As part of our series of Industry and Medical Imaging Innovator Spotlights, radRounds has the honor to introduce and interview Radiologist, Dr. Leonardo Valentin. Dr. Leonardo Valentin joins NVIDIA, first physician on the team, with a key leadership role in medical imaging at NVIDIA, enabling the future for artificial intelligence in radiology.
radRounds: Why did you go into radiology?
Dr. Valentin: I began thinking about radiology when I was in college. I was majoring in computer science through a joint pre-med informatics track at the University of Miami and I had this concern that a medical practice would take me away from computing. On the other hand, my father is a vascular surgeon and I enjoyed the innovative aspects of surgery at that time. I began exploring career paths that would allow me to practice both radiology and vascular surgery. Throughout this process I fell in love with vascular interventional radiology which seemed to marry the best aspects of both of my passions.
radRounds: Why are you working at NVIDIA and not a radiology practice or academic practice where you could be doing research in AI?
Dr. Valentin: This is a very good question. In general, the people we serve expect medicine to improve with time. From testing new pills to new procedures—innovation is as important to medicine as it is to other areas of science and technology.
Innovation does not have to always come up in the form of technology, for example, as hard as it might be to believe, washing our hands was a great innovation in the prevention of diseases. In my case, I was already drawn to the technical interests due to my background in computer science and some early experiences with computing. Finally, innovation is a core component of NVIDIA's DNA and its environment make it an ideal place to develop ideas that will ultimately improve healthcare.
radRounds: What do you think about the future of AI in radiology? Many people are both worried and excited - what is your take?
Dr. Valentin: I definitely think the excitement is more prevalent in our field. Also, a healthy amount of skepticism has been historically helpful in ensuring new systems work better by matching our expectations, and this case (AI and radiology) would not be the exception.
It is important to engage at an early stage with the individuals that are going to be responsible of supervising and approving these tools in their own practices. I think radiology and medicine in general will continue to change but we are no strangers to change. Radiologists are particularly well equipped to deal with these changes since we have dealt with so many imaging modalities throughout the years. In that sense, AI will be an even easier adaptation from an intellectual standpoint for the radiologist when you compare it to the advent of MRI.
radRounds: How can a radiologist get involved in the future where AI will be a part of all aspects of healthcare and of course radiology?
Dr. Valentin: I think we will start seeing better resources to accomplish this both in the educational realm and in the implementation/computing realm. This is a very important part of my role at NVIDIA. I want to be that link that bridges the gaps. Those gaps might come in many forms, from AI concepts to infrastructure limitations. I want to emphasize that the GPU computing revolution will be very similar to the personal computer revolution. We need to empower the end-user with this technology for important contributions to be made in this arena. Radiologists will play an important role in consuming the technology, but also in developing and perfecting it.
radRounds: What are the short term applications of AI? (where will it make impact first)
Dr. Valentin: In the short term, AI is already facilitating the tasks we are already doing, for example x-ray interpretation, segmentation and tumor-board decision analysis. In the long term, it will allow us to accomplish things we can't do without it. The acceleration of applications using GPU technology is another important part of NVIDIA's mission and we see this technology present in CT scanners and PACS. The problem I see with AI is that we often have it in our environment and we don't really notice it. This might be because when AI "solves" or addresses a given problem, the solution is often no longer considered AI. Therefore, it can often be hard for us to really "perceive it" in the way it is portrayed in Hollywood movies. But this might be a good thing! It means the solution was fully accepted by the stakeholders.
radRounds: What are some resources you recommend for someone wanting to learn more about AI? Both technical and non-technical resources?
Dr. Valentin: There are some very useful resources out there, and it might be very hard to choose which one to follow. On the textbook side of things, I still think Russell and Norvig's AI a Modern Approach is a good place to start, at the very least, the introductory chapter is very good at bridging popular perception to reality. In the popular science arena, there is a book, Automate This, that describes the history of many algorithms we now take for granted in our everyday life. But finally, I must stress, that the best and perhaps only way to really learn any subject matter is to choose a problem (large or small) that you truly care about and see how you can apply the lessons to that problem. There is no "golden road" but this is as close as it gets!
radRounds: If someone wants to speak to you and the healthcare team at NVIDIA, what should they do?
Dr. Valentin: They can reach out via email@example.com or informally through social media.
radRounds: What else would you like to add for radRounds audience?
Dr. Valentin: Stay engaged and try to read technical articles. Don't limit yourself to what you see in the news. Approach the researchers working on AI/ML since your input is very valuable to all of us. Advancing technology is a team effort.