Individually improved VCG synthesis

Some prognostic electrocardiographic variables are to be measured in a vectorcardiogram (VCG). Normally, VCGs are synthesized from a standard 12-lead ECG by a transformation matrix. This occurs at the price of some information loss that can be defined as the error ¿ between the original ECG and reconstructed ECG (ECGr). Our assumption is that the larger ¿, the less reliable the synthesized VCG. We attempted here to improve the VCG synthesis by reducing ¿ for each individual by using the Errors-In-Variables model. This technique minimizes ¿ by allowing 1-10% changes of the transformation matrix (M) that can be seen as the individualization of the conductive torso properties. We tested this procedure in 180 subjects, and used the squared correlation as a quality index for the resemblance of ECG and ECGr (R² _ECG). On average, the R² _ECG is markedly improved from 0.94 (inverse Dower) to 0.97 to 0.99 when we allow 1-10% adaptations of M. We conclude that improvement in the R² _ECG, and thus individualizing M through our method, should give us a better and individualized VCG.