Mathematical Chemical Imaging with Uncertainty Quantification

 描述（由申请人提供）：当今生物医学科学的一个关键主题是空间和分子信息的整合，以预测复杂系统的行为。化学成像（CI）是一种令人兴奋的新范式，它可以同时记录样品的空间结构和分子光谱信息，有望大大扩展当前的显微镜方法。中红外（IR）Cl成像的最新进展现在允许在几分钟内快速记录完整的图像集，使各种各样的应用，从可视化皮肤扩散，生物材料评估到癌症病理学。IR Cl数据通常由10兆像素组成，每个像素包含2000个光谱频率，并且在每个频率处的吸光度值在0和1之间（噪声水平从10-4到0.1，取决于实验参数）。与生物医学成像中常规使用的分子探针或染料相反，计算工具是从Cl数据中提取信息的唯一途径。目前限制进展的障碍是记录的数据非常大（100 GB），所有波数的吸收可能不包含有用的信息，某些频率具有冗余信息，并且通常需要1000：1的相对较高的信噪比（SNR）。在这些挑战得到满足之前，Cl的全部潜力无法实现，并且有用的生物医学成像是不可能的。计算数学组和光谱成像组之间的合作的目标是以一种新的方式解决现存的Cl挑战。我们的互补专业知识将开发从根本上提取知识的新方法，并将其整合到仪器中，以改变IR Cl的实践，并使其能够被生物医学科学家放心使用。

 DESCRIPTION (provided by applicant): A key theme in biomedical sciences today is the integration of spatial and molecular information to predict the behavior of complex systems. Chemical imaging (CI) is an exciting new paradigm that simultaneously records both spatial structure and molecular spectral information from a sample, promising to considerably extend current microscopy methods. Recent advances in mid-infrared (IR) Cl imaging are now allowing rapid recording of full image sets within minutes, enabling a wide variety of applications ranging from visualizing diffusion in skin, to biomaterial evaluation to cancer pathology. IR Cl data typically consist of 10 megapixels, with each pixel containing 2000 spectral frequencies and an absorbance value between 0 and 1 (with noise levels from 10-4 to 0.1, depending on experimental parameters) at each frequency. As opposed to molecular probes or dyes conventionally used in biomedical imaging, computational tools are the only route to extracting information from Cl data. The barriers limiting progress today are that recorded data are exceptionally large (100GB), absorbance at all wave numbers may not contain useful knowledge, some frequencies have redundant information and a relatively high signal to noise ratio (SNR) of 1000:1 is often required. The full potential of Cl cannot be realized and useful biomedical imaging is impossible until these challenges are met. The goal of this collaboration between a computational mathematics group and a spectroscopic imaging group is to do address extant Cl challenges in a novel manner. Our complementary expertise will develop fundamentally new methods for extracting knowledge and integrate them into instrumentation to transform the practice of IR Cl and make it confidently usable by the biomedical scientist.