多波長複頻驅動擴散光學立體造影系統研製與驗證

Multi-wavelength near infrared diffuse optical imaging (NIR DOI), belonging to functional imaging, enables to distinguish tumor tissue from normal one, as well as characterizes oxygen content distribution. Since middle 90s it has been applied in tumor screening and diagnosis for soft tissue such as breast, and performed for clinical trial besides continual technical improvement. Based upon the previous experience in the study on DOI, especially the current two-year project “Study on Hybrid Diffuse Optical Imaging Technique and System Integration” granted by the MOST that has been performed for improving the previous prone-type imaging system including (i) the development of DOpIm interactive imaging system, (ii) the investigation of hybrid (model based and data driving) imaging approach, and (iii) the improvement of opto-electrical measurement (periodic power driving light source, and flexible optical measurement channels). This proposed three-year project aims to enhance this stereoscopic DOI system in the following issues:(1)to extend the implementation of DOpIm imaging program including 3D edgepreserving regularization, and the best selection of regularization parameter ;(2)to deepen the development of deep-learning based hybrid (model-based and data-driven) imaging approach that combines radiance and raw image data to reconstruct optical-property images through deep-learning computation, and knowledge transfer deep learning model;(3)to further polish the prone-type DOI system through implementing combinedfrequency driving NIR, employing multi-wavelength NIR, and using line light source to reconstruct CAD model of subject to be test.This project will continue to perform human subject experiments for breast tumor detection, and collect clinical data at the Landseed International Hospital.Subsequently, we can examine the sensitivity and specificity for the developedstereoscopic DOI system.