Singular-value analysis and optimization of experimental parameters in fluorescence molecular tomography

The advent of specific molecular markers and probes employing optical reporters has encouraged the application of in vivo diffuse tomographic imaging at greater spatial resolutions and hence data-set volumes. This study applied singular-value analysis (SVA) of the fluorescence tomographic problem to determine optimal source and detector distributions that result in data sets that are balanced between information content and size. Weight matrices describing the tomographic forward problem were constructed for a range of source and detector distributions and fields of view and were decomposed into their associated singular values. These singular-value spectra were then compared so that we could observe the effects of each parameter on imaging performance. The findings of the SVA were then confirmed by examining reconstructions of simulated and experimental data acquired with the same optode distributions as examined by SVA. It was seen that for a 20-mm target width, which is relevant to the small-animal imaging situation, the source and detector fields of view should be set at approximately 30 mm. Equal numbers of sources and detectors result in the best imaging performance in the parallel-plate geometry and should be employed when logistically feasible. These data provide guidelines for the design of small-animal diffuse optical tomographic imaging systems and demonstrate the utility of SVA as a simple and efficient means of optimizing experimental parameters in problems for which a forward model of the data collection process is available.