Article Figures & Data
Figures
- Figure 1.
Experimental setup and real-time tracking system for providing continuous auditory feedback during a reaching task in mice. A, Schematic diagram of the closed-loop system integrating real-time pose estimation and auditory feedback. A high-speed camera captures the mouse's forepaw movements, which are processed by the deep neural network to track paw digits. The vertical displacement of the left forepaw is encoded into auditory tones delivered through a speaker, providing immediate feedback to the mouse. B, Left panel, An example image showing real-time tracking of the forepaw digits. Right panel, Time series plots of the auditory tone frequency (green), left paw vertical displacement (red), and right paw vertical displacement (black). The auditory tone frequency corresponds closely with the left paw movement, illustrating the system's real-time feedback capability.
- Figure 2.
Clustering analysis of reaching trajectories during motor training. A, Left panel, Diagram of the reaching task setup, where the mouse must move its left forepaw from the initial position to the target region (red square) to perform a successful reach. Right panel, Illustration of the training schedule over 4 d. B, Reaching trajectories of the left forepaw for a representative mouse in the auditory feedback group over the four training days. Each plot overlays multiple reaching movements. C, Principal component analysis (PCA) of reaching trajectories. D, t-distributed stochastic neighbor embedding (t-SNE) plots of reaching trajectories. E, Clustering of reaching trajectories using Gaussian mixture model (GMM).
- Figure 3.
Continuous movement-coded auditory feedback promotes motor skill learning. A, Schematic of the experimental setup comparing the auditory feedback group and the control group. Mice in the auditory feedback group received real-time auditory tones corresponding to their left forepaw movements during the reaching task. Both groups performed the same reaching task under the same low-intensity green LED illumination, which served solely as a contextual cue indicating active training. B, Example plots of left forepaw vertical displacement during baseline (gray) and training (red) for a mouse in the auditory feedback group. C, Left, Image of a mouse during the motor training session (video and audio are recorded by a second camera). Right panel, Plot of forepaw movements over time during a training session. D, Mel spectrogram of the real-time audio recorded by the second camera. E, Scatter plot showing the correlation between audio frequency and left paw movement intensity (Pearson's r = 0.94; slope = 0.75 [0.74, 0.77], two-tailed t test p < 0.001). The red line represents the best-fit linear regression. F, Comparison of motor learning between the auditory feedback group (n = 7) and the control group (n = 7) over 4 d of training. Violin plots represent the ratio of left paw reaches to right paw reaches, normalized to Day 1. The auditory feedback group demonstrated significantly higher reach ratios on Days 2, 3, and 4 (asterisks denoting p < 0.05). Unpaired two-sample t tests for each day revealed: Day 2 Control versus Audio (p = 0.0428), Day 3 Control versus Audio (p = 0.0289), and Day 4 Control versus Audio (p = 0.0144). A two-way repeated-measures ANOVA (day × group) revealed significant main effects of group (p < 0.05), as well as a day × group interaction (p < 0.01), indicating that continuous auditory feedback alters the progression of motor skill learning over time.
In this issue
