Fig. 3

Validation of the simplified Microwave-based Motion Detector System. a (left panel) Setup for assessing the precision of the motion detector. A small piece (2 x 2 cm) of aluminum foil is mounted on the large clock-hand of an analog clock. The motion detector is placed 30 cm away from the analog clock and was allowed to capture the movements of the clock-hand for 30 min. (right panel Basic summary of the clock experiment; total number of detected events, mean value of event intervals and the respective percentiles are given. b Intervals of detected events over time. Note the occurrence of long intervals (>2 s) which indicate slight detection problems due to blind spots in the recording setup. c Histogram (bin = 100 ms) of all intervals demonstrating, that 85.38% of the detected events are in the range of 950 ms to 1050 ms. d A magnified view of the range of 990 ms to 1010 ms shows the normally distributed nature of the recorded data. e The auto-correlation of the recorded interval data demonstrates a prominent rhythmicity with a frequency of 1 min, which is most likely due to the additional weight on the clock-hand in combination with the slackness of the low quality gear used in the ordinary analog clock, which causes the aluminum foil to vibrate (see inset). f In oder to asses the crosstalk between simultaneously recorded, neighboring cages and the effect of shielding we have conducted another experiment (<18 lux), where one CD1 mouse was introduced to cage 2 while cages 1 & 3 where unpopulated. Further we have placed an A4-sized sheet of aluminum foil (floating, not connected to GND) between cage 2 & 3. The experiment was conducted for 30 min and in addition the behavior of the animals was video-taped. The green rectangles at the cage front show the placement of the motion detector modules. While in the unshielded cage, the detector picked up 26% of the neighboring cage, the detector of cage 3 did not detect a single event. g Performance comparison of the 30 min behavior (30 s bins) in cage 2 between three different locomotion detection approaches: a) an experienced observer (OBS) manually scored either the occurrence grooming (grey bars) behavior or locomotor (red bars) activity (ambulatory activity, digging and rearing); b) locomotor activity measured by the motion detector shield (MDS) (green bars); c) frame-by-frame pixel difference (PD) as an unbiased measure of movement in the video file. The pixel noise was found to generate 5.5% differences between the frames and we used a rather liberal threshold of 8.5% to determined locomotor activity. MDS and PD datasets were initially binned at 2 s bins in a binary manner (motion = 1, no motion = 0). OBS data was also binned initially at 2 s but the data was already given in percentages due to the two different variables. Further, all data set where binned to 30 s bins and the percent presence locomotion/grooming determined. h Averaged overall locomotor activity for OBS, MDS, PD; numbers indicate the difference to OBS. i Pearson correlation analysis of MDS vs. OBS, PD vs. OBS and MDS vs. PD