Cleveland Clinic creates Center for Clinical Artificial Intelligence
Health system envisions center as a hub of collaboration between physicians, researchers, computer scientists and statisticians to find new ways to use AI technologies for improving patient outcomes.
The new Cleveland Clinic Center for Clinical Artificial Intelligence, launched by Cleveland Clinic Enterprise Analytics, will leverage machine learning, deep learning and natural language processing to power diagnostics, disease prediction, as well as treatment planning.
“We realize that there’s a lot of opportunity in the space of artificial intelligence to advance patient care and outcomes,” says Aziz Nazha, MD, who has been appointed director of the center and associate medical director for AI. “The mission is to harness the power of artificial intelligence to improve healthcare delivery and medical research by focusing on the clinical needs of the patients.”
By bringing together specialists from various departments, including genetics, IT, laboratory, pathology and radiology, the center will develop innovative clinical AI applications such as machine learning algorithms designed to reduce the risk of hospital readmission and to predict patient response to cancer treatments, according to Nazha.
“Our focus is always on the outcomes, not the methodology,” he adds. “We do focus on the technology, of course—we want to have the right model used for the right problem. But our primary focus is always on the outcomes.”
To build its machine learning models, Nazha says that the AI center will tap into the Cleveland Clinic’s large cohort of 1.5 million patient admissions. Among the projects that are currently underway are identifying patients with high risk of death during admission, predicting inpatient length of stay and readmission risk with a higher degree of accuracy than provided by existing models.
In addition, several ongoing oncology projects include developing models to provide personalized prediction of outcomes in cancer patients, improving cancer detection in pathology slides via computer vision, and predicting response or resistance to chemotherapy with multiple machine learning algorithms.
“Machine learning is not a black box—the models must make sense to us clinically,” concludes Nazha, who emphasizes the importance of clinicians understanding the internal workings of such algorithms. “We open the black box and show the physicians how and why it works. When physicians see this, then they start to believe in it.”