Outputs of the four filters in the proposed half-heart mask generation step: (a) variation level of brightness intensity (brighter pixels have higher variation); (b) highly varied region; (c) heart mask; (d) half-heart mask. The outputs require the information of all frames of a video and are generated only once for each video

A zebrafish video with larva sliding problem (ID: MVI_5269). The timestamps of (a) to (d) are 0, 7, 13 and 20 s. The top area shows the full video frame and the bottom area zooms into the red rectangle. The bottom area shows that the larva in the video slides upward. Namely, the larva slides towards the top right corner of the full video frame. In a video frame, the larva is darker than the background. If a larva slides into the screen when recording, the extracted h-signal has an overall decreasing trend; conversely, if a larva slides out the screen like this sample video, the extracted h-signal has an overall increasing trend as shown in Fig. 1b