Fricke, Christopher
[Author];
Alizadeh, Jalal
[Author];
Zakhary, Nahrin
[Author];
Woost, Timo B.
[Author];
Bogdan, Martin
[Author];
Classen, Joseph
[Author]
Evaluation of Three Machine Learning Algorithms for the Automatic Classification of EMG Patterns in Gait Disorders
You can manage bookmarks using lists, please log in to your user account for this.
Media type:
E-Article
Title:
Evaluation of Three Machine Learning Algorithms for the Automatic Classification of EMG Patterns in Gait Disorders
Contributor:
Fricke, Christopher
[Author];
Alizadeh, Jalal
[Author];
Zakhary, Nahrin
[Author];
Woost, Timo B.
[Author];
Bogdan, Martin
[Author];
Classen, Joseph
[Author]
imprint:
Lausanne: Frontiers Research Foundation, [2023]
Description:
Gait disorders are common in neurodegenerative diseases and distinguishing betweenseemingly similar kinematic patterns associated with different pathological entities is achallenge even for the experienced clinician. Ultimately, muscle activity underlies thegeneration of kinematic patterns. Therefore, one possible way to address this problemmay be to differentiate gait disorders by analyzing intrinsic features of muscle activationspatterns. Here, we examined whether it is possible to differentiate electromyography(EMG) gait patterns of healthy subjects and patients with different gait disorders usingmachine learning techniques. Nineteen healthy volunteers (9 male, 10 female, age 28.2± 6.2 years) and 18 patients with gait disorders (10 male, 8 female, age 66.2 ± 14.7years) resulting from different neurological diseases walked down a hallway 10 times ata convenient pace while their muscle activity was recorded via surface EMG electrodesattached to 5 muscles of each leg (10 channels in total). Gait disorders were classifiedas predominantly hypokinetic (n = 12) or ataxic (n = 6) gait by two experienced ratersbased on video recordings. Three different classification methods (Convolutional NeuralNetwork—CNN, Support Vector Machine—SVM, K-Nearest Neighbors—KNN) wereused to automatically classify EMG patterns according to the underlying gait disorderand differentiate patients and healthy participants. Using a leave-one-out approach fortraining and evaluating the classifiers, the automatic classification of normal and abnormalEMG patterns during gait (2 classes: “healthy” and “patient”) was possible with a highdegree of accuracy using CNN (accuracy 91.9%), but not SVM (accuracy 67.6%) orKNN (accuracy 48.7%). For classification of hypokinetic vs. ataxic vs. normal gait (3classes) best results were again obtained for CNN (accuracy 83.8%) while SVM andKNN performed worse (accuracy SVM 51.4%, KNN 32.4%). These results suggest thatmachine learning methods are useful for distinguishing individuals with gait disordersfrom healthy controls and may help classification with respect to the underlying disordereven when classifiers are trained on comparably small cohorts. In our study, CNNachieved higher accuracy than SVM and KNN and may constitute a promising methodfor further investigation.