Pure Storage AI in Healthcare
Episode 456

Esteban Rubens, Global Enterprise Imaging Principal at Pure Storage

Pure Storage AI in Healthcare

Streamlining workflows to improve outcomes

Tools We Use and Recommend

Save time, Make a Bigger Impact

Check them out

Get The Latest In Your Inbox

SUBSCRIBE

Pure Storage AI in Healthcare

Episode 456

Best Way to Contact Esteban:

Linkedin

Company Website:

Pure Storage

 

Pure Storage AI in Healthcare with Esteban Rubens, Global Enterprise Imaging Principal at Pure Storage transcript powered by Sonix—the best audio to text transcription service

Pure Storage AI in Healthcare with Esteban Rubens, Global Enterprise Imaging Principal at Pure Storage was automatically transcribed by Sonix with the latest audio-to-text algorithms. This transcript may contain errors. Sonix is the best way to convert your audio to text in 2019.

Welcome to the Outcomes Rocket podcast, where we inspire collaborative thinking, improved outcomes and business success with today’s most successful and inspiring health care leaders and influencers. And now your host, Saul Marquez.

Saul Marquez:
Welcome back to the podcast. Today I have the privilege of hosting Esteban Rubens. He serves as the Global Principal for Enterprise Imaging at Pure Storage, where he’s responsible for pure solutions, strategy, market development and thought leadership in that area of healthcare. In addition to the traditional areas of enterprise imaging, he focuses on the intersection of artificial intelligence and medical imaging with a particular emphasis on the role that I.T. infrastructure plays on both research and translational applications. He has spoken at A.I. and Imaging at the National HIMSS Conference, Regional HIMSS chapters, AI med, among other well-known conferences. He’s got almost 20 years of experience in the storage industry and fifteen plus years experience in healthcare tech. Really looking forward to having a discussion with him here today. And as you guys know, each week we sort of took on this new approach. We’re going to be telling stories, success stories and health leaders stories. And today, Esteban is our hero of the story. But before we walk through this, I want to just open up the mic to him to welcome him. Welcome, Esteban.

Esteban Rubens:
Thanks Saul, really great being here.

Saul Marquez:
Yeah, it’s a pleasure to have you. So tell me a little bit about why you decided to get into healthcare. Why storage? Why flash storage? Tell me more about that.

Esteban Rubens:
It’s interesting and also somewhat circuitous. So I was born in Argentina in a pretty medical family. My father’s a cardiologist. My grandfather was a cardiologist. But I was kind of the black sheep in that sense in that I didn’t want to be a physician. Immediately I was more drawn to science and math. So I ended up getting a scholarship for college in the US. I studied your math, but that I always had something going on with medicine, so I became basically MP. In college we had a volunteer squad and so that was sort of a great way to do all things. The other things I was being a geek. I mean, mostly I think first everything else second. So I got into two startups, even though I lived in Vermont and New Hampshire. I got into data storage with an Australian company actually that did solid state storage called Platypus Technology was or a and that was somewhat coincidental. But that sort of set me up for for this path where I did storage for a few years for a bunch of startups. Then I went to work for one of the PAX vendors. They were looking for storage geek to help them out and their customers with storage. Of course, very important to have good storage and plenty of storage when you’re doing packs and BNA and Rasam or the traditional imaging applications in healthcare. And then after quite a few years, I basically came home back to storage, a pure cure called and I just couldn’t say no. You’re an amazing company doing things in a different way. It’s been a wild ride and we’re just getting started. So it’s been pretty awesome.

Saul Marquez:
That’s brilliant, man, by the way. I have a soft spot for Argentina. And then many, many times I did a stint in med device where I was managing Latin America and spent many times and Buenos Aires, loved it there. What hospital did your folks practice in?

Esteban Rubens:
I think a lot of public hospitals, private with the goal, you know, some authority as there is certainly I can’t recall all the names, but I know.

Saul Marquez:
There’s a bunch of them.

Esteban Rubens:
There were, yes.

