Classification of patients with patellofemoral pain syndrome during running based on hip joint kinematic feature using machine learning method

Classification of patients with patellofemoral pain syndrome during running based on hip joint kinematic feature using machine learning method | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Classification of patients with patellofemoral pain syndrome during running based on hip joint kinematic feature using machine learning method Mahbubeh Keivan, Hassan Khotanlou, Mehrdad Anbarian This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8928002/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract As a prevalent knee joint disorder, Patellofemoral Pain Syndrome (PFPS) is identified by musculoskeletal issues, often involving symptoms such as pain experienced around or posterior to the patella. Hip joint kinematics may play an essential role in PFPS. Although this condition has been studied from various perspectives, there is no definitive standard clinical method for diagnosis and functional classification. This study aimed to identify the most significant hip joint kinematic features for classifying PFPS patients during running using machine learning (ML) methods. Seven females with unilateral PFPS were paired with controls considering factors such as age, weight, height, and duration of physical activity. In total, 560 running cycles were captured utilizing a motion analysis system based on an inertial measurement unit (IMU). Hip joint kinematic variables, including three-dimensional angles, velocity, and acceleration, were measured. Nineteen features were used as inputs for ML algorithms. Four models—SVM, KNN, ANN, and RF—achieved 99% accuracy in classifying healthy and PFPS patients. Maximum hip adduction emerged as the most significant kinematic feature, and the SVM model performed best for PFPS classification. In conclusion, this study demonstrates that combining IMU sensors with machine learning techniques provides an accurate approach for diagnosing PFPS during running in non-laboratory and clinical environments. Moreover, frontal plane hip joint kinematics appears to be a critical factor in identifying this condition. Health sciences/Diseases Physical sciences/Engineering Health sciences/Health care Health sciences/Medical research Hip Joint Patellofemoral Pain Syndrome Running Machine Learning Figures Figure 1 Figure 2 Figure 3 Key Points - Hip joint adduction identified as the most important kinematic feature for Patellofemoral Pain Syndrome (PFPS) detection. - Frontal hip joint kinematics appears to be a critical factor in identifying the PFPS and should be considered in clinical settings. - Combining IMU sensors with machine learning techniques provides an accurate approach for diagnosing PFPS during running. Introduction As a prevalent knee joint disorder, Patellofemoral Pain Syndrome (PFPS) is identified by musculoskeletal complications, typically manifesting as pain and inflammation behind or around the patella 1 . This condition worsens during activities like running, ascending stairs, jumping, and performing squats, where knee flexion occurs under load 2 . Anatomical and biomechanical factors, including reduced articular cartilage thickness, increased Q angle, and greater dynamic valgus 3 , 4 , contribute to the fact that women are twice as likely as men to develop PFPS 5 . The multifactorial nature of PFPS 6 , 7 and its diverse clinical manifestations make diagnosis and treatment challenging for clinicians 8 . Early detection of PFPS, however, can prevent disease progression and improve patients' quality of life. Therefore, developing efficient diagnostic methods is of critical importance. Traditional approaches, such as clinical examinations and imaging, are valuable but come with limitations, including subjectivity, high costs, and reliance on the clinician's experience. Studies have reported significant variability in diagnostic accuracy when these methods are applied in clinical trials 9 . In summary, no universally reliable and definitive method currently exists for diagnosing PFPS 10 . A key breakthrough in biomechanics has been the integration of data-driven methods leveraging wearable sensors, particularly Inertial Measurement Units (IMUs) 11 , 12 . These sensors, which include accelerometers, gyroscopes, and magnetometers, provide precise motion data in real time and enable three-dimensional analysis of joint angles and movement patterns 13 . IMUs offer several advantages, such as portability, affordability, and ease of use in both clinical and non-clinical settings. They allow users to perform natural movements continuously over extended periods without the need for laboratory setups. Consequently, IMUs facilitate the collection of multiple consecutive gait and running cycles, preserving the individual's natural movement patterns and leading to more accurate data acquisition 14 . Analyzing hip joint angles during running, a repetitive load-bearing activity, can reveal biomechanical alterations associated with PFPS 15 , 16 . Previous research has highlighted that PFPS is linked to specific changes in hip kinematics, such as increased internal rotation or reduced sagittal plane range of motion 17 , 18 , 19 . These biomechanical variations are detectable using wearable sensors. Machine learning algorithms are powerful tools for analyzing complex biomechanical data. Training these algorithms can enable the extraction and classification of features distinguishing healthy individuals from patients with PFPS 20 . Supervised learning methods have demonstrated success in analyzing running biomechanics. For instance, Halilaj et al. (2018) reviewed studies employing machine learning for investigating biomechanical patterns in musculoskeletal and neuromuscular disorders, emphasizing its success in classification, disease risk prediction, and intervention outcome estimation 21 . Combining machine learning algorithms with data from IMUs represents a promising and accurate diagnostic approach for identifying musculoskeletal disorders 22 . This study presents a novel method for distinguishing healthy individuals from those with PFPS using three-dimensional hip joint angle data recorded by IMUs and machine learning algorithms. The aim is to enhance diagnostic capabilities in rehabilitation sciences and movement analysis, grounded in the belief that integrating wearable technologies with advanced computational methods can provide clinicians with practical and precise tools. Moreover, this approach offers a deeper understanding of PFPS