This week we spoke to Daniel Pinto dos Santos, Deputy Editor at the European Society of Radiology’s flagship journal, European Radiology, and a Board Member at the European Society of Medical Imaging Informatics (EuSoMII). Pinto dos Santos is currently a Senior Radiologist at the University Hospital Cologne and University Hospital Frankfurt in Germany.
What is your background/experience with artificial intelligence and what first attracted you to the topic?
I have always been interested in computers and IT topics from a very young age, so when during my residency people started talking about AI and neural networks, I was immediately intrigued. Most of my experience and knowledge is self-taught, though.
What are the biggest challenges to AI adoption in clinical practice?
I think the main challenge currently – at least in Germany, but I would guess in other countries, too – is an economical one. The licensing fees for some AI models are often close to an annual salary of a junior doctor. And from the perspective of an institution’s management, the return on investment is often unclear or at least difficult to calculate. Therefore, many are hesitant to invest in AI if they don’t know how much it will save in the future. Apart from that, the second biggest challenge in my opinion is the lack of evidence that AI adoption really benefits the patients. In the coming years, I think it will be our responsibility as radiologists to prove this!
Give us an example (or an educated guess) of what you think AI will be able to do in 3 years. What about in 10 years?
I think in terms of diagnostic applications like the applications we see today for fracture detection, stroke detection, etc., AI will probably improve a bit in the next three years. But those AI applications perform very well already. My personal hope for the 10-year timeframe would be that companies realize that the biggest potential for improvement lies in workflow-oriented applications and not in diagnostic ones. Hopefully we’ll see more commercial AI applications for things like body composition analysis and organ segmentation, support for report generation and structuring, support for optimized scheduling and procedure planning, etc.
As a radiologist, what is your advice for younger colleagues wanting to dive into the topic of Artificial Intelligence, Machine/Deep Learning, and/or Radiomics? What are your tricks for staying up to date in this fast-evolving field?
I think there are excellent resources to get started if you are interested in those topics. There are numerous online courses you can do – for example, the ESR’s Masterclass in AI. Currently, within EuSoMII, we are planning to compile a database of such courses and resources so interested radiologists can easily find something suitable for themselves. But, of course, keeping up with the cutting edge of what is happening is still challenging after that. Personally, I get a lot of news on recent developments in artificial intelligence through social media and find it extremely helpful. Although, of course, one needs to be cautious here and there with social media and it might not be for everybody.