Saul Marquez:
That’s awesome, man. Look, I love your story. You sort of medical family got into math, storage gig, now doing some pretty impressive stuff. I mean, the area of A.I. in the cloud is something that a lot of people are trying to to use with imaging. And so tell us about the goal that you guys have been after some of the problems that you faced to achieve that goal.

Esteban Rubens:
Yes. Starting point is that A.I. in the form of deep learning, which is kind of a subset of A.I., has been around for a long time, but it’s now finally useful because technology makes it useful. So the algorithms and the math tools were there, but we just didn’t have enough computational power really readily available. We didn’t have fast storage, wouldn’t have fast networks. So finally now in the last I would say five or so years have g.p use for compute bases so that the stuff that came up for gamers and companies like Can Vidya figured out, well, this is great. We can. It’s a perfect fit for deep learning model training. We now have all flash storage, which is what we do, and you have really fast network. So when a lot of people remember datacenter networking, they think about one gig and gig. Maybe some people get into 40 gig. Well, one hundred gigabit per second networking is now fairly commonplace. It’s not something that only the big telco carriers use. So we have that amazing intersection of the right technology with these basically very well-understood tools that make this kind of research possible. This type of application of of artificial intelligence and I just want to say one thing about that palm. I really don’t care about terminology. A lot of people, I think, who are active in AI prefer to use the term augmented intelligence. And it’s it’s not just that choosing a different word to be cute. A lot of people have fears about A.I. and they’re understandable. And certainly we have you know, you had the case of that M.I.T. professor who was saying, you know, we should stop training radiologists because they’re going be replaced by. It’s ridiculous. Definitely could happen.

Saul Marquez:
I agree with you completely.

Esteban Rubens:
Yet it feeds some some fears. So I think it’s very important, especially in health care, to have this notion that we need technology because health care workers, whether physicians assistants, nurses, whoever, they’re all overworked technologists and imaging around the world, there’s a lot of burnout. And even though it might be ironic to say what technology, more technology is the answer, because burnout certainly part of the burnout is due to technology. There’s a lot of discussion about a charge and how they do it. No need to get at that. If you look at AI, it really especially imaging has the potential of alleviating a lot of these very mundane routine tasks that contribute a lot to burnout. So back to the idea of augmented intelligence, the concept this is not artificial. This is not going to replace you or take your place in any way. It’s to give you superpowers, superhuman powers so you can do your job better. I’ve heard a lot of really smart people at conferences talk about how a lot of doctors just want to go back to being caregivers, healers, whatever you want to say. There’s a lot of discussion about the lost art of the physical examination, especially in the U.S. and many places in Europe, Western Europe, doctors, and I’m looking at their patients there staring at the screens anyway. There’s the whole thing. Right. So that’s why augmented intelligence is something I prefer to use.

Saul Marquez:
Yeah. You know what I did, too, man. And the bottom line is, I mean, like boiling it down to the most simple form. You’re writing an email and G-mail, if you have G-mail and now it gives you some some suggestions. You don’t have to take the suggestion. You’re still the one writing the email. That’s basically what we got here. And the simplest form and I love that you went there, Esteban, because it is it is an augmentation, not a replacement. So as as you and your firm work to improve outcomes and make business more successful for your customers. Tell us about a typical roadblock and then how do you guys usually come in to help? What’s the plan to achieve success here?

