborating and have a lot of friends there and so we contacted them to ask then if they would be interested in working on this with us.

Alex, X-Health.show: 24:02

All right. What do they study there with Jazz? Is it like a similar to the one in Paris?

Dr. Christian Federau, AI Medical: 24:09

Yeah, exactly, it was the same. So we repeated the same findings, the fact that we found more lesions. We repeated it in three different centers and totally separate from one another, so no one knew about the other that they were actually doing this. It is good to repeat findings in different centers.

Alex, X-Health.show: 24:26

Were there any, I do not know, was any of these centers cautious about whether to implement it or not? Were there any dangers or anything they should be aware of?

Dr. Christian Federau, AI Medical: 24:40

I don't think so, because they all realize that it is a good thing that they can correct things, that they are in charge, in comparison to other automatic platforms that might make mistakes and you cannot correct them.

Alex, X-Health.show: 24:55

I can unerstand for a radiologist it speeds up the work, the radiologist can basically focus on analyzing and not on counting.

Dr. Christian Federau, AI Medical: 25:07

Yes.

Alex, X-Health.show: 25:08

But how is it difficult, or easy, to integrate the platform into their current workflows?

Dr. Christian Federau, AI Medical: 25:16

It is very simple. In the hospital there's something called the PACS, so it is an archiving system where you can exchange medical information, typically image, between different kind of devices. And so you can integrate this platform as a PACS node, it's called, and so you can exchange information as any other device in the hospital. So it's very easy.

Alex, X-Health.show: 25:41

So it is as easy as downloading a software, additional software? Dr. Christian Federau, AI Medical: Exactly. Alex, X-Health. show: Can you also use it on old images?

Dr. Christian Federau, AI Medical: 25:50

Yes, it is independent on this. It loads, it is called, DIACOM format the medical image saves as DIACOM and that is all it needs.

Alex, X-Health.show: 26:00

Let's talk a bit about the business side, right, because you have just received CE mark, before that you received ISO and the MD SAP medical device single audit program, so you are basically available for the markets, right? What's the market? Do radiologists want it?

Dr. Christian Federau, AI Medical: 26:22

Yes, we have very good feedback from radiologists and we hope every hospital in the world should be interested in working with us.

Alex, X-Health.show: 26:34

And do you already have any customers waiting? We have just spoken before the recording that you are going to many, many, many conferences now. Is that also like the start of the conversation, or what is your plan basically to reach out to those radiologists?

Dr. Christian Federau, AI Medical: 26:53

We discuss with everyone that we meet and we have a lot of interested hospitals that discuss with us. You know, hospitals are slow in taking decisions, so it goes slowly. But we have a lot of discussions and we hope to install soon our first platform.

Alex, X-Health.show: 27:09

And what is the market for AI in radiology? I mean, are there many competitors on the market or is that a worry?

Dr. Christian Federau, AI Medical: 27:18

A lot! It is very crowded. Basically, in medicine, AI applications are mainly done in radiology. It is really amazing how many. For example, FDA – hundreds of radiology software that has been cleared. And in other fields it is very, very, very little. And everyone in radiology is developing the software. So it is very crowded.

Alex, X-Health.show: 27:40

So how is Jazz different?

Dr. Christian Federau, AI Medical: 27:42

You know, I think one of the big differences that we have, we have user interface, the automatic tool. We are not trying to predict things, and you know we have developed a lot of cool features that we think are going to help the radiologists.

Alex, X-Health.show: 27:55

Like what, could you mention them?

Dr. Christian Federau, AI Medical: 27:57

For example, one thing that we did is, I'm sure you already scroll through image from top to bottom, that is how you look at image, but we developed a method that you can also scroll in time. So, rather than then scrolling in position, you scroll in time. And so that is very easy to then compare stuff. And you can also scroll in contrast too, if you want to see a lesion in a different contrast, you can just click one button and switch contrast.

Alex, X-Health.show: 28:21

S crolling in time means that you are watching one area and then Jazz changes the images for you.

