Cardiovascular disease is the number one cause of death worldwide. For example, heart failure is a significant source of mortality within the U.S; it is responsible for 1% of all emergency room presentations and contributes to one in every nine deaths. A significant proportion of heart failure patients have concurrent conduction system disease, which will lead to their eventual death. Cardiac resynchronization has proven to be a useful therapy for improving cardiac function as well as reducing mortality in some patients. However, a significant number of patients fail to respond to resynchronization therapy, often due to inadequate pacemaker lead placement. There currently exists no pacemaker system that provides the 'potential' benefits of multisite temporally and spatially precise pacing for resynchronization. To reach the audacious goal of eliminating cardiovascular diseases, new technologies must be developed that will monitor the diseased heart with unprecedented temporal and spatial precision and will manage the pacing of the heart to restore a healthy function of the heart. This project will process data recorded from multiple sites and will generate a pacing therapy specific to the patient in real-time.
The award revisits the core foundation of pacemakers to develop temporally and spatially precise pacing at multiple sites for resynchronization. The researchers will develop a robust, well-annotated database of intra-cardiac electrograms (IEGM) from multiple cardiac sites. The focus will be on identifying challenging cases that are clinically difficult to differentiate and, thus, stand to reap the greatest benefit of being able to direct overall algorithm development. This information will be added to the associated metadata file. Data will be collected from a minimum of 150 patients with at least 50 patients from each identified pathophysiology. Pathology of each patient based on data from multiple intra-cardiac recording sites will be identified. The proposed machine-learning pipeline explores the representation of time series using wavelets and then learns transformations of multiple time series using the Lie group framework. This pipeline clusters time-series data to identify the right pathology for the specific patient. In addition, the project explores implementation as an application specific integrated circuit (ASIC) that will be implanted subcutaneously to continuously process intracardiac multi-site recordings as well as to generate temporally and spatially precise pacing patterns. The overall approach is to holistically develop a methodology that can address an extremely low-power implementation of machine learning and signal processing algorithms, by not only combining, but jointly optimizing, algorithmic-, circuit- and architecture-level innovations.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
