KSKe Song

Ke Song

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Ke Song has a total of 36 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Gait biomechanics change during pregnancy and postpartum measured using markerless motion capture

    by Michelle Meyers, Ke Song, Afua Nkansah-Andoh, Devyn Russo, Max Zawel, Michelle Kwon, Douglas E Schaubel, Celeste Durnwald, Casey Jo Humbyrd, Josh R. Baxter

  • Open Access

    A Simple Minimum-Setup Pipeline for Using Leg-Worn Inertial Sensors to Track Knee Flexion: Validation on 10 Movements

    by Ke Song, Josh R. Baxter

    Knee motion is a key biomarker in chronic musculoskeletal diseases, yet conventional in-lab optical motion capture falls short of identifying how knee motion continuously impacts joint health outside the lab. Inertial measurement unit (IMU) provides a clinically attractive approach for continuous real-world motion tracking. Our goal was to establish a clinically practical, minimum-setup pipeline for leg-worn IMUs to estimate knee flexion and determine its concurrent validity to optical motion capture during various knee movements. We recorded thigh and shank-worn IMU data with concurrent marker-based and markerless optical motion capture on 10 healthy adults, who performed 10 common movements including walking, running, and stair navigation. We combined IMU functional alignment with data fusion to estimate knee flexion during each movement and compared IMU-based estimate against both motion capture systems using Pearson correlation (Rxy) and root-mean-square difference (RMSD). IMU-estimated knee flexion strongly correlated with motion capture (Rxy ≥ 0.9). RMSDs were smaller for slower movements like walking (RMSD = 4.4–6.0°) while larger during faster movements like running (RMSD = 5.4–9.4°). Wearable IMUs track knee flexion with comparable results to motion capture during daily activities typical to older adults, highlighting their potential for continuous patient monitoring. Our simple pipeline makes IMU-based knee motion tracking more practical and compatible with clinical research. Future research should seek IMU-wearing best practices to secure clinically meaningful data on real-world knee mobility.


  • Open Access

    Two-week cumulative tendon load estimated from insole sensor contact forces is associated with plantar flexor function in Achilles tendinopathy

    by Ke Song, Michelle P. Kwon, Andy K. Smith, Ryan T. Pohlig, Karin Grävare Silbernagel, Josh R. Baxter

    Tendon loading dictates rehabilitation outcomes in Achilles tendinopathy but is difficult to track in the real world. In this study, we used instrumented insole sensors to monitor Achilles tendon load for two weeks in fifteen individuals with Achilles tendinopathy, who also completed assessments of their plantar flexor strength, dynamic function, and survey-based outcomes. We used insole data to estimate two types of cumulative Achilles tendon load: overall (≥ 0.3×body weight) and high-level load (≥ 3×body weight). We determined Pearson correlations between (1) overall and high-level tendon loads, (2) plantar flexor moment, power, work, (3) heel raise height, repetitions, countermovement jump height, and (4) self-reported symptoms and activity. Overall cumulative tendon load moderately correlated to isometric plantar flexor moment ( r = 0.543) and weakly to isokinetic and dynamic functions (0.128–0.413). Cumulative high-level tendon load strongly correlated to heel raise height (0.687) and fast isokinetic moment (0.625), and moderately to other functional measures (0.470–0.592). Symptoms weakly correlated to overall (0.392) and moderately to high-level load (0.436). Self-reported activity weakly correlated to overall (0.297) and strongly to high-level load (0.617). Stronger associations with the high-level Achilles tendon load than the overall load suggest that clinical function assessments provide insight into the real-world performance of high-loading activities. In contrast, the disconnect between overall tendon loading and plantar flexor function may explain the variability in recovery outcomes. Self-reported activity and standard heel raises represent high-level tendon load well, yet they do not always suggest functional deficit. Sensor-monitored tendon load shows promise as a new biomarker for real-world plantar flexor function in Achilles tendinopathy.


