Fig 1.
The EMG sensor and vibration motor were attached to the tibialis anterior of the left and right leg, respectively. Each vibration motor was controlled by the Arduino board.
Fig 2.
The EMG signals were acquired for 2 min before and after the DVT surgery. The experimental period was a total of 188 min.
Fig 3.
The 25 PCs were extracted and reshaped into the 5 × 5 matrix with the most significant PC positioned at the center. Then, 64 filters of 4 × 4 and 2 × 2 sizes convolved with the PCs and fully connected for classification.
Fig 4.
Intracompartment pressure (mmHg), perfusion pressure (mmHg), and shear elastic modulus (kPa).
The dot indicates each data point for each group. Each bar plot presents 25th (Q1), 50th (Q2), and 75th (Q3) percentiles of the data. The whisker presents Q3 + 1.5 * (Q3 − Q1), and Q1 − 1.5 * (Q3 − Q1).
Fig 5.
The results of the LOLO-CV for binary classification (t−1 –t1) in the left (L) and right (R) legs of all four pigs.
Fig 6.
The results of the cross-validation within multiclass classification at all-time points (t−1 –t1).
Fig 7.
Confusion matrices in each case for multiclass classification.
The y-axis indicates the predicted labels and the x-axis indicates the true labels from the CNN classifier in each case.
Fig 8.
The representative fast Fourier transform results of acquired EMG signals (the right leg for Pig #4).
The peak after DVT surgery was higher than before DVT surgery.
Fig 9.
The results of the cross-validation within binary classification for 1 hour and 2 hours after the DVT surgery (t1—t4 and t5—t8).
Fig 10.
Significant correlation between IP and PP, IP and SEM, PP and SEM, and PP and EMG feature std.