Imaging through scattering

[摘要] In this thesis we demonstrate novel methods to overcome optical scattering in order to resolve information about hidden scenes, in particular for biomedical applications. Imaging through scattering media has long been a challenge, as scattering corrupts scenes in a non-invertible way. The use of near-visible optical spectrum for biomedical purposes has many advantages, such as optical contrast, optical resolution and nonionizing radiation. Particularly, it has important applications in biomedical imaging, such as sub-dermal imaging for diagnostics, screening and monitoring conditions. We demonstrate methods to overcome and use scattering in order to recover scene parameters. In particular we demonstrate a method for locating and classifying fluorescent markers hidden behind turbid layers using ultrafast time-resolved measurements with a sparse-based optimization framework. This novel method has applications in remote sensing and in-vivo fluorescence lifetime imaging. Another method is demonstrated to resolve blood flow speed within skin tissue. This method is based on a computational photography technique and coherent illumination. This method can be applied in diagnosis and monitoring of burns, wounds, prostheses and cosmetics. A particularly important application of this technology is analysis of diabetic ulcers, which is the main cause for non-traumatic amputations in India. The suggested prototype is suitable for assisting clinicians in assessing the wound healing process. The methods developed in this thesis using ultrafast time-resolved measurements, sparsity-based optimization and computational photography can spur research and applications in biomedical imaging, skin conditions diagnosis and more general modalities of imaging through scattering media.

[发布机构] Massachusetts Institute of Technology