Reconstructing Electrocardiogram Leads From a Reduced Lead Set Through the Use of Patient-Specific Transforms and Independent Component Analysis

In this exploration into electrocardiogram (ECG) lead reconstruction, two algorithms were developed and tested on a public database and in real-time on patients. These algorithms were based on independent component analysis (ICA). ICA was a promising method due to its implications for spatial independence of lead placement and its adaptive nature to changing orientation of the heart in relation to the electrodes. The first algorithm was used to reconstruct missing precordial leads, which has two key applications. The first is correcting precordial lead measurements in a standard 12-lead configuration. If an irregular signal or high level of noise is detected on a precordial lead, the obfuscated signal can be calculated from other nearby leads. The second is the reduction in the number of precordial leads required for accurate measurement, which opens up the surface of the chest above the heart for diagnostic procedures. Using only two precordial leads, the other four were reconstructed with a high degree of accuracy. This research was presented at the 33rd International Conference of the IEEE Engineering in Medicine and Biology Society in 2011.1 The second algorithm was developed to construct a full 12-lead clinical ECG from either three differential measurements or three standard leads. By utilizing differential measurements, the ECG could be reconstructed using wireless systems, which lack the common ground necessary for the standard measurement method. Using three leads distributed across the expanse of the space of the heart, all twelve leads were successfully reconstructed and compared against state of the art algorithms. This work has been accepted for publication in the IEEE Journal of Biomedical and Health Informatics.2 These algorithms show a proof of concept, one which can be further honed to deal with the issues of sorting independent components and improving the training sequences. This research also revealed the possibility of extracting and monitoring additional physiological information, such as a patient\u27s breathing rate from currently utilized ECG systems