A Conversation with Dr. Eva Lee 
by Aileen Lee
June 2007

Dr. Eva Lee is an associate professor at Georgia Tech's H. Milton Stewart School of Industrial and Systems Engineering. Some of her research topics include developing advanced optimization and predictive modeling to better treat cancer patients, early disease diagnosis, and preparing the healthcare department in the state and the local government for emergency responses in the events of bio-terrorists and infectious disease outbreaks and so on. In this conversation, Dr. Lee discussed not only her research, but also what brought her to where she is today.

Q - When you were in middle school or high school, how did you look at your future?

A - I really had no idea what I would go into besides mathematics or science. I learned a lot of mathematics before I even went to kindergarten. I was doing multiplication of hundreds when I was two years old.

I'm very numbers oriented. On the other hand, I also loved to dance; I was a ballerina. I was into painting and drawing and calligraphy (I won awards for those things). In college, [I found out that] a lot of kids interested in mathematics are also interested in the arts. It's an interesting combination, and I think they kind of complement each other.

Q - You said you started learning multiplication at the age of 2. Did your family push you into that?

A - I think I was just very obsessed with mathematics. I don't think that my parents were trying to make me love mathematics. I just chose whatever I wanted to do and they just let me explore that.

Q - You got a bachelor's degree in mathematics and computer science. Why the combination?

A - If you want to do mathematics, you want to understand the theory. The next thing you want is to be able to solve a real problem using the theory. A lot of the times, the mathematics is difficult, and the next real challenge is how you convert [mathematics] into practice. When I want to solve a real problem, in many cases I need the computational power, and that's what attracts me [to computer science]. It's almost natural.

Q - What kind of research do you do now?

A - I'm working a lot in medicine and healthcare related research: how you'd design and apply advanced mathematic models and computational algorithms that could describe clinical phenomena. For example, how you'd describe tumor cell growth and how you'd describe the treatment procedure.

If you looked at breast cancer, or if you looked at other types of diseases, when you diagnose early you can treat [the disease] very efficiently. For the patients, it's critical [to be diagnosed early], so that's why we look at the diagnosis of, say, cardiovascular diseases, or cancer. We look at [them] from the diagnosis to the treatment. If you can predict when a patient will develop the disease, how the patient would respond to a treatment, then when a patient comes in, you can predict what treatment to give him/her. That's very difficult; we're not one size fit all. We're not going to say that every patient will have his/her own regimen of treatment, because that may be too costly. On the other hand, you want to design treatment that is specific for different stages and the tumor types of the patient, and taking many of these parameters into account, we try to build a model.

Q - Sounds interesting!

A - This is really an art. The problem is complex, yet we have the ability to model it. The next step is how do you solve it, and give [the doctors] a solution that they [can] test in the clinic.

You can't just say "OK, we have a really nice system; put it in the hospital." Even if [the hospitals] see it really well in terms of the result, you still have to go through clinical trials and FDA approval. So it really takes a lot of time testing and researching. But it is good that we have that system, because a lot of the time you really don't see the effect until you look at the long term outcome.

Q - When did your research on prostate cancer begin?

A - I started researching prostate cancer in 1996, and we got really good results just a couple of years later, we are very thankful that we got recognized by the Franz Edelman competition now. It was only a few years ago that Georgia Tech licensed it to a company and put it to national usage.

Q - What exactly does the model do?

A - The model itself is for prostate cancer patients. It's for early-staged cancer where the tumor cells are still confined inside the prostate. One of the treatments for that is to put radioactive seeds inside.

Q - What are the seeds?

A - They are tiny grains of seeds. It's like little pieces of rice, but much smaller.

Q - What is it made of? Is it solid?

