Figure 1.
Four monitors are screwed into a PVC box. Top and bottom mirrors create the illusion of infinite profundity and hinder the animal from leaving the platform. A camera above the platform is used to monitor the inside of the arena. The stimulus parameters, the video recording of the behaving animal and the head-tracking can be observed by the experimenter on a fifth monitor.
Figure 2.
The program consists of three components. The component patternGen (purple) generates arbitrary stimuli, consisting of a texture (bitmap file) and a rotation protocol (text file). It also adjusts the background illumination value to gain a specific absolute intensity set by the user in the GUI. The omr monitor (light blue) implements the communication with the camera, the head tracking and the user interface during an experiment. Images are acquired through the MATLAB image acquisition toolbox and are used by the tracking module to determine the animal's head location. The live view is displayed in an on-screen GUI (arenaCtrl) in which the user can also adjust tracking parameters. The retrieved head position is used to recalculate the location of the stimulus pattern. The updated translation and rotation parameters are passed to the third component, omr arena (red), which presents the cylinder at the correct position on the four screens through SDL/OpenGL.
Figure 3.
The graphical user interface of the patternGen component, including illustrations of (A) the stimulus pattern, (B) the rotation protocol, (C) the translation protocol (not used in OMR experiments). The user sets parameter values of (D) the desired spatial frequency in cyc/, (E) the range of the pattern displayed in the preview, (F) activation of linear color space (color values are mapped according to display calibration), (G) brightness and contrast of the stimulation displays, (H) a filename prefix for the experiment (optional), (I) color and (J) contrast of the pattern, (N) selection of linear or sinusoidal stimulus rotation, (K) parameters for the selected type of rotation (for linear movement: velocity, frequency and amplitude). Moreover, the user can load (M) a preset, consisting of previously saved settings, or (L) select a different experiment type (for future paradigms other than OMR/OKR).
Figure 4.
During experiments, the gaze of the freely moving animal (red arrow) is tracked continuously (A). Algorithmic steps to determine the gaze (B): Segmentation is based on a color-threshold, which is automatically determined by analyzing the pixels in the platform region. To facilitate the detection of the nose position, coordinates are weighted with a function ranging from 0 at the border of the region around the platform that the mouse can reach with its nose to 1 at the center of the platform. The center of gravity (black X) is calculated based on weighted pixels (see color bar). It is usually located in the animal's hind-quarters region. The nose (white X in circle) is usually detected as the pixel farthest from the center of gravity. The position between the ears (white X) is determined as a center of mass in a circular region (white circle) around the location of the nose. The head-gaze (red arrow in A) is calculated as the vector from the position between the ears to the position of the nose. See supporting material S1 for an example video visualizing head tracking.
Figure 5.
Effect of the variation of the discrimination criterion.
The response curves were calculated on basis of the individual normalized responses of the 6 animals (as in figure 8) for different values of the discrimination criterion (values from 2 to 12
/s, shown to the right of each curve). The fraction of time tracked (
) increases when softening the criterion, but the shape of the response curve as well as the estimated visual acuity based on logistic fits of the median responses (shown in brackets on the right) remains qualitatively similar in a certain range around the standard value of
= 9
/s.
Figure 6.
The amount of tracking movements counted by the human observer (dashed gray line, not normalized, right axes) compared to the automatically determined response curve when using a discrimination criterion of = 9
/s (blue line, not normalized, left axes) for the six individual animals. All values are unnormalized medians over 10 stimulus presentations. The errorbars represent the variation of individual responses by showing the upper (75%) and lower (25%) quartiles of responses to each spatial frequency. The solid red line is the median for the measurements at null condition (not moving stimulus), the red dashed red lines represent the quartiles at this condition. The calculated individual spatial acuity thresholds are shown in the upper right corner of each graph.
Figure 7.
Determination of spatial frequency threshold.
The descending slope of the response curves obtained by the human observer (gray dots, A) and of the automated analysis (blue dots, B) was fitted by a logistic function (gray and blue lines). Curve fitting was done based on the median values obtained for the six animals and each spatial frequency. Dashed lines indicate the prediction bounds, in which 95% of new measurements are expected to lie. The visual threshold was defined as the spatial frequency value at the inflection point (50% of the maximum response), resulting in 0.41 cyc/ for the human observer approach and 0.39 cyc/
for the automated analysis.
Figure 8.
Automated and human observer analysis of OMR performance.
The normalized response curves determined by the human observer (dashed gray line) and by the automated analysis (blue line, discrimination criterion = 9
/s) are similar. For each spatial frequency, the median of response of all mice was calculated based on the median values for each individual mouse. Responses are normalized by setting the median value obtained for the optimal spatial frequency to one. For the automated analysis, chance level was calculated as the median of the number of head movements matching the discrimination criterion by chance at null condition during absence of movement. It was set to zero (red line) by subtracting this chance value from all values obtained by the automated analysis. Errorbars and dotted red lines represent the variation between animals by showing the smallest and highest medians obtained for the six animals during stimulation and null condition. (See Methods for details).
Figure 9.
A: Example course of the recorded head-angle over time of one animal at the optimal spatial frequency of 0.2 cyc/ (A left) and a spatial frequency of 0.6 cyc/
(A right), at which the human observer does not detect OMR behavior. Samples where the head angular velocity deviates less than
= 9
/s from the stimulus velocity of 12
/s are highlighted in green. Significantly more frames fulfill this criterion and are hence automatically detected as stimulus tracking behavior at 0.2 cyc/
than at 0.6 cyc/
. B: Movement of the grating over time. C: Histogram of head velocities of all animals induced by the gratings with spatial frequencies of 0.2 cyc/
(left) and 0.6 cyc/
(right), both moving with 12
/s. Positive numbers correspond to head-movements in the same direction as the stimulus moved, negative numbers to head-movements to the opposite direction. The green area indicates the region of head velocities that was tolerated as tracking behavior. To visualize the more strongly skewed distribution found for responses to 0.2 cyc/
in contrast to 0.6 cyc/
, head-movements that occurred in the range of tolerated speeds but in the wrong direction are highlighted by the red area.