AUTOMATED 3D TRACING AND ANALYSIS OF NEUROANATOMY

点击翻译按钮获取中文摘要

The long-term objective is the development of computer software to perform rapid and fully automated tracing of neurons. This will eliminate the tedium, slowness, cost, inaccuracy and subjectivity of current manual/semiautomatic tracing methods. It will enable new capabilities, such as rapid quantitative studies over large batches of specimens, automated quantitation of structural changes in response to stimuli, quantitative connectivity analysis, accurate measurement of geometric features such as volumes, lengths and tortuosities, and automated analysis of structures such as dendritic spines. Our earlier work demonstrated successful tracing of selectively-stained rat hippocampal neurons starting from a 3D confocal image stack. The method was based on 3D segmentation and skeleton extraction, followed by the detection of branching and terminating points, and computation of a graph-theoretic topological representation of the neuron. The proposed research will extend the present method to other microscopy modalities, incorporate needed additions such as automatic soma detection and image montaging, and establish the feasibility of performing routine, unsupervised, and robust processing of large batches of complex data sets involving a variety of morphologically diverse neurons from the visual cortex, and CA1 and CA3 regions of the hippocampus. PROPOSED COMMERCIAL APPLICATION: The most immediate, albeit limited, market is in the automated tracing of individual neurons (about 100 sites). For multiple montaged fields of multiple neurons and circuits, immediate potential sites include every research Medical School worldwide (2000-5000 sites). The product, over 5 to 10 years, will grow into clinical pathology and ophthalmology applications in tracing microvasculature. Although not an immediate market, there are another 2000 to 5000 (or more) of these future clinical sites.