The long term goal of this research is to develop a near infrared (NIR) time- resolved optical imaging system based on measurements performed in the frequency domain for the detection of tumors in breast tissue. Because of the intense scattering of light by tissue, the ability to generate high quality images from optical measurements will require evaluation of signals that have been highly scattered. The specific goal of this study is to identify and optimize conditions for data collection and analysis by evaluation of anatomically accurate optical (AAO) models of the female breast. The applicants have adopted this approach because we believe it represents an efficient and cost-effective means for evaluation and optimization of critical imaging parameters under conditions that closely resemble those that would be obtained from actual measurement, without the numerous engineering and other practical constraints and significantly higher costs that almost certainly would be encountered in similar studies involving human subjects. The applicants will construct AAO models by assigning optical absorption and scattering coefficients to the corresponding tissue structures identified in segmented 3-D MRI images of the female breast of healthy volunteers and patients having known neoplastic and non-neoplastic lesions. Determination of these coefficients will be performed using a previously described dual-integrating sphere measurement technique on freshly prepared thin slices of breast tissue obtained from surgical specimens. Definitive assignment of the tissue type will be performed by microscopic examination of stained adjacent tissue sections. Determination of optimal conditions will be based on examination of the tradeoffs required for data collection, solution accuracy, and required computing time, and will proceed in two directions. Solutions to the forward problem will involve computing detector responses and associated imaging operators for the various AAO model media as a function of target geometry and presence of simulated pathological inclusions of varying size and contrast. These will involve evaluation of newly available, highly efficient codes for solution of the transport and diffusion equations, thereby permitting a direct answer to the important question of the suitability of solutions to the diffusion equation, which is computationally more efficient than the former, under conditions that closely resemble the anatomy, geometry and optical properties of the female breast. The inverse problem for image reconstruction will be evaluated by solving a linear perturbation equation for selected data sets, using a variety of approaches that vary in their computational efficiency and ability to deal with problems stemming from ill-conditioning and ill-posedness. Previously developed layer-stripping and multigrid schemes will be evaluated and further improved; this will include development of a wavelet transform method. The sensitivity and accuracy of reconstructed images to noise in the data and imaging operators, choice of illumination scheme, view angle, and number of source-detector positions and other parameters will be explored.
