REDUCED LIPID CONTAMINATION IN IN-VIVO H-1 MRSI USING TIME-DOMAIN FITTING AND NEURAL-NETWORK CLASSIFICATION

It is a well-known problem that metabolite maps, reconstructed from in vivo H-1 MRSI data sets, may suffer from contamination caused by the presence of strong lipid signals. In the present investigation, the li pid problem was addressed by applying specific signal processing and d ata-analysis techniques, combined with pattern recognition based on th e concept of the artificial neural network. In order to arrive at imag es, cleaned from lipid artifacts, we have applied our previously intro duced iterative and noniterative time-domain fitting procedures. Furth ermore, reduction in computational time of the image reconstructions c ould be realized by using information provided by a neural network cla ssification of the spectra, calculated from the MRSI data sets.