biomechanics, paving the way for targeted therapeutic interventions. Methods Participants and Variables This applied descriptive study involved 14 female participants divided into two groups: a healthy group (n = 7, age: 22.57 ± 3.11 years, height: 162.57 ± 4.23 cm, weight: 57.71 ± 13.29 kg) and a group with unilateral PFPS (n = 7, age: 23.57 ± 2.92 years, height: 165.14 ± 4.38 cm, weight: 60.57 ± 6.49 kg). Participants were selected through purposive sampling and voluntarily took part in the study. Inclusion criteria included ages 18–25 years, 1–3 years of athletic experience, and PFPS diagnosis confirmed by an orthopedic specialist. Individuals with musculoskeletal disorders, prior lower limb injuries, or medical issues impairing running over the past six months were not included. Prior to data acquisition, each participant provided written informed consent. This research received ethical approval from the Biomedical Research Ethics Committee at Bu-Ali Sina University, with the assigned Ethics Code: IR.BASU.REC.1402.016. All experiments in this study were conducted according to the relevant guidelines and regulation of the Declaration of Helsinki. Prior to testing, all participants signed written consent form and were familiarized with the protocol. The tests were conducted in a gymnasium where the participants had trained for 1–3 years. The task consisted of running for 1 minute at a self-selected pace on a flat surface. To assess hip kinematics, a motion analysis system incorporating inertial measurement units (IMU Motion Capture Setup V2.0.7; BSN 1–15, software: B2.0.1.0) was utilized. Sensor placement was as follows: one on the torso (aligned with the T12 vertebra in the frontal plane as a reference), two on the thighs (mid-segment, sagittal plane), two on the legs (mid-segment, sagittal plane), and two on the feet. Each sensor’s X-axis was positioned parallel to the surface, while the Y-axis was oriented perpendicularly. For the foot segment, the Y-axis followed the limb’s alignment, and the X-axis remained level with the ground. Sensors were fastened with elastic bands to reduce motion artifacts. Data acquisition occurred concurrently at a 300 Hz rate via software B2.0.1.0 on a laptop. The motion analysis system has been validated in previous studies for measuring kinematic variables 23 . The recorded kinematic parameters encompassed various hip joint movements, including adduction (HADD), abduction (HABD), external rotation (HERD), internal rotation (HIRD), flexion (HFD), and extension (HED). Additionally, corresponding velocity (HADV, HABV, HERV, HIRV, HFV, HEV) and acceleration (HADA, HABA, HERA, HIRA, HFA, HEA) metrics were captured. In MATLAB R2018b, raw data underwent processing through a fourth-order Butterworth low-pass filter, configured with a cutoff frequency of 9 Hz. To account for individual variations in running speed, data normalization was performed. Step cycles were identified based on vertical acceleration of the foot, and 40 step cycles were extracted for each participant. Maximum and minimum values for each variable were calculated for each cycle. Dataset Preparation The final dataset was a 560×20 matrix comprising 19 features and one binary target variable (0 for healthy samples, 1 for PFPS samples). This dataset was formatted for machine learning algorithms and processed using Python 3. Modeling Six distinct machine learning models were compared to determine the most effective classifier for differentiating individuals with PFPS from healthy participants. The assessed models comprised K-Nearest Neighbors (KNN) 24 , Support Vector Machines (SVM) 25 , Decision Trees (DT) 26 , Random Forests (RF) 27 , Naïve Bayes (NB) 26 , and Artificial Neural Networks (ANN) 27 . To enhance model performance and minimize error, a 5-fold cross-validation approach was applied for hyperparameter tuning. The final DT model was constructed using an optimal parameter set: criterion = entropy, tree depth = 5, minimum leaf samples = 1, and minimum split samples = 2. For the RF model, the best-performing parameters included criterion = gini, tree depth = 10, minimum leaf samples = 1, minimum split samples = 2, and the number of estimators set to 50. Also, the hyperparameters euclidean=Best Metric and 1 = Best k was used in building the KNN model. A basic three-layered neural model was structured as follows: an initial layer comprising 16 neurons with a ReLU activation function, a second layer containing 8 neurons with ReLU activation, and a final output layer featuring a single node activated by the Sigmoid function. The model was trained with epochs = 50 and batch_size = 16. The NB model does not have complex parameters to adjust and after training, Cross-validation was employed to optimize the model. A comparative analysis was performed considering Precision, Recall, F1-score, accuracy, and both positive and negative prediction values (TP + TN and FP + FN). The architecture comprised three layers, utilizing the ReLU activation function for hidden layers and the Sigmoid function for the output layer. Training was conducted over 50 epochs with a batch size of 16. To mitigate overfitting, the training process incorporated Early Stopping, leading to a final accuracy of 0.99. Statistical Analyses Model implementation, feature selection, statistical analysis, and normalization were conducted using Python programming with the NumPy, Matplotlib, and Scikit-learn libraries. Results Table 1 provides a performance comparison of six machine learning classifiers, each trained using its optimized hyperparameters for diagnostic assessment. Table 1 of performance metrics of different classifiers on the training set Precision(0) Neural Network SVM (Gaussian) Random Forest (Entropy) Decision Tree (Gini) KNN Nave Bayes (Gaussian) 0.99 1.00 0.97 0.96 1.00 0.76 Precision(1) 1.00 0.99 1.00 0.99 0.97 0.96 Recall(0) 1.00 0.99 1.00 0.99 0.97 0.97 Recall(1) 0.99 1.00 0.97 0.96 1.00 0.69 f1-score(0) 0.99 0.99 0.99 0.97 0.99 0.85 f1-score(1) 0.99 0.99 0.99 0.97 0.99 0.80 accuracy 0.99 0.99 0.99 0.97 0.99 0.82 TP + TN 70 + 69 70 + 69 70 + 68 69 + 67 68 + 70 68 + 48 FP + FN 0 + 1 0 + 1 0 + 2 1 + 3 2 + 0 2 + 22 The bold numbers are the optimal values ​​of each index. The KNN, RF, SVM and ANN models had the highest accuracy with 0.99. However, the SVM and ANN models performed better on a larger number of indices. The confusion matrix (Fig. 1) for SVM and ANN indicates that both models correctly classified 139 samples and misclassified 1 sample. Feature correlation analysis (Fig. 2 ) revealed that Maximum Internal Rotation Acceleration of