**Background**
Rapid behavioral tests (e.g., open-field, elevated plus maze) are widely used for mouse phenotyping but have drawbacks: they are conducted during the day (when nocturnal mice are inactive), involve handling stress, lack acclimation, and require moving mice between devices. Home-cage imaging over several days can overcome these issues but is challenging for large numbers of mice. The authors developed an 8-cage imaging system with a projector for background illumination (white light daytime, red light nighttime) and visual stimuli at night, aiming to fill the gap between rapid tests and long-term home-cage analyses.
**Methods**
The system consists of a dark box with two webcams above eight cages and a projector below, illuminating the cage bottoms via a mirror. Cages have four quadrants: Home (red house), Wall (open area with two walls), Food (food/water tray), and Window (screened window). Mice were imaged for 22 hours (1 frame/second) starting at 2 pm, with 6 hours daytime (white light), 6 hours nighttime (red light), then 6 hours of visual stimuli (rotating yellow moth, moving red lines in the Home quadrant) on red background, followed by 4 hours daytime. Two analysis pipelines were used: (1) an ImageJ/Fiji macro that thresholds the red channel, performs particle analysis, and calculates centroid, speed, movement (% time >1 cm/s), stretch-attend posture (SAP; speed <1 cm/s and ellipse aspect ratio >2.3), and quadrant location; (2) DeepLabCut (DLC) trained on 430 frames with 48 labeled points (12 mouse body parts, 15 cage features, 21 stimuli). For validation, 8 female B6129SF2/J wild-type mice (4–5 months old) were imaged once. Subsequently, 8 female 3xTg-AD mice (same age) were imaged and compared to wild-type. Statistics used Welch’s t-test with Bonferroni correction.
**Key Results**
- **Activity in wild-type mice:** Average speed was higher in the first hour (0.97 cm/s) than the second hour (0.30 cm/s; p=2×10⁻⁴) and higher at nighttime (0.70 cm/s) than daytime (0.28 cm/s; p=3×10⁻³). Movement (% time >1 cm/s) was 20% in the first hour vs. 6% in the second (p=3×10⁻⁴) and 13% at night vs. 5% daytime (p=9×10⁻⁴). The rotating yellow moth induced a non-significant 10% increase in activity during the first 10 minutes. The first set of moving lines caused a 10% increase (not significant) then a 16% decrease (p=4×10⁻³), indicating habituation; the second set had no effect.
- **Stretch-attend posture (SAP):** SAP was more frequent in the first hour (19%) than the second (6%; p=0.007) and at night (17%) vs. daytime (6%; p=0.01). The first moving lines increased SAP during the first 10 minutes (15% vs. 3% in subsequent 50 min; p=3×10⁻³).
- **Quadrant preference:** Wild-type mice avoided the Wall quadrant for the first 11 hours (all p<0.001) and the Food quadrant for the first 6 hours (all p<0.001).
- **Comparison with 3xTg-AD mice:** AD-model mice showed reduced acclimation (difference in % moved between first and second hour: 13% wild-type vs. 8% AD; p=0.04), hyperactivity during the seventh hour (first hour of night: 8% vs. 23%; p=0.02), and less time in the Home quadrant during daytime hours 2–6 (57% vs. 14%; p=0.03).
- **DLC vs. ImageJ:** On 200 manually labeled images, Fiji had an average centroid error of 6 pixels (6.5 mm) and DLC 10 pixels (11 mm). DLC tracked 12 mouse markers vs. 2 with Fiji.
**Clinical Implications**
The 8-cage system enables high-throughput (32 mice/week), automated behavioral phenotyping with minimal handling and acclimation. The hyperactivity, reduced acclimation, and decreased home-staying in 3xTg-AD mice parallel features of Alzheimer’s disease (e.g., sundowning, wandering). This system can be used to evaluate transgenic models and potential therapeutics. Limitations include lack of bedding (interferes with bottom illumination), no acoustic stimuli, and 1 fps imaging (30 fps would produce massive files). The system is inexpensive, uses free software (ImageJ, Fiji, DLC), and includes downloadable PowerPoint stimuli and Excel templates.