Esteban Rubens:
A roadblock is that people think this is something that’s really hard and unattainable. It’s like all yeah, we hear about a it’s going to buzzworthy. It’s like we’re so far from doing that. You know, the site doesn’t even know where to start. I mean, many times that people are interested in those cases. You get that in the bad case. That’s a number two thousand seven hundred fifty five on our list of priorities that we just don’t get. And so the first thing to say is you really can’t stick your head in the sand, even though we’re all too busy. There’s no doubt about that, especially in health care, not just the reimbursement cuts and budget issues and political uncertainty, especially in the US. Totally aware of that. The thing is, this is one of those times in history in which something is happening, whether people like it or not. Right. And so that that’s kind of the inescapable effect. There are hundreds of companies doing AI in health care, a lot of them in imaging. Every major player is going that direction. If you look at companies like Philips Spice, they divested everything that wasn’t healthcare. They are now healthcare company and basically a health care AI company. So the other stuff they did is gone. So the important thing for that that I like to discuss when people think this is not even close to being relevant to them, is that, first of all, they’re sitting on a goldmine of data. Everybody has any health care institution that has done digital imaging. And that’s my focus area for more than a couple of years. And basically in the US, everybody is fax. Everybody’s had fax for close to 20 years, 50 whatever the number is. And it’s not that I’m talking about the goldmine of data in terms of monetizing it, although that is one of the aspects. But a gold mine in terms of improving outcomes for their own patients. And that is something that really can’t be overlooked. So we now have and it’s not just in radiology and cardiology. It’s a lot an apology. You know, there’s there’s now this notion of computational pathology in dermatology and ophthalmology in so many areas of healthcare are now heavily image focused or image centric and have generated images for a long time, whether they are organized in a rational way or not. You know, many times and maybe ophthalmology, there’s jpeg stored in somebody’s computer and they’re not clinical system. Not bash on them. But you know what I’m saying? Like radiology. Good at that. So going from that really fact that this is happening and that you really shouldn’t ignore it. You have to realize that everybody knows what data you have. So we I personally have talked to a lot of customers prospects in the US who have been approached by the large tech behemoths, both on the IP side as well as the healthcare side with offers to maybe digitize their whole pathology slide collection for free, if only they’ll share the data. Things of that nature. So very important for people to essentially take charge of that and realize what they’re sitting on. So that’s number one, developing some kind of policy around that data – not so data governance, data protection, how the data is used, but most importantly, really having a conversation internally about how that data can benefit their patient population and their physician population, which are huge things that a lot of people don’t talk about. So once you start talking about that, it’s a little bit easier. And then on top of everything else, we now, again, going back to the earlier point about the tech aspect being mature enough to really support this. You don’t need to deploy racks and racks and racks of very complicated stuff to start doing something with that data training models, it’s all very democratic, which is really, really great. So people start they build a P.C. and they wanted to use that. They steal from their kid who is a gamer. Right. That certainly resonates with me and with people here. They have a couple G.B. years later on the house. And then you can teach yourself data science. There are things out there like Coursera again. And Mikey has really great stuff where you can learn stuff online and get a really good education. And then the open source movement is a huge part of this because you can also download models and then start training them without having to start from scratch. So this this really amazing aspect of the democratization of A.I. in general, healthcare in particular is what what I tend to focus on because most people maybe not most. Some people are not aware that this is the case and that you can really start small and you choose a problem to solve and you can just go after it’s not boiling down the ocean.

Saul Marquez:
Esteban, how about and I think that’s why I think it’s a great promise Right. for folks to be able to do. There is, though, a lot of people that aren’t as comfortable with that. I put myself in that in that category. I could go and learn how to do these things. But I also have areas of focus that if I go and try to learn something like this, it’s going to distract from the areas of focus for the business. So what do you recommend for the leader that wants to action on this but can’t do it on their own?

Esteban Rubens:
That’s a great question. I think imaging medical goods in particular is really a great area for this because we also have a lot of geeks who become radiologists and cardiologists. So it is. And frankly, I’ve met, you know, in my 15+ years of being in this medical imaging world. I’ve met so many radiologists and other physicians involved in this world who were engineering that minors are major. So are they into math other yet they have a go to a makerspace. So the advice for the leadership is find the people who are interested because you’re going to have a…

Saul Marquez:
You’re saying that there are people in your organization today, if you’re listening that are geeks, they’re going to be interested in this.

Esteban Rubens:
You just have to find them. Yeah. And really, it actually is true,.

Saul Marquez:
Yeah, it’s a great callout.

Esteban Rubens:
But true. And sometimes I think people get lost in the day-to-day grind and maybe maybe in the hospital. Maybe the leadership doesn’t know that doctor X, Y, Z. She was an electrical engineering major. Or that somebody is out building stuff with raspberry pies or arduino is because it doesn’t come up in conversation when you’re talking about what’s going on with whatever hospital or hospital or infection rate.

Saul Marquez:
Right. Right.

Esteban Rubens:
But again, my point is these people are out there, find them and let them do what they want to do, because that can open up amazing opportunities for the whole organization.

