Sex-specific effects of appetite suppressants on stereotypy in rats
Fig 2
Automatic quantification of behavioral states induced by appetite suppressants.
A) Experimental protocol for video recording rat’s behavior. For the data acquisition, rats were filmed from top and bottom views. The first 15 minutes were used as a baseline, followed by the injection of an appetite suppressant at 15 minutes. The recording continues until completing four hours. Images were converted into videos, and a machine learning algorithm, ResNet-50, was used for training these key points: the nose, front paws, hind paws, and tail base. Subsequently, the coordinates of these key points were extracted. After correcting errors, behaviors such as locomotion, quiet-awake state, rearing, grooming, and stereotypy were defined based on these coordinates and quantitatively analyzed across all frames. B) Ethograms were built of the following behavioral states: 1) Head weaving, 2) Quiet sleep state, 3) Locomotion, 4) Grooming, and 5) Rearing. C) Pose estimation images extracted from DLC show behaviors in the same order as in panel (B). D) Representative ethogram of male rats. Left panel: saline-injected rat; right panel: rat injected with Phentermine at 20 mg/kg. Behavioral states are color-coded as follows: 1) Stereotypy (black), 2) Quiet-awake state (red), 3) Locomotion (blue), 4) Grooming (cyan), and 5) Rearing (magenta). In S1–S8 Figs, the complete ethograms for all rats and drugs can be seen, and the entire database can be found in [87].