the Hip (HIRA) had the highest correlation with the classification target. Decision Tree and Random Forest models identified Maximum Hip Adduction Angle (HADD) as the most important feature. Figure 3 shows the importance of features in building these two models. Also, to find the most important feature of the hip joint for classifying healthy and diseased samples, dimensionality reduction using PCA 28 and NCA 29 was used. Table 2 shows the importance of each feature in NCA and the amount of variance of each feature in PC1 and PC2. Maximum hip external rotation acceleration (HERA) and maximum hip extension acceleration (HEA) were the most influential features on PC1 and PC2, respectively. The maximum hip adduction angle (HADD) was also the most important feature identified by NCA, which had the greatest impact on the target. Table 2 Importance of features by NCA method and effect of features on PC1 and PC2 components in PCA method Feature PC1 PC2 NCA HADD 0.18 -0.17 1.69 HABD -0.25 -0.00 0.35 HADV 0.27 -0.03 0.05 HABV -0.28 0.07 0.05 HADA 0.27 -0.06 -0.22 HABA -0.27 0.03 -0.39 HERD -0.00 -0.48 -0.21 HIRD -0.12 -0.43 -0.10 HERV 0.27 -0.05 -0.08 HIRV -0.27 0.01 0.27 HERA 0.28 -0.07 0.13 HIRA -0.28 0.05 0.06 HED 0.25 -0.12 0.25 HFD 0.20 -0.24 -0.00 HEV 0.18 0.19 -0.18 HFV -0.06 -0.28 0.30 HEA 0.16 0.22 -0.10 HFA -0.16 -0.25 -0.03 Duration (Time) -0.17 -0.03 -0.19 PCA1 -0.20 -0.08 -0.20 PCA2 0.03 0.54 0.10 Bold numbers are the highest values ​​of each index Discussion A key aim of this research was to determine the most significant kinematic characteristic of the hip joint for differentiating individuals with patellofemoral pain syndrome (PFPS) from healthy counterparts during running. Inertial motion units (IMUs) offer advantages like portability and the ability to be used outside of a laboratory setting, while still providing acceptable accuracy in measuring joint kinematics when compared to optical motion analysis systems. Despite being prone to movement-related noise, the use of machine learning (ML) methods helps mitigate this limitation 30 . The largest number of outliers came from the PFPS group, which may be attributed to measurement errors or differences in running technique among affected individuals. Maximum Hip Abduction Velocity (HABV) showed the highest correlation with other features, suggesting its strong representational power. It also demonstrated a significant correlation (0.61) with the target, highlighting its critical role in prediction. Maximum Hip Internal Rotation Acceleration (HIRA) displayed the highest correlation (0.64) with the target variable, indicating its direct impact on the prediction outcome. Feature importance analysis using the Neighborhood Component Analysis (NCA) method identified Maximum Hip Adduction Angle (HADD) as the most important feature. Both DT and RF, as hierarchical evaluators of feature importance, also identified HADD as the most significant factor. In contrast, HIRA showed zero importance in these models, likely because correlation is a simple linear measure and does not account for interactions between features. Random Forest, Decision Tree, and NCA methods can capture these non-linear interactions, which enhances their ability to identify important features. The study demonstrated that HADD is the most crucial feature for classifying individuals with PFPS and could be considered as a key diagnostic marker. Earlier research has demonstrated that individuals with PFPS tend to show increased hip adduction angles while running, in contrast to healthy counterparts 31 , 32 , 33 , 34 , reinforcing the significance of HADD as a key differentiating factor. A secondary goal of this research was to determine the most effective ML model for differentiating PFPS patients from healthy individuals. Six models, including SVM, RF, DT, KNN, ANN, and NB, were tested and compared for performance. The models KNN, ANN, RF, and SVM achieved 99% accuracy in classifying the samples. The NB model, which is simpler and does not require specific hyperparameter adjustments, performed the poorest with an accuracy of 82%. This lower performance is likely due to the complexity of the feature relationships, which NB is less capable of modeling. Beyond accuracy, the impact of Type I (false positive) and Type II (false negative) errors was taken into account when selecting the optimal model. Among the models achieving 99% accuracy, SVM and ANN exhibited the least total error (Type I and Type II errors combined). SVM misclassified one healthy sample as affected by PFPS (Type I error), which could lead to unnecessary stress and testing. On the other hand, ANN misclassified one PFPS patient as healthy (Type II error), which is more concerning as untreated patients may experience worsening conditions. Therefore, SVM is considered the best model for distinguishing between healthy and PFPS samples. No previous study has directly compared multiple machine learning models with kinematic hip joint features for classifying healthy and PFPS samples during running. A prior study employed three joint angles (hip flexion, knee flexion, and ankle dorsiflexion) alongside EMG data as inputs for an Extreme Learning Machine (ELM) model to predict PFPS. The model’s performance was compared in terms of accuracy and computational time with other models such as KNN, SVM, RDF, NB, and MLP 35 . Conclusion To our knowledge, this is the first study utilizing kinematic features of the hip joint, collected via IMU sensors, as input to machine learning models for distinguishing individuals with PFPS from healthy participants during running. Despite the small sample size, the models SVM, RF, KNN, and ANN achieved an accuracy of 99% in classification. Based on the confusion matrix comparison, SVM demonstrated the best performance. Additionally, NCA, DT, and RF identified HADD as the most important kinematic feature for PFPS detection. While these findings are important for identifying patients without the need for expensive equipment and in real-life activity settings, further studies with larger sample sizes, including both male and female participants and a greater number of sensors, are necessary. The findings of this research indicate that integrating IMU sensors with machine learning approaches offers a precise means for detecting PFPS while running in non-laboratory or non-clinical settings. Moreover, frontal plane hip kinematics appear to play a more significant role in identifying this condition. Abbreviations PFPS Patellofemoral Pain Syndrome ML Machine learning IMU Inertial measurement unit HABD Hip abduction