The work presents a method for localising regions of myocardial infarction (MI) using magnetocardiography, which is validated by simulation techniques. We have verified that the first two principal components of magnetocardiographic signals obtained from torso mapping within a normal group are very similar even in experimental measurements. So, if we have the temporal eigenvectors of a normal group, these can be used as an orthonormal basis for estimating the signals of a patient being tested. Analysis of the residual signals using singular value decomposition (SVD) allows the localisation of regions of MI. Comparisons are made with a similar method which uses spatial eigenvectors as its basis. It is shown that our approach is able to localise the equivalent current dipoles which generate MI even when they are not orthogonal to the spatial eigenvectors.