Deep learning quantification of mouse nesting behavior for tracking cognitive decline in models of aging & Alzheimer’s disease

Abstract

Abstract The aging brain and development of Alzheimer’s disease (AD) are characterized by neurological degeneration. Nest-building performance captures deteriorating cognitive and motor functions in rodent models of AD; albeit qualitative while limited in physiological scope. To enhance sampling depth and frequency, we developed an artificial intelligence (AI)-powered image analysis pipeline (“CozyScores”) built on the ResNet18 architecture. In male & female C57BL6/J wild-type (WT) & 3xTg-AD mice (AD) mice [Young, 3–6 mo; WT, n = 28; AD, n = 32) & (Old, 21–26 mo; WT, n = 20; AD, n = 12)], we monitored progressive nest organization using photo acquisition across timepoints (0.5 to 24 h) relative to baseline on a 5-point scoring system. At the 4 h benchmark timepoint, a significant ( P < 0.05) difference in performance scores was detected as a function of age and genotype [Young WT (mean ± SEM, 3.80 ± 0.18) > Young AD (3.16 ± 0.24) > Old WT (3.08 ± 0.32) > Old AD (1.71 ± 0.14)]. At 24 h, maximum performance scores declined in Old (WT: 4.24 ± 0.13; AD: 3.59 ± 0.35) versus Young (WT: 4.60 ± 0.05; AD: 4.33 ± 0.13) groups. Males generally performed better than females throughout groups, particularly during nest construction (2 to 8 h). Overall, CozyScores is an automated, high-resolution analysis of behavioral kinetics in mouse models of aging and AD.