Three dimensional near-infrared optical imaging using ultrasound guidance for breast cancer detection and diagnosis

Date of Completion

January 2004


Engineering, Biomedical|Engineering, Electronics and Electrical|Health Sciences, Radiology|Health Sciences, Oncology




This dissertation focuses on the study of three-dimensional near-infrared (NIR) imaging reconstruction using ultrasound localization for breast cancer detection and diagnosis. With the 3D NIR forward model for reflection geometry built and investigated, successful reconstruction of both absorption and scattering coefficients has been obtained and evaluated using both a phantom study and initial clinic data. ^ In this work, we first introduce a dual mesh optical reconstruction method with depth correction using a priori ultrasound information. It can significantly improve the reconstructed absorption distributions in deeper target layers. Clinical results have shown improved correlation between reconstructed total hemoglobin concentrations and the histological micorvessel density counts. ^ Next, a combined Tikhonov regularization and dual mesh scheme based on finite element method (FEM) is proposed to solve the inter-parameter cross talk between the scattering coefficient and absorption coefficient. In this approach, the optimal regularization factor is given by the standard L-curve. Different regularization factors have been assigned to the target region and non-target region to incorporate the priori target information from the co-registered ultrasound images. Both phantom experiments and clinic examples have shown that this new approach has the ability to reconstruct the absorption and diffusion coefficients simultaneously, and the reconstructed target absorption and diffusion maps have no depth dependence effect. ^ Finally, a simple two-layer model is generated and studied to solve a clinical issue related to the chest wall, which is a mixture of bones and muscles. For small breasts, the chest wall appears within 2–3 cm distance from the skin surface. As a result, the simple one layer model is not accurate. Furthermore, for some cases, there is also an obvious chest wall mismatch between the reference site and lesion site. To solve this clinical issue, a simplified two-layer model is generated and a fitting method is adopted to estimate the optical properties of both layers. Then a correction method is applied to correct for the chest wall mismatch between the lesion site and reference site. With this scheme, phantom targets located on top of the chest wall phantom layer can be reconstructed with good contrast and resolution. ^