A - It is solid; it's a radioactive substance. You can think of it as a little piece of copper you put inside of the body (the organ where the tumor is). The radiation here we use is not copper; we use iodine, or palladium. In this case, you put the little seeds inside the organ with tumor, so the radiation is confined inside. The hardest part is to determine how you'd put the seeds inside – the seed configuration – so you get the best dose for the tumor; you want to kill all the tumor cells and you want to make sure the normal tissues aren't affected. That's the problem.

We design it as a mathematical model. The goal is to determine the best seed configuration, so all the tumor cell will receive a sufficient dose, and all the normal cells would still function after treatment. Also, we have to make sure all the clinical criteria are met. It's a simple enough problem to understand, because all you have is a little piece of tumor and you put little seeds inside. After we designed the model we discovered that it was really hard to solve. That's where computational algorithms come into place. We built the software to solve the model, and we give the solution to the clinicians.

Q - So how does that benefit the patients?

A - Patients come in and we use the imaging right away on the day of operation. [The doctors] do the planning using my software and they implant the seeds on the spot. It's a very efficient system.

Q - With the seed inside, do you do any outside drug delivery?

A - Some patients do. Sometimes what they would do is, before the treatment, they may have hormones to shrink the size of the tumor. Sometimes they also do chemotherapy. It really depends on the staging of the tumor. The path we're focusing on is how you do the seed implants in the most efficient way. In the case of breast cancer, we do put implants in the body, but it's not permanent. For prostate cancer, [the patients] would have these little seeds inside. For breast cancer, there are these little tubes that we put inside for a while, and then we take them out, so [the patients] won't have these seeds in the breast.

Q - How would you detect the tumor in the first place?

A - They take the PSA (Prostate Specific Antigen) of the patients. Once they detect it is at an escalated level, they will send the patients in for biopsy. In biopsy, they put needles in the [designated] area and they take little samples of the prostate tissues to see if there are any tumor cells. Diagnosis is not easy, because biopsy may not poke in the right spot. There is still a lot of work being done now in terms of how you diagnose different types of cancers. For breast cancer, we look at it before the cancer cells occur; we look at genomics, information like sequencing; we can look at changes in the pathways. That's early diagnosis, before the cancer cells have formed. That's really in the future, because you want to detect things [early] and treat it.

The biologists come up with biomarkers; they test them on patients and come up with numbers. Then they would tell us "OK, these are the patients that have the disease, and these are the group that doesn't." They have different measurements and stuff, and they make measurements in the biomarkers in terms of the characteristics. So, I take that information and I generate a predictive model and predictive rule to separate the groups (diseased/normal). If you have a new patient, and you take measurements, running the measurements through the predictive rule, we can tell you if the patient has the disease or not.

Q - How did you decide to use your background in mathematics and computer science to tackle the field of biological engineering?

A - I've always been interested in biology and medicine. I didn't think, at that time, that I wanted to be a clinician. I do have a premed background, but I didn't go through with the MD-PhD program, because I wasn't sure if that's what I really wanted to do. But the moment I graduated from my PhD program, I got into the medical research. That's when I started on the artificial intelligence prediction, where you can use for predicting commercial and industry applications, such as marketing trends, customer liking, and so on. But if you looked at a different side of it, you can also use it for medical diagnosis, and that part really fascinated me. That's when I decided to work on it. It was back in 1995, when I first started at Columbia [University].

Q - What advice would you give to high school students today?

A - High school kids don't have to know what they want to discover; they just have to be interested in trying to discover something. I would say, for high school kids, it is important to explore what your interests are and to excel in those areas.

I'm never satisfied with what is taught by the teacher. That is part of science. As you do experiments, you don't [always] see what you expect to see, because it is not deterministic--that is the whole reason why you're doing the experiment. Something is going to happen that you cannot predict.

I think the problem with high school is the students say, "OK, this is the subject, and I'm going to do all the homework, then I'm going to do all the tests, and that is it." That prevents them from thinking big. Students are too fixated with grades. Being creative, being able to think beyond what is being taught, and create a problem that nobody but you can solve are very important. High school kids should know that what they are taught only gives them a background.