Optimisation of the locations of multiple-dipole heart generators in a simple torso model

Abstract

Electrocardiographic inverse solutions frequently assume multiple-dipole equivalent generators with locations fixed throught the QRS complex. This paper describes investigations which extend the solution to optimise the dipole locations at any instant of time. The combined results lead to a moving multiple-dipole inverse solution which is more accurate than the conventional solutions at each instant and also contains information related to the dynamic movement of the physiological generators. Tests so far have been limited to a 3-dipole solution in a highly simplified torso model. The generators in this study were dipoles displaced varying amounts from a standard central location; the inverse solution starts by using the standard location and then is led to an improved solution by a guided search. Moderate displacements, as compared to the size of the heart model, were shown to yield root mean square errors in the inverse-solution source strengths as high as 40% when the latter were found via a least-mean-square procedure. The method of steepest descent was utilised in the second step to improve the locations and strengths of the initially assumed equivalent-dipole generators and thus improve the accuracy of the inverse solution. The percentage root-mean-square source-strength errors were reduced from as high as 40% to under 10% in most cases. The initial position error of up to 10% for the source dipoles typically was reduced to less than 5% throught the use of the search routine. The ability of the solution dipoles to track the original locations was demonstrated.