As the use of artificial intelligence across medicine increases nationwide, The Daily Pennsylvanian spoke to professors, doctors, and researchers at the Perelman School of Medicine about how they are integrating AI and machine learning into their research.
Across the University’s Health System, scientists are now using AI to enhance their understanding of biological systems and modern medicine. According to multiple Penn Med faculty members, the technology has been used in a variety of ways — from conducting translational research and analyzing data to optimizing healthcare delivery.
Pattern detection and risk prediction
Radiology and Electrical and Systems Engineering professor Christos Davatzikos described how he uses AI to identify early signs of disease and inform preventative treatment plans in an interview with the DP.
His lab — one of the first AI-guided radiation therapy projects in the field — used AI to develop methods to analyze brain MRI scans and predict the future progression of a tumor.
Davatzikos works with patients suffering from glioblastoma — a deadly form of brain cancer — and told the DP that AI pattern recognition helped lead to “much longer survival” of affected individuals.
He described AI as a “fundamental” research tool for detecting patterns, since the increasing number of biomarkers used to track diseases makes it “difficult” for one person to visualize how the brain is constantly changing.
“We look at the MRIs, and you don’t know where the tumor is infiltrating,” Davatzikos said. “But the AI is able to look at many different MRI contrasts and create a signature that predicts or detects very subtle changes that later give rise to tumor currents.”
RELATED:
Penn Medicine on ‘front lines’ of new drug crisis as withdrawal cases rise in Philadelphia
Penn Medicine researchers to present blood cancer, hematology research at national conference
Professor of psychiatry Birkan Tunç — who also serves as a research scientist at the Children’s Hospital of Philadelphia — told the DP that he uses AI to study “almost all psychiatric conditions,” including depression, anxiety, ADHD, and autism.
Tunç described how AI helps his team “capture cues and signals” from video or audio recordings and allows them to compare observations of individuals with autism and other conditions. He added that identifying these “signatures” can aid in early diagnosis, which could be “very beneficial for families.”
Perelman School of Medicine Neurology professor Brian Litt is working on a project “testing implantable brain devices that talk to their hosts,” which incorporates AI to report fluctuating levels of risk.
“Your medical device might text you and say, ‘What did you do? Your probability of having a seizure just went up by 60%,’” Litt told the DP. “And you might have had a beer, or taken a new antibiotic or something like that.”
Large-scale data analysis
Informatics in Biostatistics and Epidemiology professor Li Shen described how he integrated AI into one of his major research projects — a machine learning and informatics method that analyzes data and identifies specific Alzheimer’s disease biomarkers.
“We want to identify genetic risk factors and also protective factors, because that can be used for drug development, to try to understand the disease mechanism and accelerate therapeutic discovery,” Shen told the DP.
Associate Dean for Computing at the Medical School Li-san Wang — who serves as co-director at the Center for Artificial Intelligence and Data Science for Integrated Diagnostics — also uses AI in his Alzheimer’s disease research.
In a statement to the DP, Wang wrote that in his research, AI could help integrate genetic findings with other types of biological data — such as RNA, protein, and epigenomic data — which could prove “essential” for understanding the biological mechanisms behind Alzheimer’s disease.
Professor of bioengineering, biochemistry, and biophysics Greg Bowman spoke with the DP about using AI in multiple aspects of his work.
Bowman described how he uses AI to design peptides that bind to proteins not previously considered “viable targets,” and at the data analysis stage.
“It is really useful for helping us find patterns that are difficult to pull out by eye, given the massive amounts of data we generate over time,” Bowman said.
Streamlining research workflows
Davatzikos noted that while AI has been fundamental to his work, general improvements in computational and mathematical methods also allowed his research process to evolve.
“For example, analyzing brain scans was an art by visual inspection,” Davatzikos told the DP. “Now we have tools that precisely go in and measure certain things."
“My work has been primarily in making imaging more quantitative and more scientifically grounded,” he said.
According to Shen, using AI to help reduce “tedious time computing” has had a “big impact” on his research.
“It actually has all the knowledge rights available, say on the Internet, or whatever knowledge base it has access to ... so it can provide additional thoughts, new sales insights to the expert,” Shen said.
Litt spoke with the DP about Penn’s Center for Healthcare Transformation and Innovation, which is using AI to assist in operating room scheduling across Penn’s hospitals.
“You have to be really smart about how you schedule time in operating rooms, if you have downtime, it’s pretty expensive,” Litt said. “There’s a special Penn Health tech fellow, Kevin Shea, who has an algorithm that could potentially save half a million dollars a month in downtime from the operating room by using AI to schedule.”
Wang echoed Litt’s sentiment that AI can be beneficial for daily operations within his work — including data organization, annotation, and management — which he said remain “major challenges in large, collaborative research efforts.” He clarified, however, that AI remains “resource-intensive and expensive,” which places limitations on its potential uses.
Tunç also acknowledged that AI, in its current state, is a tool to “improve clinical workflows” and not a substitute for human labor.
“We never see it as a replacement, like trying to replace a human construct, because even the idea of psychology is a human construct,” Tunç said. “You need a human at the center.”
RELATED:
Penn Medicine on ‘front lines’ of new drug crisis as withdrawal cases rise in Philadelphia
Penn Medicine researchers to present blood cancer, hematology research at national conference
Staff reporter Addison Saji covers Penn Medicine and can be reached at saji@thedp.com At Penn, she studies English. Follow her on X at @addisonsaji.
Staff reporter Sameeksha Panda covers Penn Medicine and can be reached at panda@thedp.com. At Penn, she studies chemistry.
