gms | German Medical Science

Previously we reported inverse method for identification of local cardiac ischemia from difference QRST integral maps. Equivalent current dipole as a model of the cardiac generator was used to represent repolarization changes in the ischemic region. Aim of this study was to check applicability of the method and to estimate reliability of the located lesion in particular cases. Local ischemic lesions modeled in typical positions were inversely located from surface potentials contaminated by various degree of noise. Several criteria to assess feasibility of input data were studied: rms value to represent the strength of the source, local extrema and gradients in maps to characterize topology and noise level. Another criterion to evaluate credibility of the inverse solution was the difference between the input map and map computed from estimated dipolar source. Additional criterion to estimate uncertainty of the inverse solution was based on the shape of the criterial function used to select among possible solutions. Simulations showed that rms values of maps generated by small lesions are greater than their experimentally observed intraindividual variability in healthy persons (5-20%) what enables their identification. Relative rms difference between input and dipolar map in successful localizations varied from 3-11% when no noise was present but increased up to 50% when 10% noise was added. Hence correct interpretation of the relative rms difference depends on the estimated noise level.

Our results suggest that complex evaluation of input signal magnitude, presence and degree of disturbances in the input map, degree to which the dipolar map approximates the input map and carefully adjusted uncertainty measure can help to identify cases in which the dipole is not an adequate representation of the ischemic lesion or the result is unreliable because of low stability of the inverse solution.