  • Open Access

    Body-worn inertial measurement units track knee flexion angles with similar accuracy to optical motion capture

    by Ke Song, Josh R. Baxter

    Knee motion is altered in overuse injuries and chronic diseases like osteoarthritis. Yet, relying on marker-based motion capture in the lab limits our understanding of how knee motion continuously impacts joint health in the real world. Markerless optical motion capture improves clinical suitability, but space and time constraints remain barriers toward real-world assessments. Inertial measurement unit (IMU) sensors have enabled continuous motion tracking outside the lab. In this study, we recorded thigh and shank-worn IMU data concurrently with marker-based and markerless optical motion capture on 10 healthy adults, who performed various daily living and exercise movements. We developed an IMU virtual alignment and data fusion paradigm to estimate knee flexion angle during each movement. We compared IMU-based estimate against marker-based and markerless motion capture using Pearson correlation (R xy ) and root-mean-square difference (RMSD). IMU-estimated knee flexion angle strongly correlated with motion capture (R xy ≥ 0.9). RMSDs were small for slower movements like walking, stairs, and squats (RMSD = 4.4° – 6.0°) while larger during faster movements like running and jumping (RMSD = 5.4° – 9.4°). Our findings show that wearable IMUs track knee flexion with similar accuracy to optical motion capture during daily living activities typical to older adults, highlighting their potential for monitoring real-world mobility in knees with chronic diseases. Conversely, it remains inconclusive whether IMUs accurately track dynamic knee motion relevant to athletic injuries. Future research should seek best practice for IMU wearing and mitigate practical pitfalls to secure high-fidelity data, for identifying clinically meaningful real-world biomarkers of knee mobility.


  • Open Access

    Two weeks of cumulative tendon load monitored by insole sensors is associated with plantar flexor function in Achilles tendinopathy

    by Ke Song, Michelle P. Kwon, Andy K. Smith, Ryan T. Pohlig, Karin Grävare Silbernagel, Josh R. Baxter

    Tendon loading dictates rehabilitation outcomes in Achilles tendinopathy but is difficult to track in the real world. In this study, we used instrumented insole sensors to monitor Achilles tendon load for two weeks in fifteen individuals with Achilles tendinopathy, who also completed assessments of their plantar flexor strength, dynamic function, and survey-based outcomes. We used insole data to estimate two types of cumulative Achilles tendon load: overall (≥0.3×body weight) and high-level load (≥3×body weight). We determined Pearson correlations between (1) overall and high-level tendon loads, (2) plantar flexor moment, power, work, (3) heel raise height, repetitions, countermovement jump height, and (4) self-reported symptoms and activity. Overall cumulative tendon load moderately correlated to isometric plantar flexor moment (r = 0.543) and weakly to isokinetic and dynamic functions (0.128–0.413). Cumulative high-level tendon load strongly correlated to heel raise height (0.687) and fast isokinetic moment (0.625), and moderately to other functional measures (0.470–0.592). Symptoms weakly correlated to overall (0.392) and moderately to high-level load (0.436). Self-reported activity weakly correlated to overall (0.297) and strongly to high-level load (0.617). Stronger associations with the high-level Achilles tendon load than the overall load suggest that clinical function assessments provide insight into the real-world performance of high-loading activities. In contrast, the disconnect between overall tendon loading and plantar flexor function may explain the variability in recovery outcomes. Self-reported activity and standard heel raises represent high-level tendon load well, yet they do not always suggest functional deficit. Sensor-monitored tendon load shows promise as a new biomarker for real-world plantar flexor function in Achilles tendinopathy.


  • Seven Things to Know About Exercise Classification With Inertial Sensing Wearables

    by Vu Phan, Ke Song, Rodrigo Scattone Silva, Karin Gravare Silbernagel, Josh R. Baxter, Eni Halilaj