HADD Hip adduction HADV Hip adduction velocity HABV Hip abduction velocity HADA Hip adduction acceleration HABA Hip abduction acceleration HERD Hip external rotation HIRD Hip internal rotation HERV Hip external rotation velocity HIRA Hip internal rotation acceleration HED Hip extension HFD Hip flexion HEV Hip extension velocity HFV Hip flexion velocity HEA Hip extension acceleration HFA Hip flexion acceleration KNN K-Nearest Neighbors SVM Support Vector Machines ANN Artificial Neural Networks RF Random Forests Declarations Conflict of interest The researchers confirm that there are no conflicts of interest. Funding Financial support for this study was not provided by any public or non-profit funding organizations. Author Contribution Data Collection: Mahbubeh Keivan; Conceptualization: Mahbubeh Keivan, Mehrdad Anbarian; Methodology: Mahbubeh Keivan, Mehrdad Anbarian, Hassan Khotanlou; Software: Hassan Khotanlou; Formal Analysis: Hassan Khotanlou, Mahbubeh Keivan, Mehrdad Anbarian; Writing—Review and Editing: Mahbubeh Keivan, Mehrdad Anbarian; Supervision: Mehrdad Anbarian Acknowledgement This research is extracted from PhD dissertation of Sports Biomechanics at Bu-Ali Sina University. The authors would like to thank all subjects who participated in this study. Data Availability The data used and analyzed during the study are available from the corresponding author upon reasonable request. The data are not publicly available due to privacy on ethical issues. References Neal, B. S., Barton, C. J., Birn-Jeffery, A. & Morrissey, D. Increased hip adduction during running is associated with patellofemoral pain and differs between males and females: A case-control study. J. Biomech. 91 , 133–139 (2019). Crossley, K. M. et al. Patellofemoral pain consensus statement from the 4th international patellofemoral pain. Br. J. Sports Med. 50 (14), 839–843 (2016). Dutton, R. A., Khadavi, M. J. & Fredericson, M. Patellofemoral pain. Phys. Med. Rehabil Clin. N Am. 27 (1), 31–52. https://doi.org/10.1016/j.pmr.2015.08.008 (2016). Petersen, W. et al. Patellofemoral pain syndrome. Knee Surg. Sports Traumatol. Arthrosc. 22 (10), 2264–2274. https://doi.org/10.1007/s00167-014-2863-1 (2014). Smith, B. E. et al. Incidence and prevalence of patellofemoral pain: A systematic review and meta-analysis. PLoS One . 13 (1), e0190892. https://doi.org/10.1371/journal.pone.0190892 (2018). Hu, H. et al. Effects of neuromuscular training on pain intensity and self-reported functionality for patellofemoral pain syndrome in runners: Study protocol for a randomized controlled clinical trial. Trials 20 (1), 409. https://doi.org/10.1186/s13063-019-3454-2 (2019). Petersen, W. et al. Patellofemoral pain syndrome. Knee Surg. Sports Traumatol. Arthrosc. 22 (10), 2264–2274. https://doi.org/10.1007/s00167-013-2759-6 (2014). Nunes, G. S. et al. Clinical test for diagnosis of patellofemoral pain syndrome: Systematic review with meta-analysis. Physiotherapy Sport . 14 (1), 54–59. https://doi.org/10.1016/j.ptsp.2012.07.003 (2013). Cook, C. et al. Best tests/clinical findings for screening and diagnosis of patellofemoral pain syndrome: A systematic review. Physiotherapy 98 (2), 93–100 (2012). Xiang, L. et al. Recent machine learning progress in lower limb running biomechanics with wearable technology: A systematic review. Front. Neurorobot . 16 , 913052. https://doi.org/10.3389/fnbot.2022.913052 (2022). Mason, R. et al. Wearables for running gait analysis: A systematic review. Sports Med. 53 (1), 241–268. https://doi.org/10.1007/s40279-022-01760-6 (2023). Delgado-García, G. et al.. IMU gyroscopes are a valid alternative to 3D optical motion capture system for angular kinematics analysis in tennis. Proc. Inst. Mech. Eng. P J. Sports Eng. Technol. 235 (1), 3–12. https://doi.org/10.1177/1754337120965444 (2021). Johansen, S. K. et al. Exploring patients' and physiotherapists' visions on modelling treatments and optimising self-management strategies for patellofemoral pain: A future workshop approach. Musculoskelet. Sci. Pract. 60 , 102567. https://doi.org/10.1016/j.msksp.2022.102567 (2022). Noehren, B., Pohl, M. B., Sanchez, Z., Cunningham, T. & Lattermann, C. Proximal and distal kinematics in female runners with patellofemoral pain. Clin. Biomech. 26 (10), 1049–1055. https://doi.org/10.1016/j.clinbiomech.2011.10.005 (2011). Luedke, L. E., Heiderscheit, B. C., Williams, D. S. & Rauh, M. J. Association of isometric strength of hip and knee muscles with injury risk in high school cross-country runners. J. Orthop. Sports Phys. Ther. 46 (8), 667–676. https://doi.org/10.2519/jospt.2016.6405 (2016). Noehren, B., Hamill, J. & Davis, I. Prospective evidence for a hip etiology in patellofemoral pain. Med. Sci. Sports Exerc. 45 (6), 1120–1124. https://doi.org/10.1249/MSS.0b013e31828249d2 (2013). Souza, R. B. & Powers, C. M. Differences in hip kinematics, muscle strength, and muscle activation between subjects with and without patellofemoral pain. J. Orthop. Sports Phys. Ther. 39 (1), 12–19. https://doi.org/10.2519/jospt.2009.2885 (2009). Willy, R. W. & Davis, I. S. The effect of a hip-strengthening program on mechanics during running and during a single-leg squat. J. Orthop. Sports Phys. Ther. 41 (9), 625–632. https://doi.org/10.2519/jospt.2011.3470 (2011). Leckey, C. et al. Machine learning approaches to injury risk prediction in sport: A scoping review with evidence synthesis. Br. J. Sports Med. 59 (7), 491–500. https://doi.org/10.1136/bjsports-2024-108576 (2024). Halilaj, E. et al. Machine learning in human movement biomechanics: Best practices, common pitfalls, and new opportunities. J. Biomech. 81 , 1–11. https://doi.org/10.1016/j.jbiomech.2018.09.009 (2018). Johansen, S. K. et al. Exploring patients' and physiotherapists' visions on modelling treatments and optimising self-management strategies for patellofemoral pain: A future workshop approach. Musculoskelet. Sci. Pract. 60 , 102567. https://doi.org/10.1016/j.msksp.2022.102567 (2022). Heidarimoghaddam, R., Hosseini, M. H., Ilbeigi, S., Anbarian, M. & Tapak, L. Kinematic analysis of head and trunk in individual and team one-handed carrying. Int. J. Ind. Ergon. 