Fig 1.
Muscle sites for Delsys Trigno EMG sensor attachment.
The muscle sites included deltoid, triceps, biceps, Extensor Carpi Ulnaris (ECU), Flexor Carpi Ulnaris (FCU), Thenar Eminence (TE) for the left and right side.
Fig 2.
Knot tying task performed by participants.
Fig 3.
Pegboard transfer task performed by participants.
Fig 4.
(a) position of six muscles sites where surface EMG sensors were affixed (b)Pegboard transfer task (c) Robotic suturing task using Da Vinci (d) Knot tying task (e) Ureteroscopy task.
Table 1.
Temporal and frequency domain features extracted from EMG signals.
Table 2.
Linear and nonlinear features extracted from accelerometer data.
Fig 5.
Flow chart of feature selection-based model development and identification of parameters of muscle importance for skill classification.
Fig 6.
Linear variability as measured by RMS of EMG signals between groups of 3 skill levels.
Fig 7.
The dominant frequency of EMG signals during the performance of 3 surgical tasks (knot tying, pegboard transfer, and robotic suturing).
Fig 8.
The cumulative muscular workload for completing surgical tasks (knot tying, pegboard transfer, and robotic suturing) by three groups of different skill levels (expert, intermediate, and novice).
Fig 9.
Average muscular work done per second and significant differences in the robotic suturing task.
Fig 10.
High complexity measured by multiscale entropy.
Fig 11.
A correlation matrix shows correlation coefficients between features.
Table 3.
Feature importance as determined by recursive feature elimination method.
Table 4.
Performance metrics were evaluated by the three ML models (random forest classifier, SVM, and Naïve Bayes) with six muscle datasets.
Highest accuracies are heighted in the table as bold.
Table 5.
This table shows the performance of a combination of muscles for surgical skill evaluation for each model separately.
Accuracy, precision, recall, and F1 score are reported for all 3 classification models. Highest accuracies are highlighted in the table as bold.
Fig 12.
ROC curves showing performance of two muscles (ECU and Biceps) for skill level classification among a) novice, b) intermediate, and c) expert using three classifiers random forest (blue line), Naïve Bayes (orange line), SVM (green line).
Fig 13.
ROC curves showing performance of two muscles (ECU and Deltoid) for skill level classification among a) novice, b) intermediate, and c) expert using three classifiers random forest (blue line), Naïve Bayes (orange line), SVM (green line).
Fig 14.
ROC curves showing performance of two muscles (ECU, Biceps and Deltoid) for skill level classification among a) novice, b) intermediate, and c) expert using three classifiers random forest (blue line), Naïve Bayes (orange line), SVM (green line).
Fig 15.
Total time taken for completing surgical tasks (knot tying, pegboard transfer and robotic suturing) by three groups of different skill level (expert, intermediate and novice).