Saul Marquez:
Love it. Great. Great call to action there. And so what happens Esteban if people don’t heed this course. What’s the tragic failure?

Esteban Rubens:
I like the way you’re posing that question because it’s not about doing it like I can’t sell you the fire and brimstone. But yeah, it’s about tragedy. It’s about patients not getting help that they could otherwise help. You know, it’s about diagnoses being delayed or not being delivered as quickly as they can. It’s about maybe physicians who get burnt out and leave the profession because they just are not doing what they thought they signed up to do. There’s just it’s mostly about avoiding bad things. That, of course, there are a lot of positive things. Right. But in terms of the human outcome, it’s not doing things that you could have done that would have benefited both your patients and your caregivers whichever whatever role they have. And also, certainly there’s a bottom line aspect to this right.. There is money involved in that. Certainly. I totally get it. We’re not that’s not number one. But it’s also especially, again, unfortunately, us. It’s it’s a big deal. So don’t underestimate that aspect, that this is something that can contribute to your bottom line. We have a lot of health care organizations in the US who ended up marketing some of this stuff that they created or derive really tangible benefits from doing AI with their data, whatever, whatever it ends up being, you know, it’s a very general term, but essentially turning their data from a boat anchor, you know, something that’s kind of annoying, especially imaging data. Why are we keeping this 20 year old chest x rays? No one cares about? Well, they are incredibly valuable, even if you have a bunch of normal things with no pathology. So that that realization that basically data is never cold, data is always useful, particularly to get insights in healthcare, particularly in imaging. So we have get to this concept that sometimes people scoff at this. But I think more and more we’re seeing that it’s being people realizing that it’s true that there is no such thing as cold data.

Saul Marquez:
Esteban, great illustration here. What’s there to gain? Tell us maybe a story of one of your customers or somebody that you know that’s been able to leverage this plan that you’ve given the listeners? Right. find a way to do it on your own open source. Find the technologies to get you there or find the people in your organization to get you there. Give us an example of somebody that has.

Esteban Rubens:
Yeah. So there’s an example. I’ll keep the names confidential because I talk to anybody about this or on their permission. But this is a research university in the Midwest, and it’s actually not a unique case where there is a very fascinating collaboration. So it turns out that the chief of radiology, informatics was an engineer. The first training this this person had was as an engineer. So they now have a collaboration between the medical school, the university hospital that’s affiliated with the university, of course, and the School of Engineering, where they’re doing data science within the hospital, using tools developed by the data scientists at the School of Engineering that then benefit the overall population of that state. So in this is, it wasn’t a given because you have a lot of other university hospitals that are affiliated with medical schools and have those universities have schools of interior engineering where they’re not doing it because you didn’t have the people who said, no, we really should be doing together. And it also goes back to this kind of old thing now about interdisciplinary studies and collaboration, you know, to breaking down barriers. Why should the medical school not be talking to the school of Engineering about data science? And this is also a very interesting point for universities, because data scientists are in extremely high demand. It’s one of these things that maybe, you know, a lot of people talk about that extend a very fast growing profession. Salaries are off the charts because demand far outstrips supply. And of course, universities can do more to get students interested and then certainly to get students in the medical schools and other parts of the university, at least aware of the science. So it can become this really virtuous cycle of amplifying what is good about this kind of AI research and helping people with their careers too.

Saul Marquez:
Yeah. And again, another example of how using your current resources to get things going is the right way versus hiring out this task and spending tons of money. Great, great example, Esteban. And it’s been really interesting to have this talk with you. I mean, if you’re a health care thing, you’re listening to this thinking about how to get this going, you just need to get it done. You’ve got some examples today, and even so, we focused on providers today, but how about health care companies, industry and payers?

