Machine learning (ML), a branch of artificial intelligence (AI) and computer science, is an emerging field which uses algorithms to analyze and draw inferences from patterns in data. ML and AI approaches are the basis of sensational innovations – like self-driving cars or ChatGPT – as well as something as familiar as Netflix’s recommendation engine.
There are growing exciting possibilities of ML applications in medicine including predicting a patient’s prognosis or monitoring symptoms to allow more proactive management and care. Another exciting potential use of ML is as a tool to support radiologists and pathologists in interpreting medical images. This application could potentially accelerate productivity and improve diagnostic accuracy.
While interest in ML for medical imaging is explosive, taking new methods and technologies from “code to clinic” remains a major challenge.
“Since the deep-learning revolution a few years ago, there has been an exponential increase in the number of papers that come out each year on machine learning and medical imaging,” explains Pallavi Tiwari, PhD, who leads UW-Madison’s Machine Learning for Medical Imaging (ML4MI) Initiative. “The goal is clinical implementation: How do we make this happen? And that’s a big bottleneck in the field right now.”
She believes interdisciplinary collaboration will be the key to achieving this goal. The ML4MI Initiative – sponsored by the Departments of Radiology, Medical Physics, and Engineering – fosters exactly that.
“We are hoping to bring together the researchers who are developing these sophisticated algorithms and methods and the clinicians/radiologists who make clinical decisions,” she explains.
With a world-class engineering program and an outstanding team of machine learning researchers in the School of Medicine and Public Health, UW-Madison is uniquely suited to this collaborative endeavor. By bridging the gap between these groups, the ML4MI Initiative aims to catalyze the translation of these emerging technologies into clinical practice.
The team includes:
• Alan McMillan, PhD, the Lab Director of Molecular Imaging/Magnetic Resonance Technology Lab (MIMRTL)
• John Garrett, PhD, the Director of Informatics and Co-Lead of the Opportunistic Screening Consortium in Abdominal Radiology (OSCAR) Project
To kick-start the initiative, an ML4MI retreat was held on March 15 at Memorial Union Terrace. Over 100 members of the UW-Madison community attended the event, where leading researchers at UW presented their work.
While these talks highlighted the immense potential of machine learning to solve challenging problems in image reconstruction, image processing, and computer-aided diagnosis, they also emphasized the need for further breakthroughs before these methods can be integrated into clinical practice.
“This retreat is just the first step,” said Dr. Tiwari. “At that event, we got to gather the brilliant minds here at UW-Madison, but in the future, we hope to have events that are open to an even wider audience.”
Already, the ML4MI Initiative offers an open-access seminar series featuring external speakers at the forefront of the field. In the future, Dr. Tiwari and her team hope to organize additional events, to foster growing collaborations between College of Engineering, SMPH, and the newly established Data Science Institute at the intersection of AI and precision medicine
To read more about ML4MI and stay up-to-date on upcoming events, resources, and opportunities, check out their website here.