Utilizing Machine Learning to Improve Clinical Care and Scientific Research

Lee Cooper, PhD, is associate professor of Pathology and Director of the Division of Computational Pathology. He is also director of the Center for Computational Imaging and Signal Analytics in Medicine in the Institute for Artificial Intelligence and Medicine, a member of the Robert H. Lurie Comprehensive Cancer Center of Northwestern University and a professor at the McCormick School of Engineering. His research is dedicated to improving the accuracy, accessibility and efficiency of machine learning solutions to problems like diagnosis, prognostication and treatment response prediction. 

What are your research interests? 

I investigate how machine learning can be used to improve diagnostics and to acquire knowledge from data generated by the pathology lab. I primarily work with whole-slide images, which are high-resolution images of glass slides that each contain billions of pixels. I also have a focus on software and computing since the data in this area can be very large. A pathology lab can produce a million or more slides annually, which translates to multiple petabytes of data when digitized. This data contains tremendous untapped clinical and scientific value. More recently I’m exploring how we can use cloud computing out of necessity. 

What is the ultimate goal of your research? 

We want to improve the accuracy, “explainability” and efficiency of machine learning solutions to problems like diagnosis, prognostication and treatment response prediction. We also produce software tools to work with enterprise-scale data and to make machine learning more accessible to biomedical investigators.  

How did you become interested in this area of research?  

I had an encounter as a student that sparked an interest in biology. I was an engineer studying math, physics and high-performance computing at the time. I became intrigued with how complex biology and disease are and I found my way to pathology from there. The digital pathology community was tiny when I graduated but has grown into a billion-dollar industry with a large and vibrant academic community. 

What types of collaborations are you engaged in across campus (and beyond)? 

Here, I collaborate with Jeff Goldstein who is our medical director of Digital Pathology. We work together on implementing digital pathology at NM and share a research program for translating AI. My field is more active in Europe and so most of my academic collaborators are located there. I also work closely with industry partners like NVIDIA and others to create software tools. Some of our open-source software has thousands of users and is actively developed by community members that we don’t interact with directly. It is an interesting form of collaboration that is unique to software. 

How is your research funded? 

We do not focus on a specific disease and so our NIH funding has come from diverse sources - the National Cancer Institute, National Institute of Biomedical Imaging and Bioengineering, National Institute of Neurological Disorders and Stroke, National Institute of Diabetes and Digestive and Kidney Diseases and National Library of Medicine. Many of our industry collaborations are also funded through the Small Business Innovation Research program.  

Where have you recently published papers?   

We recently published an article in Nature Medicine describing an AI system to predict clinical outcomes for patients diagnosed with invasive breast cancer. We mostly publish in specialty journals like Modern Pathology, where we have three articles this year.