The potential to perform both feature localisation and tissue characterisation within a clinical MRI scanner is an attractive one. Many of the latest scanners have the capability to identify a region on the MRI image an then obtain spectroscopic data from it. That spectroscopic data comes from the magnetic spin resonances of different atomic species within the region of interest. From information of the spatial distribution of atomic species it may be possible to determine the cellular characteristics of tissue within the region. Such tissue characterisation, if it were successful, would have a large number of applications - particularly in the detection and localisation of tumours.

Previous work, predominantly on in-vitro tissue, has studied the amount of 1H and 31P in tissue using magnetic resonance spectroscopy. These are abundant in both the underlying tissue and in metabolites which also allows the activity of tissue to be studied. There are problems in characterising the spectra obtained and neural networks and other artificial intelligence techniques have commonly been employed. However, the classification success is dependent on the quality of the training set data which potentially limits the application of the technique. In addition, the classification provides no information of the structure of the tissue and it is this information which informs diagnosis.

We have developed a theoretical three compartment model of tissue based on the intracellular fluid, extracellular fluid and the cell membranes. These models relate the concentration of chemical species in different compartments to tissue structure. We are currently exploring the use of these models to characterise the cellular structure of tissue using measurement of tissue cultures using NMR spectroscopy facilities in the Department of Physics.