Welcome back to Beyond the Patterns. So today I have the great pleasure to introduce Dr.

Christian Baumgartner, who is currently heading the Machine Learning for Medical Image Analysis

Group at the University of Tübingen. Christian completed his PhD in 2016 under the joint

supervision of Professor Andy King and Professor Daniel Rickard at King's College London in

the School of Biomedical Engineering and Imaging Sciences. He further pursued his research

interest in machine learning for medical images as a postdoc, first at Imperial College London

with Professor Daniel Rickard, then at ETH Zurich with Professor Ender Konokoglu. Christian

worked then as a senior research engineer at PTC Wulforia Zurich for a year before joining

the University of Tübingen in his current role in February 2021. So today I have the

great pleasure to announce a presentation that is entitled The Next Decay in Automated

Medical Image Analysis. And I actually did this recording when I was traveling and you

can see that during the Q&A session I'm actually walking in the streets of the wonderful town

of Leeds. So I hope you enjoyed the presentation, the Q&A session and I very much recommend

having a look at this wonderful presentation.

Yeah, thanks Andreas for the introduction and also for inviting me. I'm very happy to

have this opportunity to talk to you guys a little bit about my research. So as Andreas

already mentioned, I had this sort of break in my academic career. So recently I've kind

of been thinking a lot about, you know, what we should be doing and what is actually kind

of missing and what we have already done. Well, before we move on, so just briefly,

so in this talk I want to on one hand show you some of my past research, speak about

some things that I perceive to be limitations of the field in general, and then maybe try

to give a little bit of a perspective. So I think this quote has become quite famous

in the meantime. So Jeff Hinton in 2016 said that people should stop training radiologists

now. It's just completely obvious that within five years, deep learning is going to do better

than radiologists. So I imagine he has probably come to regret this quote by now a little

bit. But it is actually now exactly five years since he made this quote. So maybe it's a

good moment to kind of reassess where the field is at this point and what the future

perspectives are. So I want to start with this Gardner hype cycle. I don't know, maybe

some of you have heard about this already. So Gardner is a consulting company, a technology

consulting company, and they came up with this sort of typical trajectory of hype that

many technologies undergo. So I think there is a lot of criticism also about this chart.

So take it with a grain of salt, but I will try to sort of use it as a red thread through

this presentation. So what happens often is that there is a technology trigger, like let's

say the advent of deep learning, and then there is a peak of inflated expectations.

Then all of a sudden it doesn't work, the bad press starts, the public gets on the other

side of the argument, and there is this trough of disillusionment before you kind of make

the technology work really and you end up in this plateau of productivity. So a good

example is in 1995 everybody was talking about interactive agents, chat bots and things like

this, or intelligent assistants. And there was a huge peak in 1995, and then Microsoft

implemented this assistant, this clippy guy. I don't know, maybe this audience is too young

to remember, but this was a complete failure basically. Everybody was annoyed about this.

And this interest in this sort of interactive assistant technology stopped until very recently,

and now we have Siri and all of these things that kind of work, or at least better than

this guy did. So yeah, obviously I think medical image analysis is undergoing something similar,

and I will be talking about this a little bit. So as I said previously, I want to explore

how machine learning can be used in medical image analysis. And since I've been thinking

a lot about this, I want to give you a little bit of a structure that I personally use to

categorize research. And then I want to try to make some predictions about the next decade

and, you know, keep in mind, this is a quote by a famous baseball player. It's tough to

make predictions, especially about the future. So, you know, we're all kind of guessing at