Development and Dissemination of MuscleMiner: An Imaging Informatics Tool for Mus

描述（由申请人提供）：骨骼肌活检图像评估是研究和临床实践中必不可少的程序。虽然被广泛使用，但目前的肌肉图像形态测量、存档、可视化、查询、搜索、检索和挖掘过程存在几个主要的局限性：1)虽然传统的形态测量参数，如横断面积（CSA）和最小Feret直径等，是评估肌肉功能的关键指标，但目前的测量仍然主要基于人工或半自动化的方法，导致巨大的人工成本和潜在的观察者之间的差异。2)目前肌肉图像的归档仍然主要基于过时的工具，如Excel电子表格和电脑文件夹。给定一个新的肌肉图像，几乎不可能快速交叉比较、可视化、查询、搜索和检索以前的病例，这些病例表现出相似的图像内容，具有可比性的形态计量测量，目的是在实验室发现新的生物学相关性，或在床边提供个性化的诊断和预后。3)尽管一张典型的肌肉图像通常包含数百万个数据点（像素），但在临床实践中，医生经常将这些丰富的信息浓缩成一两个诊断标签，而丢弃其余的。新的图像标记，并不总是通过目视检查明显或不能人工量化，但潜在地代表了精确医学的关键诊断和预后价值，尚未得到严格的检查。同样，在基础科学研究中，只考虑非常有限数量的已知测量（例如，CSA）。一些非传统的测量方法，如肌纤维形状，有可能作为肌肉功能的新指标，尚未得到充分的研究。4)目前的肌肉图像分析和搜索功能吞吐量较低。云计算作为一个前沿研究领域，可以提供高吞吐量的计算能力，以分布式的方式处理大图像数据。然而，它在肌肉图像上的应用从未被探索过。MuscleMiner将为自动图像形态测量、存档、可视化、查询、搜索、基于内容的图像检索、生物信息学图像挖掘和云计算提供一套完整的工具。本提案的目标是：1)开发自动形态测量单元，基于内容的图像检索（CBIR）单元，以及图像存档和可视化单元。2)开发先进的生物信息学图像挖掘单元，以协助快速发现和验证新的图像标记。3)开发云计算单元，实现大图像数据处理和搜索功能。向肌肉研究社区传播这个免费可用的云成像信息系统。

DESCRIPTION (provided by applicant): Image evaluation of skeletal muscle biopsies is a procedure essential to research and clinical practice. Although widely used, several major limitations exist with respect to current muscle image morphometric measurement, archiving, visualization, querying, searching, retrieval, and mining procedures: 1) Although traditional morphometric parameters, such as cross-sectional area (CSA) and minimum Feret diameter, etc., serve as critical indicators for assessing muscle function, current measurements are still largely based on manual or semi-automated methods, leading to significant labor costs with large potential inter-observer variability. 2) The current archiving of muscle images is still mainy based on outdated tools such as Excel spreadsheets and computer file folders. Given a new muscle image, it is almost impossible to quickly cross-compare, visualize, query, search, and retrieve previous cases exhibiting similar image contents with comparable morphometric measures, for the purpose of either discovering novel biological co-correlations at benchside, or providing personalized diagnosis and prognosis at bedside. 3) Although a typical muscle image often contains millions of data points (pixels), in clinical practice, doctors often condense this rich information into one or two diagnostic labels and discard the rest. Novel image markers, which are not always apparent through visual inspections or not manually quantifiable, but potentially represent critical diagnostic and prognostic values for precision medicine, have not been rigorously examined. Similarly, in basic science research, only a very limited number of known measures (e.g., CSA) are considered. Some non-traditional measures, such as myofiber shapes that hold the potential to serve as new indicators of muscle functions, are not fully investigated. 4) Current muscle image analysis and searching functions are fairly low throughput. As a frontier research area, Cloud computing can handle big image data in a distributed manner by providing high-throughput computational power. However, its application to muscle images has never been explored. MuscleMiner will provide a complete suite of tools for automated image morphometric measurements, archiving, visualization, querying, searching, content-based image retrieval, bioinformatics image mining, and Cloud computing. The objectives of this proposal are to: 1) Develop the automated morphometric measurement unit, content-based image retrieval (CBIR) unit, and the image archiving and visualization unit. 2) Develop the advanced bioinformatics image mining unit to assist in the rapid discovery and validation of new image markers. 3) Develop the Cloud computing unit to enable big image data processing and searching functions. Disseminate this freely available, Cloud-enabled imaging informatics system to muscle research community.