Dr. Christian Federau, AI Medical: 28:30

Exactly. So you see one lesion and then if you want to see how was it last time, you can just click one button and then you see what it was last time. And if you want to see it two times ago, you click another button. So you can basically go back in time. So, for example, a slowly growing lesion, you are going to be able to see this slowly growing lesion very clearly. And so we have a lot of cool features that makes the software very easy to use and I think that differentiates us.

Alex, X-Health.show: 28:55

I also read that you are a founding member of Artificial Intelligence Committee, European Society of Neuro radiology, so I can imagine there are enthusiasts of implementing AI into neuro radiology. Do they also have their own startups, or do they look for solutions. What are the ways you work at this committee?

Dr. Christian Federau, AI Medical: 29:20

They are all other neuradiologists that are very interested in discussing this topic and applying it. They do not have their own startups, but they are very interested in this. And so what we do there is typically, we try to discuss and write papers about stuff that might be interested for the members of the society. For example, what is the ethical, what is the legal aspect of this? That is the last paper we did. So we try to find topics and research and make some kind of papers that help others of the society to understand what matters about AI and neuro radiology.

Alex, X-Health.show: 29:57

There is always this question about who owns the data, right? So what is the patient say? Do they have a say? How much can they own, can they "take home the images or the results?

Dr. Christian Federau, AI Medical: 30:11

This depends, in every country is going to be different jurisdiction. But in general, the patients own their own data and I think this is how it should be. And so you know, I hope. I think in general hospitals do not give the data easily enough. Patients should actually be constantly having their own data. But I think the evolution is coming. I think this is slowly coming that patient should have directly, constantly access to their own data. And they should not be in some kind of system in the hospital and you should not even have to ask about it, so you should have directly access to it.

Alex, X-Health.show: 30:44

And does the Jazz allow it to share the data with the patient"

Dr. Christian Federau, AI Medical: 30:50

Jazz is running inside the hospital firewall, so basically, it is inside the hospital. And so nothing leaves the hospital. And we are actually working on a platform, we actually want to make a platform where doctors could upload the data on the platform, where then the patient could have access directly to it. And so and so that is something that we actually work on, and for me it is very important.

Alex, X-Health.show: 31:13

What is there also in the future? Because you also said you are a bit different from other startups I spoke with because you did not have investors so far.

Dr. Christian Federau, AI Medical: 31:23

We have one. I mean we have one angel investor. Yeah, we will see, I mean we might raise some money later next year but yeah, so far we did not even look for money. The main reason is that I knew how to do most of the things, so I did not need to raise money. I thought I'd rather move forward and get this, you know, stuff done and have a product and find customers, and actually don't need money.

Alex, X-Health.show: 31:47

Are you saying you programmed it by yourself?

Dr. Christian Federau, AI Medical: 31:51

Yes, I mean I programmed a large part of it, another programmer helped me, too. Yeah, but most of it.

Alex, X-Health.show: 31:58

What did you use for it?

Dr. Christian Federau, AI Medical: 31:59

You mean languages? So it is programmed on Mac, so we use basically Mac languages, basically Objective- C and S wiftUI. And then we have like a base of C++. So most of the algorithms are going to run on C++.

Alex, X-Health.show: 32:16

Why would you even learn that as a radiologist?

Dr. Christian Federau, AI Medical: 32:19

I am a physicist too, that is why. So I learned about how to program and I did a lot of research where I programmed a lot and learned more and more And yeah, I enjoyed it. Because you can do these kinds of things, then you can solve problems.

Alex, X-Health.show: 32:32

So was it like a parallel, say, hobby, or was it a parallel project, that you were having on the side?

Dr. Christian Federau, AI Medical: 32:42

Parallel to my clinical work, my hobby was making research and so I spent my evenings and nights doing research and I programmed more and more. The previous research I was doing, I learned how to make a user interface to show the image, and then I learned more and more about this, how to do it, and then every time I thought how I could do this more efficiently. And so I start to automatize every step that I need to do my research, and that' s how it moves slowly into this project. But then when you start to do really a medical device, you have to start a clean way, of course. But but yeah, over 10 years, over 15 years, I learned constantly more about how to program and how to manipulate image. Alex, X-Health. show: And so it was always an image- based programming that you worked on?

Dr. Christian Federau, AI Medical: 33:33

Dr. Christian Federau, AI Medical: What do you mean with image- based?