Esteban Rubens:
That’s I would say there is somewhat more activity there. There’s hundreds of startups that are doing different AI things with images. The FDA in the US has a specific approval path for A.I. based software. So the government’s it’s not just in the U.S., it’s Europe, it’s China, it’s Australia. Everywhere there is this realization that this is actually a big deal. And there’s also this. What I’ve heard at some conferences that these regulators are saying we’re going to approve things based on the science. We understand that there’s a lot of liability issues, but we’re leaving that for the legislature to figure out. So my takeaways, they’re not trying to stop progress because of liability concerns. That’s something that needs to be addressed and during good talk for hours about that. But it’s happening. So it’s, of course, not just startups, although that’s very interesting because since right by 2017, there’s been fairly geometric growth in or even exponential in the funding of private equity and venture capital into companies that use AI in health care in general and certainly medical imaging in particular. Now we are into the billions of dollars, a quarter of money, private money being invested into these companies. And then all the big established that they were saying before, all the big established players that you can think of are either planning or testing or deploying products with some kind of AI, you know, sort of there’s all sorts of different things. In radiology, we have things like workless prioritization using AI with natural language processing, which is not really Excel analysis. But then we have just straight out the second opinion type of information being displayed as overlays where the radiologist can agree or disagree with that finding that’s provided by the software. But it’s also not something that they’re forced to do. This is everywhere. You scratch the surface and you go to Arseny Arseny until 2016 I think had no I speak off and then started 2017, just boomed and they started a machine learning showcase that doubled in twenty eighteen and now and nineteen. If you’ve been to Arseny, the hall of basement of the North Hall at the McCormick, it’s going to be for A.I. companies, the same at HIMSS. You know you can walk around hands without hitting somebody talking about AI just name it, it’d go on and on. But for companies this is it’s really a revolution because it’s created a whole market segment. Really AI in healthcare

Saul Marquez:
Love it, huge opportunity here, folks, if you’re not taking advantage of it, you’re missing out. And don’t miss out. Step on it. Really appreciate you sharing the insights that you have and in the work that you do. What closing thought would you leave us with?

Esteban Rubens:
The first thing you said, this is really a global effort. This is not U.S. based. It’s not Europe based. There they’re really interesting papers that have come out in the last few months that show that even if you have a really well-trained, deep learning model that helps a lot, in some cases you have to train it with all sorts of data. So if you’ve trained a model with data within from the northeast of the US, it may not be great for rural India or any other pair like that, but you can think of. So this is a global effort that everybody should be involved in. There are some voices which I really like, maybe a little optimistic that that this is something that’s really going to help global cooperation because everybody has a stake in making sure that this will be better, actually. Remember meeting a doctor from Senegal at a conference and just going back to your earlier question. He trained himself into doing this and then he’d start a company in Senegal. Doing this is really a wonderful thing. And again, thinking about the patient benefits is amazing. The speed that the increase in speed of diagnosis that these technologies allow for is mind boggling. And also things like avoiding unnecessary biopsies or avoiding the sequencing of a tumorous genome to know if certain mutations present when instead you can feed MRI images to an algorithm and get the same answer. But in the fraction of the time with nothing, nothing traumatic. So it’s just really it’s amazing.

Saul Marquez:
Huge promise.

Esteban Rubens:
Absolutely.

Saul Marquez:
Huge promise. Well, Esteban, really appreciate your time here today. If the listeners wanted to continue the conversation with you. Where can they reach out or follow you?

Esteban Rubens:
I’m on Twitter. @pureesteban on LinkedIn is where I do most of my activities. And certainly. Any conference such as RSNA and HIMS and if you go to the pure storage health care site, that’s another way to find me.

Saul Marquez:
Outstanding, Esteban. Hey, listen, just when I’ve give you a big thanks. Appreciate you sharing your story and your passion for the AI in health care.

Esteban Rubens:
Thank you so much. It’s been great.

Thanks for listening to the Outcomes Rocket podcast. Be sure to visit us on the web at www.outcomesrocket.com for the show notes, resourses, inspiration and so much more.

Quickly and accurately convert audio to text with Sonix.

Sonix uses cutting-edge artificial intelligence to convert your mp3 files to text.

Thousands of researchers and podcasters use Sonix to automatically transcribe their audio files (*.mp3). Easily convert your mp3 file to text or docx to make your media content more accessible to listeners.

Sonix is the best online audio transcription software in 2019—it’s fast, easy, and affordable.

If you are looking for a great way to convert your mp3 to text, try Sonix today.