Calibration techniques and datatype extraction for time-resolved optical tomography

This article describes the preprocessing and calibration methods currently applied to data acquired with the University College London multichannel time-resolved optical tomography system. We briefly outline the imaging system and describe the features of our experimentally collected data, sources of stochastic noise, and systematic errors. We examine two methods of calibrating data: "difference imaging" using two image data sets with and without the features of interest to produce an image, and "absolute imaging" using an independent calibration measurement. We describe the methods developed to apply each calibration to raw data. Although the difference imaging performed is found to produce images with fewer artifacts, analysis indicates that it will not be directly applicable for clinical applications. Also examined are the effects of using a two dimensional (2D) reconstruction scheme to produce images from measured data. For absolute imaging, artifacts are shown to dominate such images even in the case of a homogeneous third dimension. The feasibility of deriving an ad-hoc correction factor to allow the use of a 2D reconstruction for measured data is examined, and is shown to reduce artifact. Difference imaging is demonstrated to be more robust to such effects.