Smart: Tracking single subject position

Tracking position is probably the most important feature offered by the video analysis systems. Even if only this features where available, tracking position relieves researchers of an extremely time-consuming task.

Figure 8Figure 8. Screen of SMART when tracking a subject in the water maze.

SMART offers several different approaches to determinate the position of the experimental subject in the working scenery. Two of them are similar to those offered by other systems (such as Videomex-V from Columbus) and are based in a high contrast between the experimental subject and the background. Supposed this contrast (either black on white or viceversa), is an easy task to transform the image into a binary matrix (with 256 x 192 elements in the case of Videomex-V and ONE and with 512 x 512 in our case), diminishing dramatically the amount of information to be processed.

In the binary matrix, the majority of elements are black (or white) and correspond to the background (that needs to be uniform), while only a portion of them, white (or black), correspond to the experimental subject. It is an easy task to determinate the gravity centre of the spot corresponding to the subject and obtain their co-ordinates.

This procedure is efficient but stringent respect of the experimental situation as the high contrast between the experimental subject and the background can not always be achieved.

Additionally, SMART offers a third processing mode, in which this high contrast becomes unnecessary, so widening the scope of application of the procedure.

What SMART do is to establish a threshold stating the minimum difference (in luminance, 0-255) between the experimental subject and the scenery (understanding this later as a picture of the experimental set-up without the animal). Once the animal is present, a subtraction is done between the scenery (initial image) and the current image, and what remains is only the newly added object (in fact, the experimental subject, see figure 9, left). Thus, what SMART "see" is a spot, showing the experimental subject; their position co-ordinate is calculated determining its gravity centre. In this way, position co-ordinates can be determined even if the contrast between the subject and the background is very low, as it is shown in figure 8, in which a white animal is tracked while swimming in white coloured water.


	Figure 9a Figure 9. Aspect of a mouse as result after the subtraction of the scenery (left) and after two levels of erosion (middle and right); only a portion of the real screen is shown; images have been converted into inverted binary matrixes.
	Other important feature included in this procedure is the capability to erode, in an user selected degree, the processed image. The erosion is a process by mean of which successive external layers of pixels are eliminated. The result of this process is the progressive overshadowing of the image (as shown in figure 9, middle and right), depending on the amount of layers eliminated (2 and 10 in the middle and right images respectively in figure x). Additionally, this allows for the elimination of artefacts disturbing tracking, such as random variation of luminance above the threshold level or faeces. As shown in figure 9 (see left), shape of the spot grossly corresponds to the form of the contour of the animal and covers an area that must be known (SMART include tools for entering this value). When this area diminishes below the 40% of their initial value, a rearing is understood to be occurring, lasting until area recovers the 60% of the initial value. The time of beginning/end of the rearing is included in the track data file and it can be seen when doing track analysis.

Panlab s.l.
^{c/ Loreto 50
Phone: 34-934190709 Fax: 34-934197145
08029 Barcelona
Spain
E~mail:}^{info@panlab-sl.com}