OSU Wexner Medical Center
Dr. White and Dr. Prevedello working together to analyze new images.
Machine learning, deep learning, augmented intelligence… No, I’m not pitching you concepts for the next installment of The Matrix. These three concepts are the backbone of the latest work at The Ohio State University Wexner Medical Center.
Artificial Intelligence, or AI, is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and other animals. The use of AI is growing throughout the healthcare field, and the Wexner Medical Center continues to stay at the forefront by integrating this tech for patient imaging.
“Radiology is already driven by the latest technologies, more so than other medical fields,” says Richard D. White, M.D. and chair of the department of radiology at OSU. “While AI is starting to improve radiology workflows at Ohio State, we’re now asking, ‘How can AI improve patient care?’”
Over the past two years, White’s department began applying concepts of AI to diagnostic imagining. Since then, the Laboratory for Augmented Intelligence in Imaging was developed and has set out on a number of projects with one goal in mind – to improve people’s lives.
“The overall goal is to improve patient care in many ways,” says Luciano Prevedello, M.D. and leader of the LAI2 team. “There are things in medical images that we cannot see with the naked human eye, but with the massive computing power we have access to, we can extract information that is available and correlates with pathology.”
Speeding up the process
An overarching issue at hospitals throughout the world is that scans are often marked urgent by doctors who believe their patient’s condition could be critical; thus creating a backlog of presumed urgent scans that require immediate attention.
For example, at larger academic hospitals like OSU, more than 40 percent of its CT scans that are ordered are marked as urgent.
OSU’s recent partnership with AI computing business, NVIDIA has led the university to the creation of an algorithm that can sort and prioritize scans much faster than a human can be based on imaging findings.
“The idea of the algorithm is to speed up the interpretation process,” Prevedello says. “Imagine if you could have a computer looking at these images right when the scanner acquires them. This allows us to detect diseases and problems so much earlier and provide care as soon as possible.”
The program itself can read a scan in roughly six seconds, and once those scans are digested, workflow is shifted so that the more critical findings are at the top of the list for interpretation.
“The program works in real time to alert physicians about serious health issues sooner so we can speed care for those patients,” Prevedello says. “While not perfect, it’s highly accurate, recognizing more than 90 percent of bleeding, water and tumors in the brain during our testing.”
Before delving into the world of healthcare, NVIDIA was chiefly known for producing graphic cards for video games.
While this technology is vastly improving workflow and diagnoses, Prevedello notes that this is not a permanent replacement for a skilled radiologist. In many ways, the algorithm is meant to shorten the time frame at the time of acquisition.
“There are several ways in which these tools can fail and the safest approach is the way we are using it – that being said, these tools are very good screeners,” Prevedello says. “Once it detects a positive case, it notifies the radiologist for the final and appropriate clinical interpretation.”
Through the continued development of the team’s algorithms and technology, efficiency and effectiveness are afforded to the radiologists. OSU’s team has created a deep learning algorithm that is able to understand clinical text entered by physicians.
To provide context, by using AI, it can analyze a situation like a 56-year-old female with double vision and then suggest the best imaging parameters to answer the doctor’s clinical question.
Once a biopsy comes back, the algorithms correlate image features with the result and can actually learn to predict the disease in future cases.
“We’re excited for what’s coming in the near future,” White says. “We’re a part of something right now that will make medical care faster, safer and more personalized for patients everywhere.”
Rocco Falleti is an assistant editor. Feedback welcome at rfalleti@cityscenecolumbus.com.