Molecular and cellular imaging of bone biopsies using AI augmented deep UV Raman microscopy

点击翻译按钮获取中文摘要

An exploratory research project will develop deep-UV Raman microscopic hyperspectral imaging for molecular and/or cellular analysis of biological tissues with a goal of the early detection, improved screening, and clinical diagnostics of cancer. Raman microscopy is often used in cancer biology to identify occurring chemical changes; however, the sensitivity and specificity of detection remain to be a challenge. This gap of fundamental knowledge on how to improve the information context of such images will be addressed by utilizing deep UV excitation, which, through resonance excitation of specific molecules will enhance specificity of molecular detection and improve the sensitivity by enhancing the signal against the background. To further improve the image-based analysis and screening, a novel hyperspectral image analysis platform will be developed. The proposed research program fills the technology gaps by developing an instrument, capable of performing Raman imaging at least 100 times faster, acquire new information through assessing low-frequency Raman modes, while reducing the cost and the footprint to accelerate the wide-spread availability of the instrument. The new imaging system augmented with novel hyperspectral imaging algorithms to handle multidimensional imaging data will be applied to advance a challenging biopsy of bone tumors, one of the most devastating consequences of many cancers with the goal to achieve 95% specificity. In Aim 1, a novel, patent-pending, wide-field deep UV hyperspectral Raman imaging platform will be optimized for cancer tissue samples. A working prototype will be built, and its performance will be experimentally characterized. In Aim 2, a data analysis platform with machine and deep learning algorithms for pathology of bone tissue will be developed. Advanced imaging algorithms that take into account many small changes in addition to a traditional analysis of Raman spectra will be used. Machine learning and deep learning techniques will be developed to automatically determine abnormalities beyond current yes/no tumor paradigm. In Aim 3, the developed platform will be validated as a novel analysis strategy. Research will focus on distinguishing tumors in the animal model of metastatic bone cancer and developing a set of optical markers to enable rapid identification of tumors. The proposed strategy offers a novel enabling technology to elucidate basic mechanisms underlying cancer initiation and progression and will facilitate early cancer detection, screening, and/or cancer risk assessment, by differentiating, evaluating and/or observing cancer stages and progression. The overall approach targets the wide spread of the technology, its relatively low-cost and seamless transition to clinical setting. The R33 phase will improve the sensitivity of detection and identify the pathways toward commercialization. The research study will also provide a roadmap to develop a new advanced approach for studying a variety of bone-related tumors and identify novel preclinical and clinical assays.