A tool to automate assessment of regional brain atrophy in mouse models of neurodegenerative disease

As life expectancy rises, so too does the prevalence of neurodegenerative diseases. Neurodegeneration causes progressive regional brain atrophy, typically initiating prior to symptom onset. Researchers measure the impact of potential treatments on atrophy in mouse models to assess their effectiveness. This is important because treatments designed to combat neuropathology are more likely to modify the disease, per contra to symptom management. Magnetic resonance imaging, while accurate in measurement of brain region structure volumes, is prohibitively expensive. Conversely, stereological volume assessment, the process of estimating the volume of individual 3D brain regions from imaged 2D brain sections, is more commonly used. This involves manually tracing brain region(s) of interest in regularly spaced imaged cross-sections to determine their 2D area, followed by application of the Cavalieri principle to estimate the volume. The pertinent caveats of this approach are the labor-intensive manual tracing process, and potential inaccuracies that arise due to human variation. To overcome these challenges, we have created a Neuropathology Assessment Tool (NAT) to automate regional brain tracing and identification using artificial intelligence (AI) and concepts from topological data analysis. The NAT was validated by comparing manual and NAT analysis of striatal volume in Huntington disease model mice. The NAT detected striatal atrophy with higher efficiency, 93.8% agreement with manual measurements, and lower inter-group variability. The NAT will increase efficiency of preclinical neuropathology assessment, allowing for a greater number of experimental therapies to be tested and facilitating drug discovery intractable neurodegenerative diseases.