Computer scientist developing predictive models for diagnosing heart disease
The University of Delaware’s Hagit Shatkay is spending her sabbatical at an unusual place for a computer scientist — a hospital.
An expert in computational biology and medical informatics, Shatkay is developing predictive models for diagnosing and assessing the severity of heart disease in collaboration with cardiologists at Johns Hopkins Medicine.
The work is funded by a grant from the National Science Foundation through the EAGER (Early-concept Grants for Exploratory Research) program, which supports potentially transformative research that explores new subjects, different methods, or interdisciplinary approaches.
Shatkay’s work is aimed specifically at hypertrophic cardiomyopathy (HCM), a condition in which the septum — the wall that divides the two lower chambers of the heart — thickens, making it difficult to pump blood. HCM can affect people of all ages and is a common cause of sudden cardiac death in young people, often reported in young athletes.
HCM is difficult to screen for, predict and diagnose, especially in young and active people. Detecting subtle early signs and accurately measuring septum thickness is important for HCM diagnosis and for identifying people who are at high risk.
“While MRI provides clear images that may help gauge septum thickness, it is expensive and time-consuming, and its interpretation requires a high level of expertise,” Shatkay says. “As such, it cannot be used on a large-scale as a screening tool.”
“On the other hand, simpler monitoring techniques such as echocardiograms and electrocardiograms (ECG) can provide information quickly and inexpensively but are difficult to interpret. Identifying potential HCM from ECG and gauging septum thickness from ‘noisy’ and fuzzy echocardiograms is particularly difficult and error-prone.”
Shatkay plans to apply machine learning methods and tools to noisy electrocardiogram data and echo-images that have been annotated by experts at Hopkins. This will enable a computer to identify which signals in the electrocardiograms may indicate trouble and where to measure the heart on an echocardiogram for the greatest accuracy.
“Developing computational tools that improve screening, diagnosis and prediction while making the basis for such tools simpler and less expensive stands to benefit large populations, especially in underprivileged areas where state-of-the-art imaging technology is scarce,” Shatkay says.
While the current project focuses on computer-assisted HCM detection, the research paves the way toward using similar methods in other disease domains, as well as in other data-intensive applications outside medicine.
About the research
Hagit Shatkay is an associate professor in UD’s Department of Computer and Information Sciences and an adjunct associate professor in the School of Computingat Queen’s University in Canada. She is also a member of the Center for Bioinformatics and Computational Biology at DBI, and affiliated with the Biomedical Engineering Department.
She is collaborating with Drs. Roselle and Theodore Abraham, cardiologists at the Johns Hopkins Medicine Heart and Vascular Institute, on this project.
The NSF EAGER grant, “Toward Predictive Models for Diagnosis and Severity Assessment of Heart Disease from Diverse Temporal Data,” provides Shatkay with $100,000 in funding for one year.