Related Experiment Video
Updated: Jun 4, 2026
14:56
The Combined Use of Transcranial Direct Current Stimulation and Robotic Therapy for the Upper Limb
Published on: September 23, 2018
Rehab-DRLX: explainable neurorehabilitation prognosis using deep reinforcement learning and transformer-based models
Hadeel Alsolai1, Shakir Khan2, Rakesh Kumar Mahendran3
1Department of Information Systems, College of Computer and Information Sciences, Princess Nourah Bint Abdulrahman University, Riyadh, Saudi Arabia.
Frontiers in Computational Neuroscience
|June 3, 2026
Summary
Rehab-DRLX, a hybrid deep learning model, enhances neurorehabilitation by combining deep reinforcement learning and explainable AI for accurate patient recovery prognosis. This approach improves clinical decision-making and builds trust through interpretable insights.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Clinical Informatics
Background:
- Neurorehabilitation is critical for patients with motor impairments and neurological conditions affecting daily activities.
- Current prognostic tools often lack interpretability and accuracy, hindering effective clinical decision-making.
Purpose of the Study:
- To introduce Rehab-DRLX, a novel hybrid deep learning model for interpretable and accurate neurorehabilitation prognosis.
- To process multimodal patient data, including clinical records, motion data, and neuroimaging, for dynamic recovery pattern analysis.
Main Methods:
- Utilized a hybrid deep learning framework combining deep reinforcement learning (DRL) with an explainable transformer model.
- Integrated a reinforced representation learning (RRL) module with a convolutional neural network (CNN) for spatiotemporal feature encoding.
- Employed an explainable prognosis transformer (XPT) with hierarchical attention for transparent decision-making.
Main Results:
- Achieved high accuracy (94.6%), F1-score (0.93), and low RMSE (0.082) and MAE (0.061).
- Demonstrated significant improvements over existing prognostic methods.
- Ablation studies confirmed the essential contribution of each architectural component.
Conclusions:
- Rehab-DRLX offers a practical and viable solution for neurorehabilitation prognosis.
- The model provides both accuracy and interpretability, enhancing clinical trust and decision support.
- This approach addresses the limitations of traditional 'black box' prognostic tools.
