Optical Nevoscope reconstructions using photon Diffusion theory

Analysis of 3D shape and size of skin lesions and features obtained by analyzing the images of skin-lesions, may help early diagnosis of melanoma. Our efforts in non-invasive imaging of skin lesions have been in the direction of characterization of the lesion volumes with transillumination based optical imaging modality. Light photons after entering into the skin interact with matter for multiple scattering events. The back scattered radiation re-emerging from the skin forms two dimensional projection images of the transilluminated skin-lesion. These projection images are obtained through direct view and mirror reflections. The problem of solving for the volume of the lesion from these projections is a non-linear inverse problem. We attempt to solve this using dipole solutions to the diffusion equation. The objective here is to obtain the jacobian (voxel weights) of the system of equations and progressively solve linear inverse problem. An iterative non-linear inversion method is proposed, which incorporates the salient features of the rigorous perturbation Monte Carlo method at much lower computational overheads, to compute the jacobian. The jacobian obtained based on diffusion theory is used in reconstructions of the lesions based on algebraic reconstruction techniques. A milk-gelatin phantom of tissue and embedded lesion is constructed in which the concentration of milk determines extent of scattering. The absorber is simulated by adding ink to the same mixture in an embedded capillary tube. The phantom is trans-illuminated using the Nevoscope. Reconstructions of the phantom based on Diffusion theory are obtained. Our results show smoother reconstructions for the Diffusion theory based method.