Signal source separation and localization in the analysis of dynamic near-infrared optical tomographic time series

The emerging sub-field of dynamic medical optical tomography shows great potential for conferring significantly enhanced early diagnosis and treatment monitoring capabilities upon researchers and clinicians. In previous reports we have showed that adoption of elementary time-series analysis techniques can bring about large improvements in localization and contrast in optical tomographic images. Here we build upon the earlier work, and show that well-known techniques for extraction and localization of signals embedded in a noisy background, and for decomvolution of signal mixtures, also can be successfully applied to the problem of interpreting dynamic optical tomography data sets. A general linear model computation is used for the signal extraction/localization problem, while the decomvolution problem is addressed by means of a blind source separation technique extensively reported.