Hand movement prediction enables the creation of personalized rehabilitation programs that align with individual neural patterns. It is particularly beneficial for patients recovering from strokes or other motor impairments, allowing therapists to adjust treatments in real time based on brain activity feedback, leading to more effective patient care. To this end, this study aims at predicting hand movements using functional near-infrared spectroscopy (fNIRS) signals and artificial intelligence (AI) algorithms. We employed fNIRS and collect cortical activity associated with hand movements and utilized machine learning (ML) algorithms to predict the hand movement. For the prediction task, various Machine Learning algorithms, including Random Forest (RF), Decision Tree (DT), k-Nearest Neighbors (kNN), Logistic Regression (LR), Support Vector Machines with polynomial (SVM Poly) and radial basis function (SVM RBF) kernels, Gaussian Naive Bayes (GNB), Neural Networks (NN), Gradient Boosting (GB), and AdaBoost, were used. The performance of the RF algorithm was notably superior, demonstrating high accuracy in predicting hand movements. In contrast, LR, SVM (Poly), SVM (RBF), GNB, and NN algorithms exhibited poor performance, while kNN, DT, GB, and AdaBoost algorithms showed moderate effectiveness. In addition, the Receiver Operating Characteristic (ROC) curves indicated that the RF algorithm had significantly larger areas under the curve (AUC) compared to other algorithms. The integration of fNIRS with AI improves hand movement prediction, suggesting advancements in rehabilitation and personalized medicine. This study emphasizes the potential of these algorithms. More research is required to expand their applications across various clinical settings.
Jafari et al. (Fri,) studied this question.