Choice of data types in time resolved fluorescence enhanced diffuse optical tomography

Abstract

In this paper we examine possible data types for time resolved fluorescence enhanced diffuse optical tomography (FDOT). FDOT is a particular case of diffuse optical tomography, where our goal is to analyze fluorophores deeply embedded in a turbid medium. We focus on the relative robustness of the different sets of data types to noise. We use an analytical model to generate the expected temporal point spread function (TPSF) and generate the data types from this. Varying levels of noise are applied to the TPSF before generating the data types. We show that local data types are more robust to noise than global data types, and should provide enhanced information to the inverse problem. We go on to show that with a simple reconstruction algorithm, depth and lifetime (the parameters of interest) of the fluorophore are better reconstructed using the local data types. Further we show that the relationship between depth and lifetime is better preserved for the local data types, suggesting they are in some way not only more robust, but also self-regularizing. We conclude that while the local data types may be more expensive to generate in the general case, they do offer clear advantages over the standard global data types.