TL;DR
This paper introduces a novel, quantitative method for assessing upper limb motor function in stroke patients, combining kinematic and sEMG data. The proposed five features show significant differences across stroke severity levels and correlate well with the established Fugl-Meyer assessment scale, offering a more nuanced evaluation for personalized rehabilitation.
What Problem Does This Paper Address?
Stroke is a leading cause of disability, often resulting in significant upper limb motor dysfunction. While clinical scales like the Fugl-Meyer (FM) assessment are widely used to evaluate recovery, they rely on subjective scoring and can miss subtle functional changes. This limits the ability to tailor rehabilitation strategies precisely to individual patient needs. The paper aims to bridge this gap by developing a quantitative, objective method that captures more detailed aspects of motor control, including movement quality, efficiency, and underlying muscle activation patterns.
Methods
The study proposes a set of five evaluation features derived from a combination of kinematic and surface electromyography (sEMG) data collected during a voluntary upward reaching task.
- Kinematic Data: Inertial Measurement Units (IMUs) attached to the waist, upper arm, forearm, and hand capture motion at 50 Hz. From this, the researchers extract:
- Max Shoulder Joint Angle: Measures the range of motion of the shoulder.
- Upper Arm Peak Speed: Quantifies the maximum instantaneous speed during the reaching movement.
- Upper Arm Average Speed: Reflects the overall efficiency of the reaching movement.
- Torso Balance Degree: Assesses compensatory torso movements, calculated from Euler angles derived from torso IMU data.
- sEMG Data: sEMG signals from seven upper limb muscles are recorded at 1000 Hz.
- Muscle Synergy Similarity: Non-negative Matrix Factorization (NMF) is used to extract muscle synergies, which represent modular muscle activation patterns. The similarity of these patterns to those of healthy individuals is then quantified.
The sEMG data undergoes standard preprocessing (filtering, rectification, integration) and normalization. All features are then normalized using a min-max approach for comparison.
Datasets
The study recruited:
- 34 stroke survivors (diagnosed for the first time, within 6 months post-stroke, able to lift shoulder ≥ 30°, no other musculoskeletal disease).
- 25 age-matched healthy volunteers as a control group.
Stroke patients were further classified into "severe impaired" (FM ≤ 30) and "mild-to-moderate impaired" (FM > 30). The Fugl-Meyer (FM) assessment score for the upper limb was obtained from a professional rehabilitation therapist for each patient. No new benchmark datasets were introduced; existing clinical scales and standard data acquisition methods were used.
Key Insights and Ideas
- Beyond the FM Score: The FM scale, while useful, has limitations in capturing subtle motor deficits, especially those related to dyskinesia or abnormal muscle patterns. The proposed features offer a more granular view.
- Muscle Synergies as Biomarkers: Alterations in muscle synergy patterns in stroke patients, particularly the co-activation of shoulder muscles for compensation and reduced activation of stabilizing muscles (like TRI), provide insights into underlying neural control strategies and deficits.
- Torso Compensation is Key: Increased torso balance degree in stroke patients (especially severe cases) highlights the reliance on compensatory strategies when arm movement is impaired.
- Integrated Approach is Powerful: Combining kinematic data (for movement execution) with sEMG data (for muscle activation) provides a more comprehensive understanding of motor control than either modality alone.
Key Results
- Significant Group Differences: All five proposed features showed statistically significant differences (p ≤ 0.001) between the severe, mild-to-moderate, and control groups.
- Correlation with Clinical Severity: The features demonstrated significant correlations (p ≤ 0.001) with the Fugl-Meyer (FM) assessment scores.
- Max shoulder angle, peak speed, average speed, and synergy similarity increased with improved motor function (higher FM scores).
- Torso balance degree decreased with improved motor function.
- High Consistency with FM Scale: Bland-Altman plots showed good agreement between the proposed evaluation features and the clinical FM scale, indicating the reliability of the new method.
- Specific Findings:
- Severe stroke patients exhibited significantly lower shoulder joint angles, peak speeds, and average speeds compared to mild-to-moderate and control groups.
- Muscle synergy similarity was lower in stroke patients, suggesting altered motor control strategies.
- Torso balance degree was higher in stroke patients, indicating greater compensatory trunk movement.
Limitations and Future Work
- Poststroke Duration: The study recruited patients within 6 months post-stroke to minimize the effect of rehabilitation. Future work should investigate how these features change over longer durations and with different rehabilitation interventions.
- Rehabilitation Treatment Effects: The influence of various rehabilitation therapies (e.g., robot-assisted training) on kinematic parameters and clinical scales needs further exploration.
- Automatic Grading: The authors suggest establishing a comprehensive evaluation model based on these features for automatic patient grading, which would require developing classification algorithms and expanding the dataset.
- Broader Application: While this study focused on reaching, future work could explore the application of these features to other upper limb movements and potentially other body parts.
Summary updated