94 , 103422. https://doi.org/10.1016/j.ergon.2023.103422 (2023). Liu, H. Value evaluation of knee joint sports injury detection model-aided diagnosis based on machine learning. Front. Phys. 11 , 1166275. https://doi.org/10.3389/fphy.2023.1166275 (2023). Tayfur, B., Charuphongsa, C., Morrissey, D. & Miller, S. C. Neuromuscular function of the knee joint following knee injuries: Does it ever get back to normal? A systematic review with meta-analyses. Sports Med. 51 (2), 321–338. https://doi.org/10.1007/s40279-020-01386-6 (2021). Iacobescu, P., Marina, V., Anghel, C. & Anghele, A. D. Evaluating binary classifiers for cardiovascular disease prediction: Enhancing early diagnostic capabilities. J. Cardiovasc. Dev. Dis. 11 (12), 396. https://doi.org/10.3390/jcdd11120396 (2024). Khairuddin, M. Z. F. et al. Occupational injury risk mitigation: Machine learning approach and feature optimization for smart workplace surveillance. Int. J. Environ. Res. Public. Health . 19 (21), 13962. https://doi.org/10.3390/ijerph192113962 (2022). Ray, P., Reddy, S. S. & Banerjee, T. Various dimension reduction techniques for high dimensional data analysis: A review. Artif. Intell. Rev. 54 (3), 3473–3485. https://doi.org/10.1007/s10462-020-09928-0 (2021). Raghu, S. & Sriraam, N. Classification of focal and non-focal EEG signals using neighborhood component analysis and machine learning algorithms. Expert Syst. Appl. 113 , 18–32. https://doi.org/10.1016/j.eswa.2018.06.042 (2018). Mohammadi Moghadam Sh, Yeung, T. & Choisne, J. A comparison of machine learning models’ accuracy in predicting lower-limb joints’ kinematics, kinetics, and muscle forces from wearable sensors. Sci. Rep. 13 , 5046. https://doi.org/10.1038/s41598-023-31906-z (2023). Dierks, T. A., Manal, K. T., Hamill, J. & Davis, I. S. Proximal and distal influences on hip and knee kinematics in runners with patellofemoral pain during a prolonged run. J. Orthop. Sports Phys. Ther. 38 (8), 448–456. https://doi.org/10.2519/jospt.2008.2490 (2008). Mündermann, A., Asay, J. L., Mündermann, L. & Andriacchi, T. P. Implications of increased medio-lateral trunk sway for ambulatory mechanics. J. Biomech. 41 (1), 165–170. https://doi.org/10.1016/j.jbiomech.2007.09.020 (2008). Lin, C. C. et al. Machine learning-enhanced prediction of running-related injuries. PLoS One . 18 (2), e0281427. https://doi.org/10.1371/journal.pone.0281427 (2023). Malinauskas, R. A., Ogami, T., Witzke, K. A. & Doyle, W. A. Development of a methodology for computing patellofemoral joint mechanics. J. Biomech. Eng. 142 (4), 041010. https://doi.org/10.1115/1.4045753 (2020). Lavagnino, M. et al. Patellofemoral loading and knee flexion during squatting. J. Orthop. Res. 26 (8), 1084–1090. https://doi.org/10.1002/jor.20621 (2008). Donnelly, C. J., Elliott, B. C., Doyle, T. L., Finch, C. F. & Dempsey, A. R. Lloyd D.G. Changes in knee joint biomechanics following balance and technique training and a season of Australian football. Br. J. Sports Med. 46 (13), 917–922. https://doi.org/10.1136/bjsports-2012-090936 (2012). Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8928002","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":601465296,"identity":"4882cffe-7322-459c-bc5d-c611d79129e2","order_by":0,"name":"Mahbubeh Keivan","email":"","orcid":"","institution":"Bu-Ali Sina University","correspondingAuthor":false,"prefix":"","firstName":"Mahbubeh","middleName":"","lastName":"Keivan","suffix":""},{"id":601465299,"identity":"aca18869-8d36-4f72-9d36-4c3f4854091c","order_by":1,"name":"Hassan Khotanlou","email":"","orcid":"","institution":"Bu-Ali Sina University","correspondingAuthor":false,"prefix":"","firstName":"Hassan","middleName":"","lastName":"Khotanlou","suffix":""},{"id":601465302,"identity":"284275ff-7704-4751-a670-7b3a52baf080","order_by":2,"name":"Mehrdad Anbarian","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA00lEQVRIiWNgGAWjYFACHiA2YGDgl4AJMBOrRXIGaVpAum4Q6yz+2b0HP1cU2OUZ325+/JmHwU6egZ33AV4tEnfOJUueMUguNrtzzEyahyHZsIGZ3QC/NTdyDCQbDJgTt91IMGPmYWBOYGBmw69D/kaO8c8Gg/rEzTPSPwMdVk9Yi8GNHDOgLYcTN0jkGAAddpiwFsMbeWmWDQbHE2fcOVMmOcfguGEbIS1yN3IP32z4U53YP7t984c3FdXy/PzH8GtBdycDAwE7RsEoGAWjYBQQAwDtkTxNy87Z9gAAAABJRU5ErkJggg==","orcid":"","institution":"Bu-Ali Sina University","correspondingAuthor":true,"prefix":"","firstName":"Mehrdad","middleName":"","lastName":"Anbarian","suffix":""}],"badges":[],"createdAt":"2026-02-20 17:10:01","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8928002/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8928002/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104403901,"identity":"6ad9e5e9-9802-4ad7-a632-8bc190631c9e","added_by":"auto","created_at":"2026-03-11 12:19:20","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":41895,"visible":true,"origin":"","legend":"\u003cp\u003eConfusion matrix: Final ANN model (A), Final SVM model (B)\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8928002/v1/8370c19ff3cf795bed869b5d.jpg"},{"id":104206938,"identity":"063227e2-959e-40ca-b64a-3cb91d78a44d","added_by":"auto","created_at":"2026-03-09 07:05:09","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":54667,"visible":true,"origin":"","legend":"\u003cp\u003eCorrelation of features with the target\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8928002/v1/48c0b3ad6fdc7bdd12e1ca08.jpg"},{"id":104206939,"identity":"d0f0fbfb-7e63-4c5f-ad05-d7c28407fba3","added_by":"auto","created_at":"2026-03-09 07:05:09","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":71591,"visible":true,"origin":"","legend":"\u003cp\u003eAnalysis of the importance of features: RF model (A), DT model (B)\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8928002/v1/c863d341c3f6c2f9f1e8d7e5.jpg"},{"id":104409190,"identity":"c57cee34-8b08-48f8-a421-8f30ff1063d5","added_by":"auto","created_at":"2026-03-11 12:44:21","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":911479,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8928002/v1/a89d3ca9-3f23-4566-b66d-9f9cca006520.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Classification of patients with patellofemoral pain syndrome during running based on hip joint kinematic feature using machine learning method","fulltext":[{"header":"Key Points","content":"\u003cp\u003e- Hip joint adduction identified as the most important kinematic feature for Patellofemoral Pain Syndrome (PFPS) detection.\u003c/p\u003e\u003cp\u003e- Frontal hip joint kinematics appears to be a critical factor in identifying the PFPS and should be considered in clinical settings.\u003c/p\u003e\u003cp\u003e- Combining IMU sensors with machine learning techniques provides an accurate approach for diagnosing PFPS during running.