Validation of vertical activity and rearing measurements by temporal correlation.

<p>(a) Broken column charts showing the automated detection of vertical activity and manual analyses of vertical activity and rearing, in 1 h samples of light and dark phase video footage, arranged in sequential 15-min bins (videos 1–4). Gray columns indicate the time spent in vertical activity by automated detection, red columns indicate the time spent in vertical activity by manual annotation, blue columns indicate the time spent rearing by manual annotation. The frame-by-frame accuracy against the gray column of each bin was determined. (b) Correlation between time spent in vertical activity detected automatically versus manual annotation of vertical activity; R² = 0.9236; ICC = 0.97. (c) A Bland-Altman plot, a visual tool to compare two techniques by plotting the difference against the signal to assess for bias in any one technique over the dynamic range assessed, was used to compare a manual measure of vertical activity and an automated measure of vertical activity. (d) Correlation between time spent in vertical activity detected automatically versus manual annotation of rearing; R² = 0.4774; ICC = 0.91. Two notable outliers are highlighted; red-encircled point: probable cause is rat on top of play tunnel but not actually rearing; blue-encircled point: rat rearing at rear of cage but nose just beneath horizontal line cut-off (illustrated in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0181068#pone.0181068.g005" target="_blank">Fig 5</a>). (e) Bland-Altman plot comparing vertical measure between manual assessment of rearing and an automated assess of vertical activity.