Worldwide, more than 230 million major surgical procedures are performed each year. And while anesthesia is generally safe, its use does not come without risks, such as a spike in blood pressure, an allergic reaction or seizures. The National Science Foundation recently awarded $1.18 million to the University of Pittsburgh Swanson School of Engineering to support research into using machine learning and big data to analyze electronic anesthesia records and prevent these postoperative complications.
Heng Huang, the John A. Jurenko Professor in Computer Engineering at Pitt, is principal investigator on the study that will analyze more than 2 million cases of anesthesia data taken from 303 UPMC clinics and treatment centers.
“A human doctor uses guidelines from manuals in combination with subjective experience to determine patients’ risk factors and needs,” said Huang. “We are using artificial intelligence and machine learning to develop an objective way to predict surgical outcomes based on historical patient data.”
Huang is collaborating with Dan Li, an assistant professor in the School of Nursing, and Fei Zhang, a certified registered nurse anesthetist in the Department of Anesthesiology and Perioperative Medicine.
The team will design new deep learning algorithms and software to mine patient data and identify common risk factors in patients about to receive anesthesia. They will then develop a “decision support system” to better inform doctors when patients are at high risk for postoperative complications and in-hospital mortality.
“Many patients come in to the hospital with so much information about them on file that doctors don’t have a comprehensive way to consider all the variables and their interactions,” Huang said. “With a computer, you really can do a better job than a human of determining how all that data is going to predict patient outcomes.”
To create a large-scale, machine learning framework capable of predicting patient outcomes, Huang will employ several emerging computational technologies including deep learning, semi-supervised learning and large-scale optimization.
Huang has been creating new machine learning techniques for biomedical applications throughout his career. Some of his past projects involved analyzing big imaging genomic data to help identify Alzheimer’s disease at earlier stages, data mining electronic medical records to personalize patient treatment, integrating pathology images and cancer genomics for personalized medicine and building interactive gene expression databases.
“I’ve focused on applying computational techniques to biomedical applications for about the past 15 years because you can really make a big impact on improving people’s quality of life and benefiting humanity with A.I. in ways humans cannot achieve alone,” Huang said.