Multispectral Retinal Imaging and Mapping of Naturally Occurring Fluorophore and Chromophore Distributions in Health and Early Pathology

本提案的总体目标是将鲁棒的自动化多光谱图像分析方法与多光谱视网膜反射和自身荧光成像相结合，以提供视网膜内自然发生的荧光团和发色团分布的定量图。由此产生的高分辨率多组分分子图谱将用于更好地理解正常视网膜衰老和早期视网膜疾病进展中的分子事件。-利用标准眼底相机中使用特殊滤光片组获得的不同多光谱自荧光和反射图像，通过自动图像分析方法生成分子图像地图，优化其质量。设想作为一个迭代的过程细化过滤器，成像协议和分析工具。-应用光谱线性解混模型提取纯亚像素光谱特征，以获得多种内源性分子物种局部浓度的临床重要变化-开发基于降维算法的光谱分类方案，以对局部细胞功能和病理进行分类-开发局部支持的多尺度（小波）表示系统，用于可重复特征的稀疏表示，以便跟踪随时间变化。特别是不同类别的早期病理病变在100微米量级。

0854233CzajaThe overall objective of this proposal is to integrate robust automated multispectral image analysis methods with multispectral retinal reflectance and autofluorescence imaging to provide quantitative maps of naturally occurring fluorophores and chromophore distri-butions within the retina. The resulting high-resolution multi-component molecular maps would be applied to better understand molecular events in normal retinal aging and in early retinal disease progression.Tasks:- To optimize the quality of molecular image maps generated by automated imageanalysis methods from different multispectral sets of autofluorescence and reflectance images obtained using special filter sets in standard fundus cameras. Envisioned as an iterative process of refinement of filters, imaging protocols and analysis tools.- To apply spectral linear demixing models for extraction of pure subpixel spectral signatures to obtain clinically important changes in local concentrations of multiple endogenous molecular species- To develop spectral classification schemes based on dimension reduction algorithms to classify local cellular function and pathology- To develop locally supported multi-scale (wavelet) representation systems for sparse representation of reproducible features in order to follow changes overtime, particularly for different classes of early pathologic lesions on the order of 100 microns.