Data subset algorithm for computationally efficient reconstruction of 3-D spectral imaging in diffuse optical tomography

Three-dimensional (3-D) models of light propagation in diffuse optical tomography provide an accurate representation of scattering in tissue. Here the use of spectral priors, shown to improve quantification of functional parameters in 2-D, has been extended to 3-D. To make 3-D spectral imaging computationally tractable, a novel technique is presented to deal with the large data set. The basic principle consists of using a dynamic criterion to select optimal data subsets that capture the major changes in the imaging domain. Results from three test cases showed comparable image quality and accuracy with less than 4% difference between the uses of data subset approach versus the entire dataset. Tested on simulated data from two different models, the algorithm was able to discern multiple objects successfully with an average error of 30% in quantifying multiple regions and less than 1% in quantifying the background.