Alex, X-Health.show: 33:35

That was with radiology in mind all the time, how to improve it?

Dr. Christian Federau, AI Medical: 33:39

Exactly, exactly.

Alex, X-Health.show: 33:40

So you spent 15 years and then you arrived at this Jazz platform that we are speaking about now.

Dr. Christian Federau, AI Medical: 33:49

I did research on a perfusion method in the brain and called IVIM. Alex, X-Health. show. What was that

Dr. Christian Federau, AI Medical: 33:54

perfusion method? Dr. Christian Federau, AI Medical: With MRI you can measure diffusion. And so diffusion is motion of molecules, basically. And so this can come from thermal motion, but it can also come from blood motion. If you use the diffusion sequence to measure diffusion, a small part of it is also coming from a perfusion. And so if you manipulate this correctly, you can extract this perfusion information. That is what I was working on, and so I worked a lot on this. When I was starting, people were not believing that this is even working, and so I was able to show to some extent that I was able to make some image, perfusion image, on this. It requires a lot of very precise work, and so I learned how to work very precisely with all this image, and I had to learn how to manipulate image in a very precise way. And all this knowledge then, finally, I used also in Jazz.

Alex, X-Health.show: 34:53

So did you do it in collaboration with any I don't know, university? We are now at ETH Entrepreneurs Club. So did you do the research for ETH, or was it like you're just part-time sitting at home?

Dr. Christian Federau, AI Medical: 35:09

During my career I switched several times the institutions. So I was in Lausanne and then I was in Stanford, then I was also in Basel, and I had my last grant at ETH, an Abizione grant from the Swiss National Science Foundation. So I was at different institutions.

Alex, X-Health.show: 35:30

And how do you see, in general, the AI-supported radiology, say, in five, 10 years?

Dr. Christian Federau, AI Medical: 35:38

I am sure it is going to evolve. It might not evolve as fast as people think. One of the reasons is that is the data that we discussed before. The major progress that you saw in AI is on fields where there is a lot of data available. In medical field is not a lot of data available and so I think it is very difficult to make this kind of progress if you do not have the data. But of course it is going to evolve. And I hope for the good.

Alex, X-Health.show: 36:03

So what are the other, say roadblocks on the way? I can imagine also, hospitals need to have, or clinics need to have, where to store the data, right, or clouds are available, right?

Dr. Christian Federau, AI Medical: 36:18

They have this PACS archive systems, so they have the data, but, yeah, they usually do not share with anyone else. So that is the major roadblock. The second thing is that hospitals, as we said before, are slow in taking decisions and it is not helping fast progress.

Alex, X-Health.show: 36:35

And where do you see AI Medical in these five, 10 years. What is the vision?

Dr. Christian Federau, AI Medical: 36:42

We hope we can develop the product capable of doing this data extraction in a general way for all pathologies. And that this is going to help improve patient outcome and our understanding of all these pathologies significantly. That is basically what we want to do.

Alex, X-Health.show: 37:04

So thank you very much, Christian, it was a pleasure speaking with you. And you guys, if you want to check AI Medical, please go to AI-medical. ch

Dr. Christian Federau, AI Medical: 37:23

Thank you so much, Aleksandra. It was a real pleasure talking to you.

Alex, X-Health.show: 37:27

[POST-ROLL] I'm totally impressed by the audacity of researchers turned startup founders, doctors turned entrepreneurs, or ordinary parents turned healthcare innovators. People battling the battles that no one fought before. For the eXtra health of the future. So if you see a startup posting on LinkedIn, show them some love, hit Like, comment, "that's fabulous. If you have a couple drops more of that altruism, follow the X-Hulse show, leave a review here. I'll be able to bring more of these visionaries to you. So a big thank you. You're awesome. See you next week.

Speaker: 38:00

The information in this podcast is for informational purposes only and should not be considered medical advice. If you have any medical questions, please consult your healthcare practitioner. The opinions on the show are Alex's or her guests. The podcast does not make any responsibility or warranties about guests' statements or credibility. While the podcast makes every effort to ensure that the information shared is accurate, please let us know if you have any comments, suggestions or corrections.