Near-infrared (NIR) tomography breast image reconstruction with a priori structural information from MRI: algorithm development for reconstructing heterogeneities

A combined magnetic resonance and near-infrared (MRI-NIR) imaging modality can potentially yield high resolution maps of optical properties from noninvasive simultaneous measurement. The main disadvantage of near-infrared (NIR) tomography lies in the low spatial resolution resulting from the highly scattering nature of tissue for these wavelengths. MRI has achieved high resolution, but suffers from low specificity. In this study, NIR image reconstruction algorithms that incorporate a priori structural information provided by MRI are investigated in an attempt to optimize recovery of a simulated optical property distribution. The effect of high levels of tissue heterogeneity are evaluated to determine the limitations of incorporating prior information into a realistic set of patient breast images. We assume absorption coefficient (/spl mu//sub a/) variations near /spl plusmn/40%, and transport scattering coefficient (/spl mu//sub s//sup //) variations near /spl plusmn/20%, in a coronal breast MRI geometry. Changes in tissue pathology due to tumor growth can be observed with NIR tompgraphy, and so the goal here is to determine how best to quantify these tumor-based contrast regions within the presence of high tissue heterogeneity. By applying knowledge of tissue's layered structure in reconstruction through various constraints in the iterative algorithm, quantitative recovery of the tumor optical properties improves from 69% to 74%, and localization improves as well. However, only when the true heterogeneity of the tissue distribution was included was accurate quantification of the tumor region possible. Using a good initial guess of /spl mu//sub a/ and /spl mu//sub s//sup //, derived from the regional structure of the model, quantification of the region reaches 99% of the true value, and spatial resolution retains a similar value to the original MRI image.