Analysis of Pelvis and Lower Limb Coordination in Stroke Patients Using Smartphone-Based Motion Capture Technology

Academic Background

Stroke is one of the diseases with the highest incidence, disability, and mortality rates worldwide, with up to 15 million new cases each year. Among them, 20%-30% of stroke patients develop hemiplegic gait, which is one of the most severe functional impairments after a stroke. Hemiplegic gait not only affects the walking ability of patients but also increases the risk of falls, significantly impairing their quality of life. Traditional gait analysis often focuses on spatiotemporal parameters and joint kinematics of the lower limbs, but these discrete parameters often fail to fully reveal the mechanisms behind abnormal gait patterns in stroke patients. The coordination between the pelvis and lower limbs plays a crucial role in maintaining gait stability, balance, and propulsion; however, research in this area remains limited.

Currently, assessments of pelvis-lower limb coordination typically rely on high-cost reflective marker-based motion capture systems, limiting their widespread clinical application. With technological advancements, markerless motion capture techniques based on smartphone videos provide a convenient, low-cost, and portable alternative for gait analysis. This study aims to use smartphone video capture technology, combined with Statistical Parametric Mapping (SPM) and Continuous Relative Phase (CRP) analysis methods, to thoroughly explore the coordination and kinematic characteristics of the pelvis and lower limbs in stroke patients, offering new assessment tools and intervention strategies for clinical rehabilitation.

Paper Source

This study was jointly conducted by Yinghu Peng, Wei Wang, Yangkang Zeng, Zhenxian Chen, Hai Li, and Guanglin Li from the Shenzhen Institutes of Advanced Technology (SIAT), Chinese Academy of Sciences (CAS), the Neurorehabilitation Laboratory, Department of Rehabilitation Medicine, Shenzhen Hospital, Southern Medical University, and the School of Mechanical Engineering, Chang’an University. The paper was published in the IEEE Transactions on Biomedical Engineering journal in 2025.

Research Workflow

1. Participant Information

The study recruited 17 hospitalized stroke patients (average age 49.9 years) and 20 age-matched healthy controls (average age 49.4 years). Inclusion criteria included: aged 18-65 years, non-obese (BMI < 30 kg/m²), unilateral hemiparesis (hemorrhagic or ischemic), ability to walk independently for 10 meters, and Brunnstrom scores ≥4. Exclusion criteria included: inability to understand instructions, presence of other mental illnesses or systemic neurological conditions, and severe organ dysfunction.

2. Equipment and Experimental Procedures

The study used two iPhones (iPhone 11 and 12 mini) to collect motion data. The experiment was divided into the following steps:
1) Camera Calibration: A checkerboard was used to calibrate the camera position, and video data were processed using the OpenPose algorithm.
2) Static Trial: Static data from participants were collected to create customized musculoskeletal models.
3) Dynamic Trial: The OpenCap web application triangulated synchronized 2D video keypoints and estimated the 3D positions of 43 anatomical markers using an LSTM network trained on public motion capture datasets.

3. Data Processing and Joint Kinematics Calculation

The open-source software OpenSim was used to estimate joint kinematics. Data processing included:
1) Custom Scaling: A customized musculoskeletal model was created based on the participant’s height and static trial data.
2) Inverse Kinematics: Marker trajectories from walking trials were used to calculate pelvis and lower limb joint angles, normalizing the gait cycle to 100%.

4. Intersegment Coordination Analysis

The Continuous Relative Phase (CRP) analysis method was employed to assess the coordination between the pelvis and lower limbs. The study analyzed CRP for five segment pairs: pelvis-thigh (sagittal, coronal, and transverse planes), thigh-shank (sagittal plane), and shank-foot (sagittal plane). CRP calculations were based on Hilbert transforms and phase angle differences, with time-series analysis performed using Statistical Parametric Mapping (SPM).

Key Results

1. Intersegment Coordination

The study found significant differences in CRP curves of the pelvis-thigh in the coronal plane throughout the entire gait cycle (p < 0.001). The CRP of the thigh-shank in the control group during the swing phase (65%-91%) was significantly higher than that in both paretic and nonparetic limbs. Significant differences in CRP of the shank-foot were observed during mid-stance (13%-29%) and mid-swing (57%-68%).

2. Pelvic Kinematics

The range of motion (ROM) of pelvic tilt in the control group was significantly greater than in stroke patients (p < 0.001). During early and mid-stance phases (2%-35%) of the gait cycle, pelvic tilt values in the control group were significantly lower than those in stroke patients, while they were significantly higher during late stance and early swing phases (47%-71%).

3. Lower Limb Joint Kinematics

The ROM of hip flexion-extension in the paretic limb was significantly lower than in the control and nonparetic limbs (p = 0.001). The maximum knee flexion angle in the paretic limb was significantly lower than in both the control and nonparetic limbs (p < 0.001). Ankle dorsiflexion angles were significantly lower in the paretic limb during mid-stance compared to the control and nonparetic limbs.

Study Conclusions

This study comprehensively evaluated the coordination and kinematic characteristics of the pelvis and lower limbs in stroke patients using smartphone video capture technology combined with SPM and CRP analysis methods. The study found significant differences in pelvis-lower limb coordination between the paretic and nonparetic limbs, which may be related to muscle dysfunction such as spasticity and weakness. The findings provide new assessment tools for clinical rehabilitation and offer scientific evidence for targeted interventions (e.g., reducing spasticity and strengthening muscles).

Highlights of the Study

1. Technological Innovation: For the first time, smartphone video capture technology was applied to evaluate pelvis-lower limb coordination in stroke patients, providing a convenient and cost-effective gait analysis solution for clinical settings.
   
2. Methodological Innovation: By combining SPM and CRP analysis methods, the study achieved comprehensive evaluation of time-series data across the gait cycle, revealing deeper mechanisms underlying abnormal gait patterns in stroke patients.
   
3. Clinical Value: The results provide new assessment metrics and intervention strategies for stroke patient rehabilitation, helping optimize gait mechanics and improve functional outcomes.
   

Other Valuable Information

The study also outlined future research directions, such as including patients with lower Brunnstrom scores, assessing additional muscle function indicators, and considering factors like BMI, gender, and age for their impact on gait. Additionally, the study discussed the limitations of OpenCap technology, such as errors in identifying joint keypoints during dynamic and complex movements and the influence of lighting and clothing on measurement accuracy.

This study not only provides new technical means for gait analysis in stroke patients but also offers important scientific evidence for clinical rehabilitation treatment, demonstrating significant academic value and clinical application prospects.