Optimal Sensor Placement for Space–Time Potential Mapping and Data Fusion

Current ECG imaging (ECGi) systems deploy a large number of ECG sensors to provide the high-resolution body surface potential mapping (BSPM). The availability of BSPM was shown to substantially improve the early detection of life-threatening heart disease. However, most existing ECGi systems employ an approximately uniform distribution of hundreds of ECG sensors on the body surface. Very little has been done to investigate the optimal sensor placement for BSPM. In this article, we propose a new optimal sensing strategy to search the optimal number and locations of the sensors. First, we develop a greedy algorithm to sequentially place ECG sensor on the body surface, which will maximize the information gain at each step. Second, we leverage the available BSPM data to develop a spatiotemporal model of cardiac electrical activity. Third, we study the algorithmic convergence and stopping criteria by evaluating diminishing return of the placement of two sequential ECG sensors. Experimental results show that the optimal sensing strategy with 30 sensors yields large
$R^2$
statistics (
$>$
97%) for BSPM during the P, QRS, and T waves, as well as an average
$R^2$
statistics of 97.71% for 12-lead ECG, and 99.44% for 3-lead VCG. The proposed methodology has strong potentials to help further improve the design of ECGi systems.