Dibo Ntuba

Reconstructing Hemodynamic Parameters in Diffuse Optical Tomographic Breast Imaging

Thursday June 29 �

Abstract:

Diffuse Optical Tomography (DOT) is an emerging imaging technique with great potential for detecting, diagnosing and monitoring breast tumors. DOT provides information on absorption and scattering behavior of near-infrared light (650-900 nm) in tissue, from which functional parameters like hemoglobin concentration and tissue oxygenation saturation levels can be deduced These parameters are thought to be indicators of tumor angiogenesis and metabolism in breast tissue. However, DOT suffers from low spatial resolution due to the highly scattering, and thus diffusive, nature of tissue, and the resulting inverse problem is ill-posed and, in practical configurations, underdetermined.

This work concerns modeling and then estimating oxygen consumption during mammographic-like compression of the breast. We first extend an on-going study of the effects of compression on hemodynamic parameters in normal subjects under compression. Based on parameters drawn from this study, we then develop and test a simulation model that performs a regularized least-squares reconstruction of oxygen consumption and blood flow using both an "indirect" and a "direct" approach. The indirect approach first reconstructs oxy- and deoxy-hemoglobin from photon fluence measurements and then post-fits the results to an exponential equation model for oxygen saturation that imposes constant oxygen consumption and blood flow constraints. The direct approach incorporates this exponential equation into the forward model and solves a joint spatio-temporal system to directly reconstruct oxygen consumption and blood flow from photon measurements. Both methods show contrast between the background and tumor regions; performance comparison between the direct and indirect methods as implemented are inconclusive, despite the considerably greater computational cost of the direct method.

Committee:

Prof. Eric Miller
Dr. Stefan Carp (MGH)
Prof. Dana Brooks (advisor)