    Objective: Exercise monitoring with low-cost wearables could improve the efficacy of remote physicaltherapy prescriptions by tracking compliance and informing the delivery of tailored feedback. While a multitude of commercial wearables can detect activities of daily life, such as walking and running, they cannot accurately detect physical-therapy exercises. The goal of this study was to build open-source classifiers for remote physical therapy monitoring and provide insight on how data collection choices may impact classifier performance. Methods: We trained and evaluated multi-class classifiers using data from 19 healthy adults who performed 37 exercises while wearing 10 inertial measurement units on the wrist, pelvis, thighs, shanks, and feet. We investigated the effect of sensor density, location, type, sampling frequency, output granularity, feature engineering, and training-data size on exercise-classification performance. Results: Exercise groups (n = 10) could be classified with 96% accuracy using a set of 10 inertial measurement units (IMUs) and with 89% accuracy using a single pelvis-worn IMU. Multiple sensor modalities (i.e., accelerometers and gyroscopes), high sampling frequencies, and more data from the same population did not improve model performance, but in the future data from diverse populations and better feature engineering could. Conclusions: Given the growing demand for exercise monitoring systems, our sensitivity analyses, along with open-source tools and data, should reduce barriers for product developers, who are balancing accuracy with product formfactor, and increase transparency and trust in clinicians and patients. The open-source data and code are available at https://simtk.org/projects/imu-exercise.


  • Top Cited Papers

    Markerless motion capture estimates of lower extremity kinematics and kinetics are comparable to marker-based across 8 movements

    by Ke Song, Todd J. Hullfish, Rodrigo Scattone Silva, Karin Grävare Silbernagel, Josh R. Baxter


  • Are Abnormal Muscle Biomechanics and Patient-reported Outcomes Associated in Patients With Hip Dysplasia?

    by Tina Ba Wu, Keith R. PhD Lohse, Linda Van Dillen, Ke Song, John C. Clohisy, Michael D. Harris

    Background: Developmental dysplasia of the hip (DDH) is a major risk factor for the early development of hip osteoarthritis. Recent studies have demonstrated how DDH alters hip muscle moment arms and elevates muscle-induced biomechanical variables such as joint reaction forces and acetabular edge loads. Understanding the link between abnormal biomechanics and patient-reported outcome measures (PROMs) is important for evidence-based clinical interventions that improve patient symptoms and functional outcomes. To our knowledge, there are no reports of the relationships between muscle-induced biomechanics and PROMs. Questions/purposes: (1) Are there associations between PROMs and muscle-induced hip biomechanics during gait for patients with DDH and controls? (2) Are there associations among PROMs and separately among biomechanical variables? Methods: Participants in this prospective cross-sectional comparative study included 20 female patients with DDH who had no prior surgery or osteoarthritis and 15 female individuals with no evidence of hip pathology (controls) (age: median 23 years [range 16 to 39 years]; BMI: median 22 kg/m2 [range 17 to 27 kg/m2]). Muscle-induced biomechanical variables for this cohort were reported and had been calculated from patient-specific musculoskeletal models, motion data, and MRI. Biomechanical variables included joint reaction forces, acetabular edge loads, hip center lateralization, and gluteus medius muscle moment arm lengths. PROMs included the Hip Disability and Osteoarthritis Outcome Score (HOOS), the WOMAC, International Hip Outcome Tool-12, National Institutes of Health Patient-Reported Outcome Measure Information System (PROMIS) Pain Interference and Physical Function subscales, and University of California Los Angeles activity scale. Associations between PROMs and biomechanical variables were tested using Spearman rank-order correlations and corrected for multiple comparisons using the Benjamini-Yekutieli method. For this study, associations between variables were considered to exist when correlations were statistically significant (p < 0.05) and were either strong (ρ ≥ 0.60) or moderate (ρ = 0.40 to 0.59). Results: Acetabular edge load impulses (the cumulative acetabular edge load across the gait cycle), medially directed joint reaction forces, and hip center lateralization most commonly demonstrated moderate or strong associations with PROMs. The strongest associations were a negative correlation between acetabular edge load impulse on the superior acetabulum and the HOOS function in daily living subscale (ρ = -0.63; p = 0.001), followed by a negative correlation between hip center lateralization and the HOOS pain subscale (ρ = -0.6; p = 0.003), and a positive correlation between hip center lateralization and the PROMIS pain subscale (ρ = 0.62; p = 0.002). The University of California Los Angeles activity scale was the only PROM that did not demonstrate associations with any biomechanical variable. All PROMs, aside from the University of California Los Angeles activity scale, were associated with one another. Although most of the biomechanical variables were associated with one another, these relationships were not as consistent as those among PROMs. Conclusion: The associations with PROMs detected in the current study suggest that muscle-induced biomechanics may have wide-reaching effects not only on loads within the hip, but also on patients’ perceptions of their health and function. As the treatment of DDH evolves, patient-specific joint preservation strategies may benefit from targeting the underlying causes of biomechanical outcomes associated with PROMs. Level of Evidence: Level III, prognostic study.


