Validation of video motion-detection scoring of forced swim test in mice

Abstract

Background: The forced swim test (FST) is used to predict the effectiveness of novel antidepressant treatments. In this test, a mouse or rat is placed in a beaker of water for several minutes, and the amount of time spent passively floating is measured; antidepressants reduce the amount of such immobility. Though the FST is commonly used, manually scoring the test is time-consuming and involves considerable subjectivity.

New method: We developed a simple MATLAB-based motion-detection method to quantify mice's activity in videos of FST. FST trials are video-recorded from a side view. Each pixel of the video is compared between subsequent video frames; if the pixel's color difference surpasses a threshold, a motion count is recorded.

Results: Human-scored immobility time correlates well with total motion detected by the computer (r=-0.80) and immobility time determined by the computer (r=0.83). Our computer method successfully detects group differences in activity between genotypes and different days of testing. Furthermore, we observe heterosis for this behavior, in which (C57BL/6J×A/J) F1 hybrid mice are more active in the FST than the parental strains.

Comparison with existing methods: This computer-scoring method is much faster and more objective than human scoring. Other automatic scoring methods exist, but they require the purchase of expensive hardware and/or software.

Conclusion: This computer-scoring method is an effective, fast, and low-cost method of quantifying the FST. It is validated by replicating statistical differences observed in traditional visual scoring. We also demonstrate a case of heterosis in the FST.

Keywords: Automation; Depression; Forced swim test; Heterosis; Overdominance; Strain differences; Video analysis.

Fig. 1

Sample motion record. The motion record of a B6 mouse in the Acute Stress group is displayed here, after applying a smoothing function. 6 minutes of the FST is displayed. This mouse started with high activity while struggling to escape the beaker, ceased activity abruptly for several seconds before resuming swimming, then settled into alternating bouts of passive floating and swimming.

Fig. 2

Computer scoring vs. visual scoring. (A) Comparison of computer-scored total motion and visual-scored immobility time. Units on the computer axis are arbitrary. Each point represents one mouse (n = 134). Pearson correlation r = −0.80. (B) Immobility time measured by computer-scoring and visual-scoring. Pearson r = 0.83. (C) Latency to immobility. Pearson r = 0.79, Spearman ρ = 0.85.

Fig. 3

Summed motion per minute of mice on (A) the first FST trial and (B) the second FST trial. Motion was computer-scored. All error bars are SE. One-way ANOVA testing for effect of genotype was computed for each minute; * denotes p<0.05, ** denotes p<0.01, and *** denotes p<0.001.

Fig. 4

Group comparisons. Day 1 vs. Day 2 differences for each strain are graphed using the following measures: (A) visual-scored mobility, (B) computer-scored total motion, and (C) computer-scored mobility. Time mobile is plotted in (A) and (C) instead of time immobile to make it visually easier to compare with the total-motion graph; time mobile is equal to 360 seconds minus time immobile. Error bars represent standard deviations.