\u003c/p\u003e"},{"header":"Introduction","content":"\u003cp\u003eAs a prevalent knee joint disorder, Patellofemoral Pain Syndrome (PFPS) is identified by musculoskeletal complications, typically manifesting as pain and inflammation behind or around the patella\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. This condition worsens during activities like running, ascending stairs, jumping, and performing squats, where knee flexion occurs under load\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Anatomical and biomechanical factors, including reduced articular cartilage thickness, increased Q angle, and greater dynamic valgus\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e, contribute to the fact that women are twice as likely as men to develop PFPS\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. The multifactorial nature of PFPS\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e and its diverse clinical manifestations make diagnosis and treatment challenging for clinicians\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Early detection of PFPS, however, can prevent disease progression and improve patients' quality of life. Therefore, developing efficient diagnostic methods is of critical importance. Traditional approaches, such as clinical examinations and imaging, are valuable but come with limitations, including subjectivity, high costs, and reliance on the clinician's experience. Studies have reported significant variability in diagnostic accuracy when these methods are applied in clinical trials\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. In summary, no universally reliable and definitive method currently exists for diagnosing PFPS\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eA key breakthrough in biomechanics has been the integration of data-driven methods leveraging wearable sensors, particularly Inertial Measurement Units (IMUs)\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. These sensors, which include accelerometers, gyroscopes, and magnetometers, provide precise motion data in real time and enable three-dimensional analysis of joint angles and movement patterns\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. IMUs offer several advantages, such as portability, affordability, and ease of use in both clinical and non-clinical settings. They allow users to perform natural movements continuously over extended periods without the need for laboratory setups. Consequently, IMUs facilitate the collection of multiple consecutive gait and running cycles, preserving the individual's natural movement patterns and leading to more accurate data acquisition\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eAnalyzing hip joint angles during running, a repetitive load-bearing activity, can reveal biomechanical alterations associated with PFPS\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. Previous research has highlighted that PFPS is linked to specific changes in hip kinematics, such as increased internal rotation or reduced sagittal plane range of motion\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. These biomechanical variations are detectable using wearable sensors.\u003c/p\u003e \u003cp\u003eMachine learning algorithms are powerful tools for analyzing complex biomechanical data. Training these algorithms can enable the extraction and classification of features distinguishing healthy individuals from patients with PFPS\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. Supervised learning methods have demonstrated success in analyzing running biomechanics. For instance, Halilaj et al. (2018) reviewed studies employing machine learning for investigating biomechanical patterns in musculoskeletal and neuromuscular disorders, emphasizing its success in classification, disease risk prediction, and intervention outcome estimation\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. Combining machine learning algorithms with data from IMUs represents a promising and accurate diagnostic approach for identifying musculoskeletal disorders \u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThis study presents a novel method for distinguishing healthy individuals from those with PFPS using three-dimensional hip joint angle data recorded by IMUs and machine learning algorithms. The aim is to enhance diagnostic capabilities in rehabilitation sciences and movement analysis, grounded in the belief that integrating wearable technologies with advanced computational methods can provide clinicians with practical and precise tools. Moreover, this approach offers a deeper understanding of PFPS biomechanics, paving the way for targeted therapeutic interventions.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eParticipants and Variables\u003c/h2\u003e \u003cp\u003eThis applied descriptive study involved 14 female participants divided into two groups: a healthy group (n\u0026thinsp;=\u0026thinsp;7, age: 22.57\u0026thinsp;\u0026plusmn;\u0026thinsp;3.11 years, height: 162.57\u0026thinsp;\u0026plusmn;\u0026thinsp;4.23 cm, weight: 57.71\u0026thinsp;\u0026plusmn;\u0026thinsp;13.29 kg) and a group with unilateral PFPS (n\u0026thinsp;=\u0026thinsp;7, age: 23.57\u0026thinsp;\u0026plusmn;\u0026thinsp;2.92 years, height: 165.14\u0026thinsp;\u0026plusmn;\u0026thinsp;4.38 cm, weight: 60.57\u0026thinsp;\u0026plusmn;\u0026thinsp;6.49 kg). Participants were selected through purposive sampling and voluntarily took part in the study. Inclusion criteria included ages 18\u0026ndash;25 years, 1\u0026ndash;3 years of athletic experience, and PFPS diagnosis confirmed by an orthopedic specialist. Individuals with musculoskeletal disorders, prior lower limb injuries, or medical issues impairing running over the past six months were not included. Prior to data acquisition, each participant provided written informed consent. This research received ethical approval from the Biomedical Research Ethics Committee at Bu-Ali Sina University, with the assigned Ethics Code: IR.BASU.REC.1402.016. All experiments in this study were conducted according to the relevant guidelines and regulation of the Declaration of Helsinki. Prior to testing, all participants signed written consent form and were familiarized with the protocol. The tests were conducted in a gymnasium where the participants had trained for 1\u0026ndash;3 years. The task consisted of running for 1 minute at a self-selected pace on a flat surface.\u003c/p\u003e \u003cp\u003eTo assess hip kinematics, a motion analysis system incorporating inertial measurement units (IMU Motion Capture Setup V2.0.7; BSN 1\u0026ndash;15, software: B2.0.1.0) was utilized.