Spectral estimation of human blood for use in cancer detection

The methyl and methylene resonances in the NMR frequency spectrum of blood plasma are known to be affected by the presence of cancer. Here the parameters of the individual resonances comprising the two resonances are estimated using a nonlinear least-squares technique based on the Levenberg-Marquardt method. The most significant parameters for detecting cancer are identified. Functions for classifying given samples into cancer and non-cancer categories are presented. The discriminant functions correctly classified at least 78% of the samples.<<ETX>>