A human brain consists of billions of nerve cells (neurons) interconnected by their projection fibers called axons. It is believed that the number of axons, their size and shape are essential in understanding the normal or abnormal, the developing or aging, as well as the injured or diseased brains. Therefore, a quantitative knowledge of the number of axons, as well as their size and shape, is essential in understanding the rain. A method that automates the analysis of axon imagesis presented. The heart of our system relies on a segmentation scheme called the active contour model (snake). Though the framework of snake has already been established, a number of unresolved issues remains for practical applications. Specifically, we have addressed three problems associated with the analysis...[ Read more ]

A human brain consists of billions of nerve cells (neurons) interconnected by their projection fibers called axons. It is believed that the number of axons, their size and shape are essential in understanding the normal or abnormal, the developing or aging, as well as the injured or diseased brains. Therefore, a quantitative knowledge of the number of axons, as well as their size and shape, is essential in understanding the rain. A method that automates the analysis of axon imagesis presented. The heart of our system relies on a segmentation scheme called the active contour model (snake). Though the framework of snake has already been established, a number of unresolved issues remains for practical applications. Specifically, we have addressed three problems associated with the analysis of the axon images : (1) initialization of snakes; (2) energy functional formulation and (3) elimination of false alarms. A transform approach, based on a piecewise elliptical contour approximation to the axon boundaries, is proposed to locate an interior point for each axon, thereby serving as a clue for snake initialization. An energy functional favoring equal separation between the inner and outer boundaries is formulated to facilitate the detection of outer boundaries. False alarms are eliminated by means of a conflict resolution scheme which is reduced to an optimization problem implemented as O-l integer programming. Our developed method has been tested on a number of nerve images and its results are favorable.