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

The study proposed and applied an AI-powered image analysis pipeline, “CozyScores,” built on the ResNet18 architecture, to automatically and quantitatively capture the kinetics of mouse nesting behavior at high resolution. Compared with prior assessments of nest-building performance that were qualitative and limited in physiological scope, the pipeline used photo acquisition from 0.5 to 24 h and a 5-point scoring system to enhance sampling depth and frequency. Evidence comes from the abstract’s description of the pipeline architecture (ResNet18) and monitoring approach (0.5 to 24 h, 5-point scoring); the loaded text does not include full methodological details, so assessment of pipeline specifics and scoring agreement remains pending the full article.

At the 4 h benchmark timepoint, nesting performance scores showed a significant difference by age and genotype (P < 0.05), ordered as Young WT 3.80±0.18 > Young AD 3.16±0.24 > Old WT 3.08±0.32 > Old AD 1.71±0.14. This result quantifies nesting deficits in aging and AD models as comparable kinetic scores and covers male and female, young/old, and WT/3xTg-AD groups. The abstract reports P < 0.05, mean±SEM, and group sample sizes (Young WT n=28, Young AD n=32, Old WT n=20, Old AD n=12); full statistical models and multiple-comparison corrections require the original article.

Maximum 24 h performance scores were lower in Old groups than in Young groups (Old WT 4.24±0.13, Old AD 3.59±0.35; Young WT 4.60±0.05, Young AD 4.33±0.13), and males generally performed better than females, particularly during the 2 to 8 h nest-construction period. This adds observations of time-dependent nesting performance and sex differences beyond a single endpoint score. Based on the group mean±SEM and sex-trend descriptions reported in the abstract; effect sizes and statistical details for sex differences are not elaborated in the abstract.

The work is aimed at preclinical aging and AD research settings: it uses male and female C57BL6/J wild-type and 3xTg-AD mice in young (3–6 months) and old (21–26 months) groups, monitoring nest organization with photo acquisition from 0.5 to 24 h and a 5-point scoring system. It enables researchers to obtain denser kinetic readouts during nest construction and to compare age, genotype, and sex within one framework; it can subsequently be used in experimental settings that require automated, repeated behavioral scoring.

Readers would still watch how well CozyScores AI scores agree with manual 5-point scoring, whether the model generalizes across cage setups, lighting, bedding, or strains, how 4 h and 24 h scores relate to other cognitive/motor tests and AD pathology measures, and how robust the sex differences are. The currently loaded text is only the abstract and metadata, lacking figures, tables, and methodological detail, so these questions remain open within what can be summarized here.