  • Open Access

    Patellofemoral Joint Loading Progression Across 35 Weightbearing Rehabilitation Exercises and Activities of Daily Living

    by Ke Song, Rodrigo Scattone Silva, Todd J. Hullfish, Karin Grävare Silbernagel, Josh R. Baxter

    Background: Exercises that provide progressive therapeutic loading are a central component of patellofemoral pain rehabilitation, but quantitative evidence on patellofemoral joint loading is scarce for a majority of common weightbearing rehabilitation exercises. Purpose: To define a loading index to quantify, compare, rank, and categorize overall loading levels in the patellofemoral joint across 35 types of weightbearing rehabilitation exercises and activities of daily living. Study Design: Descriptive laboratory study. Methods: Model-estimated knee flexion angles and extension moments based on motion capture and ground-reaction force data were used to quantify patellofemoral joint loading in 20 healthy participants who performed each exercise. A loading index was computed via a weighted sum of loading peak and cumulative loading impulse for each exercise. The 35 rehabilitation exercises and daily living activities were then ranked and categorized into low, moderate, and high “loading tiers” according to the loading index. Results: Overall patellofemoral loading levels varied substantially across the exercises and activities, with loading peak ranging from 0.6 times body weight during walking to 8.2 times body weight during single-leg decline squat. Most rehabilitation exercises generated a moderate level of patellofemoral joint loading. Few weightbearing exercises provided low-level loading that resembled walking or high-level loading with both high magnitude and duration. Exercises with high knee flexion tended to generate higher patellofemoral joint loading compared with high-intensity exercises. Conclusion: This study quantified patellofemoral joint loading across a large collection of weightbearing exercises in the same cohort. Clinical Relevance: The visualized loading index ranks and modifiable worksheet may assist clinicians in planning patient-specific exercise programs for patellofemoral pain rehabilitation.


  • Open Access

    Markerless motion capture estimates of lower extremity kinematics and kinetics are comparable to marker-based across 8 movements

    by Ke Song, Todd J. Hullfish, Rodrigo Scattone Silva, Karin Grävare Silbernagel, Josh R. Baxter

    Motion analysis is essential for assessing in-vivo human biomechanics. Marker-based motion capture is the standard to analyze human motion, but the inherent inaccuracy and practical challenges limit its utility in large-scale and real-world applications. Markerless motion capture has shown promise to overcome these practical barriers. However, its fidelity in quantifying joint kinematics and kinetics has not been verified across multiple common human movements. In this study, we concurrently captured marker-based and markerless motion data on 10 healthy subjects performing 8 daily living and exercise movements. We calculated the correlation (R xy ) and root-mean-square difference (RMSD) between markerless and marker-based estimates of ankle dorsi-plantarflexion, knee flexion, and three-dimensional hip kinematics (angles) and kinetics (moments) during each movement. Estimates from markerless motion capture matched closely with marker-based in ankle and knee joint angles (R xy ≥ 0.877, RMSD ≤ 5.9°) and moments (R xy ≥ 0.934, RMSD ≤ 2.66 % height × weight). High outcome comparability means the practical benefits of markerless motion capture can simplify experiments and facilitate large-scale analyses. Hip angles and moments demonstrated more differences between the two systems (RMSD: 6.7° – 15.9° and up to 7.15 % height × weight), especially during rapid movements such as running. Markerless motion capture appears to improve the accuracy of hip-related measures, yet more research is needed for validation. We encourage the biomechanics community to continue verifying, validating, and establishing best practices for markerless motion capture, which holds exciting potential to advance collaborative biomechanical research and expand real-world assessments needed for clinical translation.