Reconstructed Near Infrared Diffusion Imaging for Breast Cancer Detection

Finite element based reconstruction algorithms for optical property profiling in biological tissues have been developed, implemented and evaluated using simulated data and laboratory measurements obtained with near infrared continuous wave (cw) and amplitude-modulated laser illumination systems. The algorithms are based on finite element solution of the diffusion equation where Newton's method is used to update an initial estimate of the optical property distribution. Extensive simulations and laboratory experiments using tissue-like phantoms have been conducted in both the steady-state and frequency domains in order to assess the overall imaging capabilities of the reconstruction algorithms and the measurement systems. Image enhancement schemes including spatial filtering, total variation minimization and dual meshing have been proposed and developed in order to improve the resulting image quality. The important issues of imaging resolution and contrast in the presence of single and multiple heterogeneities have also been experimentally studied. Simulations and experiments have shown that both absorption and scattering coefficient maps can be simultaneously reconstructed using absolute imaging techniques. These images are the first of their kind to show recovery of single and multiple objects having both absorption and scattering contrast with the background which are based on absolute imaging procedures. More importantly, the reconstructed images obtained from experimental frequency -domain data have been quantitative not only in terms of the "tumor" size, location and shape, but also the optical property values, themselves. This quantitative nature of the recovered optical properties is an important step towards clinical breast cancer diagnosis since the differentiation between normal tissues and benign and malignant tumors will likely rely on this quantitative capability. The application of image enhancement schemes in the reconstruction algorithms developed have shown that the quality of the images produced can be improved considerably. In particular, the most impressive gains are found in the quantitative nature of the recovered optical property values when these methods are included. Further, imaging experiments using both steady-state and frequency-domain measurements have demonstrated that an object as small as 4 mm in diameter in an 86 mm background region can be recovered quantitatively and multiple objects as close as 20 mm (center-to-center) can also be resolved quantitatively over a wide range of contrasts with respect to the background. This work has shown that laboratory scale experiments using representative contrast levels and reasonably small sizes of targets (approximately 20:1, background thickness relative to embedded anomaly size) result in quantitative images which suggests that near infrared diffusion optical imaging using reconstruction schemes may become a useful clinical tool for breast cancer detection.