Troy Long, IOE PhD Student

troy long interview photo

Troy Long’s research is focused on treatment planning for radiation therapy. He has worked on lexicographic optimization for IMRT with an emphasis on sensitivity analysis. Currently he is working on stochastic adaptive treatment planning and beam orientation optimization for IMRT, as well as other stochastic optimization problems in radiation therapy treatment planning. He also is working on non-coplanar VMAT treatment planning with Massachusetts General Hospital. He is a fifth-year candidate in Industrial and Operations Engineering and is interested in large-scale optimization, modeling, and algorithms with healthcare applications. His advisor is H. Edwin Romeijn.

Disclaimer: This is a transcript from a verbal interview

What kind of research are you excited about?
The trend for engineering healthcare is to look at existing methods and improve upon them. Now, we are at the point in research in which the core of what people are working on has caught up to what has been used in practice. What we are starting to do is develop algorithms for either non-existent technology or hardware that will be approved in the near future so that whenever they are implemented, there is a better chance they are going to be effective just because engineers have been working on it in the past.

I do radiation treatment planning and am working on an algorithm for a beam that continuously moves around the body in non-coplanar space, but this machine really doesn’t exist right now. It should in the next few years and so we are future-proofing things. I think that is an exciting transition point in healthcare research. Instead of just being physicians, there are also engineers that are on the forefront of actual treatment.

What research problems are keeping you up at night?
This is actually more of a problem of the state of healthcare research than my specific research, but being able to share data in engineering is very important and it is very difficult whenever you are working with patient data. Say I’m working with collaborators at the University of Michigan and I am publishing a paper on beam orientation optimization for radiation therapy. The cases we use belong to the University and they are not comfortable with us putting those cases out for other people to test their algorithms against. I think that now we are in this era of big data where we can save everything and storage is cheap that with a little bit of foresight into who is going to be using this data, we can share it in a better way. We should have data scientists working with the hospital and the researchers to build up these suppositories of data or a framework that will allow easy data sharing that people can actually have their inputs. That’s the thing that keeps me up at night. There are a lot of things we could do, but because of this almost clerical issue, we can’t.

How do you think engineers and healthcare professionals can work together to improve patient care?
There are lots of answers to that! Just work together is the first thing. I was talking with a collaborator and he drew something on the board that was very interesting. You take a rectangle and you draw a line from the top left corner to the bottom right corner and each of these areas represent the amount of involvement of healthcare professionals and engineers. You can kind of scale a vertical line based on the involvement, where the project is, and how integrated each of them are. That was the middle step. The first step is that you have these two separate entities and they interact and now you have this cross over when you are interacting, but where they really should be is two long horizontal boxes along the entire process where the team doesn’t consist of one physician and eighteen engineers that are working on things, but there is more of a collaborative effort more than a back and forth. I think that is where it should be going.

What surprised you most about working with healthcare professionals?
Other than how early all of them get up in the morning? I think there is such a big divide between the establishment and the new people who are willing to have this outside information. You can see in these larger faculty meetings that certain people make comments saying, “We really don’t need this,” and other people saying, “Oh yes, this is the future of things.” It’s just as when engineers had a problem interacting with businesses even twenty years ago saying, “If you use these types of models to show where you should put the next Wal-Mart, then there is a higher chance the store will do better,” and businesses responding, “No, that’s nonsense” and then they kind of turned to that kind of thinking. It’s going to take some time, but I think we are going in the right direction. There still exists that prejudice against the non-clinical person. It is a lot less prevalent in radiation therapy because physicians need physicists or else everything breaks, but if you go into other more direct care things, you run into those kinds of problems.

How has working with healthcare changed or altered your way of thinking about healthcare problems?
I see almost everything like a math optimization problem at this point. Quality Adjusted Life Years or QALYs are used to calculate the value somebody’s existence numerically. It is a weird way of thinking about things, but I have found myself thinking like that now working in healthcare. It is weird how that is the focus of all of these problems. The life years of a patient still reduces to money, unfortunately.

How has working in healthcare changed or altered your approach to tackling healthcare problems? 
It hasn’t. I don’t think healthcare problems are inherently different than other problems. There are other problems where you work with people and there are problems where you don’t work with people. There are problems with stochastic elements and ones that don’t. You are still using the same modeling paradigm for healthcare as you are for other thing, but there could be other considerations. For example, in radiation therapy, the main way we characterize the quality of dose inside a patient uses the same mathematical structure as portfolio optimization in financial engineering. They use something called “Value at Risk” and we use something called “Dose Volume Hisotograms,” but mathematically they are the same thing.

 

Originally published on 12/18/14.