Neural network classification of Barnes maze search strategy utilization

Abstract

The Barnes maze is a commonly used test of allocentric spatial reference memory, consisting of an elevated circular table with holes around the perimeter. Spatial cues surrounding the maze are intended to allow the animal on the maze to locate a target hole from which they can escape and return to their home cage during a period of acquisition trials. Following the acquisition period, the target box under the target hole is removed and a probe trial is performed to test spatial memory. One of the limitations of Barnes maze testing is that non-spatial strategies can be employed to locate the target hole, such as systematic serial searching hole to hole, which may mask the signal of spatial learning or memory deficits on gross outcome measures, such as the latency to find the target hole. Quantifying the search strategies used during Barnes maze testing can provide a more direct measurement of spatial memory impairments, and even allow for the detection of impairments to executive functioning and working memory that may impact the ability to utilize non-spatial systematic search strategies. We have developed a machine learning algorithm to automatically quantify search strategy utilization in an unbiased manner. Traumatic brain injury (TBI) is known to cause impairments in spatial learning and memory, and in our testing with a model of repetitive mild TBI we have found significant deficits in the utilization of both spatial and systematic non-spatial search strategies 3 months after TBI. Understanding the factors driving gross outcome measures on the Barnes maze may provide greater insight into the effectiveness of potential treatment strategies designed to mitigate the chronic effects of TBI and other neurodegenerative diseases.