Skip to main content

Faculty Spotlight: Katherine Heller, PhD

Monday, October 30, 2017
Katherine Heller

At first glance, machine learning and statistical science aren’t the most obvious skillsets for a member of a neurology department. Expertise in these areas, however may be key to helping patients and clinicians better understand patterns and trends in multiple sclerosis and other complex diseases. In this “Faculty Spotlight” interview, Katherine Heller, PhD, talks about her development of the MS Mosaic iPhone app, the similarities and differences between human and machine learning, and her travels from Machu Picchu to the Western desert of Egypt.

What does your average work day look like? How do your contributions to the Department of Statistical Science overlap with those of the Neurology Department?
I am very fortunate that I don't have an average work day. But my days usually involve some combination of teaching, going to talks, meeting with graduate students and others about research, writing papers and grants, service work here at Duke (e.g. PhD admissions committees or thesis defense committees), service work for my field internationally (e.g. overseeing the acceptance of papers at our major publishing venues or being on the board of an organization for women in machine learning), or travel to talk about research. I answer a lot of email.

A good deal of the research that I do revolves around the application of machine learning, Bayesian statistics, and artificial intelligence methods to medicine. My training was partly in Brain and Cognitive Sciences, so I have a particular interest in the parts of medicine that deal with the brain.

How did you first get involved with the Neurology Department?
I first got involved with the neurology department through a combination of my interest in what turned into MS Mosaic, and through my being part of the Center for Cognitive Neuroscience.

You co-lead the development of the recently released MS Mosaic iPhone app and ongoing research study. What was your role within this project?

My role within the project is as the machine learning expert. I am very interested in the benefits that data collection can bring to patients. Right now most of the disease-relevant information that people with chronic illnesses have is not collected as data in a clinical setting. Finding mechanisms to collect data on what is happening in patients' day to day lives outside of clinic is necessary to better understand their disease, and best improve their care. This is why we developed MS Mosaic.

What are you most proud of about this app? What’s next for MS Mosaic?

That it exists. It's been a very long road.

Now comes the fun part, where we start collecting data on people who are using the app and develop methods to learn more about multiple sclerosis and its treatment.

What is your role on the board of directors for Women in Machine Learning? What are the challenges and opportunities for women in this field where women have been traditionally underrepresented?
I've been involved in WiML since its founding. I'm on the board of directors, and held the grant that funds our annual workshop for several years. There are a significant number of challenges to women in the field, and in many fields, as we are now seeing in the media. Sexual harassment aside, being in an underrepresented field means that there are often characteristics that women have, that are stereotyped as being the mark of people who aren't good machine learning researchers, so there ends up being bias against women as a group, that is very hard to prove for any individual. It's not unusual that, e.g. a journal board or a department asked for names of editors or potential speakers, and receives a list of 60 men and 0 women.

How much overlap is there in the fields of human and machine learning? How does your understanding of each of these fields inform the other?

There's a great deal of overlap. Deep neural networks, now the most en vogue area of machine learning, were originally inspired by neurons in the brain. Artificial intelligence (the umbrella under which machine learning falls) cuts both ways. Fundamentally people often have a bias in their view of "intelligence" towards human intelligence, and the brain in particular, is used for inspiration in AI methods. At the same time, machine learning has a huge amount to offer in its ability to understand human and neural data, and to offer models of how the brain might work.

A decade ago you visited Egypt. What was the most memorable part of that trip? What’s the most interesting or exotic place you’ve been to since then?

The most memorable part of the trip was going through the Western Desert. It was incredibly beautiful. But also politically unstable. We had more police escorts than travelers. It was a family vacation in my family.

I've been to a number of really interesting places since then. I was really impressed at Rwanda's recovery from the genocide. I probably had the most fun hiking the Inca trail and through the sacred valley in Peru, or just hanging out with a bunch of howler monkeys on top of Mayan pyramids in Tikal (Guatemala).

K Heller

What passions or hobbies do you have outside of the Department?
I climb with friends at the Triangle Rock Club in Morrisville. I travel a lot. Most of the rest of the time I'm generally trying to make my kids' lives as awesome as possible.

K Heller