The Use of Support Vector Machines and Hierarchical Linear Modeling in Morris Water Maze Analysis

Abstract

This thesis describes a methodology that can be used in the analysis of Morris water maze data. The Morris water maze is a test which measures an animal’s spatial learning performance. In the Morris water maze, an animal is placed in a pool of opaque water, rendering it unable to visibly locate a platform used to escape. Thus the animal is made to learn the location of the platform relative to external spatial cues. This thesis sought to take videos of Morris water maze trials, analyze them to produce a series of coordinates, and then use these data to produce variables describing paths swum in the Morris water maze, such as trial duration, rotation, and measures describing the distance from the platform and perimeter. These values were then used to classify trials into a search strategy using support vector machines. An analysis was then performed and described to demonstrate how this output can be examined. Different mouse models for Alzheimer’s disease were used (Wildtype, Amyloid-only, Tau-only, Dual transgenic) for analysis, allowing us to examine how different aspects of AD pathology affect search strategy. Output indicated that as more advanced aspects of Alzheimer’s disease pathology are modeled, animals show increasing rates of a non-spatial search strategy (thigmotaxis).