\u003c/p\u003e \u003cp\u003eSensor placement was as follows: one on the torso (aligned with the T12 vertebra in the frontal plane as a reference), two on the thighs (mid-segment, sagittal plane), two on the legs (mid-segment, sagittal plane), and two on the feet. Each sensor\u0026rsquo;s X-axis was positioned parallel to the surface, while the Y-axis was oriented perpendicularly. For the foot segment, the Y-axis followed the limb\u0026rsquo;s alignment, and the X-axis remained level with the ground. Sensors were fastened with elastic bands to reduce motion artifacts. Data acquisition occurred concurrently at a 300 Hz rate via software B2.0.1.0 on a laptop. The motion analysis system has been validated in previous studies for measuring kinematic variables \u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe recorded kinematic parameters encompassed various hip joint movements, including adduction (HADD), abduction (HABD), external rotation (HERD), internal rotation (HIRD), flexion (HFD), and extension (HED). Additionally, corresponding velocity (HADV, HABV, HERV, HIRV, HFV, HEV) and acceleration (HADA, HABA, HERA, HIRA, HFA, HEA) metrics were captured.\u003c/p\u003e \u003cp\u003eIn MATLAB R2018b, raw data underwent processing through a fourth-order Butterworth low-pass filter, configured with a cutoff frequency of 9 Hz. To account for individual variations in running speed, data normalization was performed. Step cycles were identified based on vertical acceleration of the foot, and 40 step cycles were extracted for each participant. Maximum and minimum values for each variable were calculated for each cycle.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eDataset Preparation\u003c/h3\u003e\n\u003cp\u003eThe final dataset was a 560\u0026times;20 matrix comprising 19 features and one binary target variable (0 for healthy samples, 1 for PFPS samples). This dataset was formatted for machine learning algorithms and processed using Python 3.\u003c/p\u003e\n\u003ch3\u003eModeling\u003c/h3\u003e\n\u003cp\u003eSix distinct machine learning models were compared to determine the most effective classifier for differentiating individuals with PFPS from healthy participants. The assessed models comprised K-Nearest Neighbors (KNN) \u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e, Support Vector Machines (SVM) \u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e, Decision Trees (DT) \u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e, Random Forests (RF) \u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e, Na\u0026iuml;ve Bayes (NB) \u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e, and Artificial Neural Networks (ANN) \u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. To enhance model performance and minimize error, a 5-fold cross-validation approach was applied for hyperparameter tuning.\u003c/p\u003e \u003cp\u003eThe final DT model was constructed using an optimal parameter set: criterion\u0026thinsp;=\u0026thinsp;entropy, tree depth\u0026thinsp;=\u0026thinsp;5, minimum leaf samples\u0026thinsp;=\u0026thinsp;1, and minimum split samples\u0026thinsp;=\u0026thinsp;2. For the RF model, the best-performing parameters included criterion\u0026thinsp;=\u0026thinsp;gini, tree depth\u0026thinsp;=\u0026thinsp;10, minimum leaf samples\u0026thinsp;=\u0026thinsp;1, minimum split samples\u0026thinsp;=\u0026thinsp;2, and the number of estimators set to 50. Also, the hyperparameters euclidean=Best Metric and 1\u0026thinsp;=\u0026thinsp;Best k was used in building the KNN model. A basic three-layered neural model was structured as follows: an initial layer comprising 16 neurons with a ReLU activation function, a second layer containing 8 neurons with ReLU activation, and a final output layer featuring a single node activated by the Sigmoid function. The model was trained with epochs\u0026thinsp;=\u0026thinsp;50 and batch_size\u0026thinsp;=\u0026thinsp;16. The NB model does not have complex parameters to adjust and after training, Cross-validation was employed to optimize the model. A comparative analysis was performed considering Precision, Recall, F1-score, accuracy, and both positive and negative prediction values (TP\u0026thinsp;+\u0026thinsp;TN and FP\u0026thinsp;+\u0026thinsp;FN). The architecture comprised three layers, utilizing the ReLU activation function for hidden layers and the Sigmoid function for the output layer. Training was conducted over 50 epochs with a batch size of 16. To mitigate overfitting, the training process incorporated Early Stopping, leading to a final accuracy of 0.99.\u003c/p\u003e\n\u003ch3\u003eStatistical Analyses\u003c/h3\u003e\n\u003cp\u003eModel implementation, feature selection, statistical analysis, and normalization were conducted using Python programming with the NumPy, Matplotlib, and Scikit-learn libraries.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eTable 1 provides a performance comparison of six machine learning classifiers, each trained using its optimized hyperparameters for diagnostic assessment.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e of performance metrics of different classifiers on the training set\u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cb\u003ePrecision(0)\u003c/b\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNeural Network\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSVM (Gaussian)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRandom Forest\u003c/p\u003e \u003cp\u003e(Entropy)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDecision Tree\u003c/p\u003e \u003cp\u003e(Gini)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eNave Bayes\u003c/p\u003e \u003cp\u003e(Gaussian)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.99\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e1.00\u003c/b\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e1.00\u003c/b\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.76\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePrecision(1)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e1.00\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e1.00\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRecall(0)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e1.00\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e1.00\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRecall(1)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e1.00\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e1.00\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.69\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ef1-score(0)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ef1-score(1)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.80\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eaccuracy\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.99\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTP\u0026thinsp;+\u0026thinsp;TN\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