AI and Healthcare, an interview with Dr Jack Kreindler
Conducted by Matthew Miller (edited for style and fluency)
Who are you and what is your mission in life?
My background is in medicine. I specialized in high-altitude physiology, so if you wanted to climb Mt Everest you would come see someone like me. I paid my way through medical school by doing stuff with computers, ranging from designing user interfaces to working with Douglas Adams on The Hitchhiker’s Guide to the Galaxy online (the first Wikipedia effectively). I realized, at that time, that IT would transform all of healthcare.
After qualifying and practicing full-time medicine in a hospital for a year, I started my first venture-backed company and that was basically collecting lots of data from healthy people and not-so-healthy people working in big companies and finding out who would get sick and why. So, that was my first entrée into the world of health informatics and later that grew into deeper AI-driven problem solving.
Nearly 20 years ago I started my first company. Since then, we have come a long way in terms of much data there is available in healthcare and the power required to digest it and turn it into insight and actions.
I’ve delved into a whole bunch of things. For example, we have collected a lot of data from people in order to test and detect when things were going right and when things were going wrong. This would allow us to act just in time or far quicker than you could in medicine when you are measuring things really late.
One of the projects we set up was a spinout of my foremost institute here in London (CHHP) which was: how could we use wearables to earlier detect decompensation in people before they even had symptoms. This is not a trivial thing to do, and at that time all we were using was Excel.
Other personal matters, with respect to my family and unfortunate incidents of people getting cancer, led me down a path of thinking about how we could use AI to earlier detect these anomalies which should have been detected on the MRI scan.
Ranging from micro event processing to deep learning, image recognition, and pattern recognition, AI has become a key part of my life in medicine.
CHHP is a beautiful and impossible-to-replicate example of where you can get very clever people with big brains to work together to solve really difficult problems for, sadly, only a few people.
Sentrian is a first attempt of taking the type of pattern recognition which goes on in the brains of these people and allow many more people who haven’t had their training to make the decisions that they would, using machine intelligence.
My team can make better, earlier decisions than most AI can at the moment, since it is hard to get the data that we as clinicians can get as we are looking at a patient. There is some really subtle pattern recognition which goes on there, which I call ESP (extraordinarily subtle pattern recognition). In a sense what Sentrian is doing is giving the kinds of pattern recognition which is going on in the brains of very brilliant clinicians to nurses, patients, loved ones and caregivers which they otherwise wouldn’t have access to.
Also, it can help people like me improve our algorithms and challenges us to think about why we ignored a pattern, for example. Sentrian is one of the first examples out there of ‘clinician-directed machine learning’, but this machine learning is also helping direct clinicians. It is trying to bring the power of AI into practice in a hybridised manner.
BXR Gym, which just opened up in London (and which you are involved in), is incorporating data and data analysis in their gym. Is this the gym of the future?
It will be interesting to see how they, as well as all sport/fitness clubs adopt data and early detection of illnesses and ailments for general people who want to get fit, want to learn how to box better, want to learn how to avoid head injury, etc. It will interesting to see it.
It is the most cutting-edge of gymnasia. We will try to make it as data-driven as possible. In CHHP we have a 21st century gym which does not just have weights in it. We experimented with Milon equipment that measures exactly what you are doing with each of the machines. The question is: how can we have this sort of experience for all people, regardless of location and expenses?
The idea of a 21st century gym is to have a normal gym with things like TRXs, little cycling machines, and maybe even just a chair or door frame. Alongside that, you would have a lab (not an expensive one), but one which could objectively measure your body composition, your power output, oxygen consumption, etc.
It is taking what is only inside olympic sports institutes and elite sports clubs and turning that lab-based stuff into something basically every gym has. It would be bringing real sports science into the gym.
It is not possible right now for a few reasons: 1) the kit is expensive and 2) it is hard to get the right people to analyse the data. However, there are now cheaper ways to measure these things, which used to cost 100s of thousands. Now you can use a sticky plaster for much of this analysis. All of this will probably shrink to the size of a pencil case at some point.
How do you scale the kinds of brains to interpret the data? This really should not be too difficult with the proliferation of cloud computing and our ability to store, compute, and share data. The gym of the future will be able to fit in your pocket- your environment is your gym and your trainer is still a person if you want it to be. AI will help analyse the results and help you.
What sort of advice do you have for aspiring doctors of the 21st century as healthcare becomes more data-driven?
Medicine relies in many ways on having a universal language, so every doctor knows how to prescribe the same thing, for example (at least ideally). Locking that down and ensuring it spreads slows the whole thing down. If you make a change somewhere you must ensure it will all percolate out and gets settled before you can make a change. This can take decades sometimes. Sometimes the whole lifetime of a doctor.
On the other hand, there is a need to make transformative changes. How do you marry and face the challenge of these two needs? We have to move from relying on just the data that is gathered from trials and move toward real world data. Doctors need to know how to collect data for every patient in the real world after they have prescribed something. We have the ability to get that data and aggregate it so the next patient that does not fit the evidence-based data can benefit from the mistakes that we have made in the past and the cool things things we have gotten right.
If you disrupt things, people die (in medicine). Teaching about the importance of data, personalized responses, and AI to help us make important decisions is really essential. It should be in every curriculum. Stanford, St Bartholomew’s hospital, and a couple of others have started bringing data innovation and data (with emphasis on AI and real-world outcomes) into the curricula at the grassroots level, which is a great start.
Where else can AI solve big problems?
Medicine is more horrendously screwed up than other industries. Maybe it is because it must move more slowly- experiments take time in healthcare. People may lose money if you do an experiment in ecommerce or banking, but people lose life and limb if you have a bad experiment in healthcare. As a result it is slightly behind the times. If I were to pick one, which is sort of related to medicine, it would be food.
Knowing how best to grow (what to grow, where to grow it, and when to grow it) in order to meet demand and to reduce waste will be a big one. Energy might be a big one as well. We could learn better ways of how to generate energy. They are probably both ahead of medicine in regards to their efficiency.
What major trends and innovations have you seen in healthcare?
More accurate and consistent diagnoses, which we have already mentioned. AI is also helping new and established drug companies to more rationally work out what we need to build to address different diseases that are not as well addressed. Antimicrobial resistance is a big problem- there are many efforts in AI to design drugs which are more resilient to it.
Robotics, such as robotic surgery or robotic targeted radio-therapy would be another example.
The understanding of how cancers evolve and how genes mutate (and in what order) are also good examples of innovations.
It won’t be long before AI (or whatever you want to call it) influences our ability to make healthcare more affordable and equally-available for people. These are exciting times, it is a good time for AI.
What sort of AI-related material would you like to recommend to the community?
I love movies. One of the things I am doing is trying to get science to get much better press. I think screen and stage are the two ways we can do that. We adore the people on the screen and stage because they are really good storytellers. We are very bad at storytelling in science.
A little project I am working on now, is called Weird and Wonderful, a movement of people who want to communicate important things in science to wider audiences.
We have seen some really interesting films like Vanilla Sky, Momento, and Ex Machina which address this general theme about how humanity will be changed by our ever-more hybridised, melded world of human and machine intelligence.
In general, film and stage is going to provide the best way to understanding how important these topics are.
There are also a lot of interesting talks and lectures which I have attended, from The Future of Humanities Institute, CSER, Royal Society of Medicine, Singularity University, and WIRED Health. Some of the festivals such as the Hay Festival even have tents and stages where scientists are standing alongside famous actors and singers. I think we are going to see a lot more of that, bridging the knowledge gap in interesting ways. The Science Museum is also a great place (with their new robotics and mathematics exhibitions).