e70\u0026thinsp;+\u0026thinsp;69\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e70\u0026thinsp;+\u0026thinsp;69\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e70\u0026thinsp;+\u0026thinsp;68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e69\u0026thinsp;+\u0026thinsp;67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e68\u0026thinsp;+\u0026thinsp;70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e68\u0026thinsp;+\u0026thinsp;48\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eFP\u0026thinsp;+\u0026thinsp;FN\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0\u0026thinsp;+\u0026thinsp;1\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0\u0026thinsp;+\u0026thinsp;1\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u0026thinsp;+\u0026thinsp;2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u0026thinsp;+\u0026thinsp;3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2\u0026thinsp;+\u0026thinsp;0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2\u0026thinsp;+\u0026thinsp;22\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe bold numbers are the optimal values ​​of each index.\u003c/p\u003e \u003cp\u003eThe KNN, RF, SVM and ANN models had the highest accuracy with 0.99. However, the SVM and ANN models performed better on a larger number of indices. The confusion matrix (Fig.\u0026nbsp;1) for SVM and ANN indicates that both models correctly classified 139 samples and misclassified 1 sample.\u003c/p\u003e \u003cp\u003eFeature correlation analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003e) revealed that Maximum Internal Rotation Acceleration of the Hip (HIRA) had the highest correlation with the classification target.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eDecision Tree and Random Forest models identified Maximum Hip Adduction Angle (HADD) as the most important feature. Figure\u0026nbsp;3 shows the importance of features in building these two models.\u003c/p\u003e \u003cp\u003eAlso, to find the most important feature of the hip joint for classifying healthy and diseased samples, dimensionality reduction using PCA \u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e and NCA \u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e was used. Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows the importance of each feature in NCA and the amount of variance of each feature in PC1 and PC2. Maximum hip external rotation acceleration (HERA) and maximum hip extension acceleration (HEA) were the most influential features on PC1 and PC2, respectively. The maximum hip adduction angle (HADD) was also the most important feature identified by NCA, which had the greatest impact on the target.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eImportance of features by NCA method and effect of features on PC1 and PC2 components in PCA method\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFeature\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePC1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePC2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNCA\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHADD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e1.69\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHABD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.35\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHADV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHABV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.07\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHADA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.06\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e-0.22\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHABA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e-0.39\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHERD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e-0.21\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHIRD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e-0.10\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHERV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e-0.08\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHIRV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.01\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.27\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHERA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.28\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.07\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHIRA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.06\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHED\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.25\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHFD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e-0.00\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHEV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e-0.18\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHFV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.06\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.30\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHEA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.22\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e-0.10\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHFA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e-0.03\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDuration (Time)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e-0.19\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePCA1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e-0.20\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePCA2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.10\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eBold numbers are the highest values ​​of each index\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eA key aim of this research was to determine the most significant kinematic characteristic of the hip joint for differentiating individuals with patellofemoral pain syndrome (PFPS) from healthy counterparts during running. Inertial motion units (IMUs) offer advantages like portability and the ability to be used outside of a laboratory setting, while still providing acceptable accuracy in measuring joint kinematics when compared to optical motion analysis systems. Despite being prone to movement-related noise, the use of machine learning (ML) methods helps mitigate this limitation \u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. The largest number of outliers came from the PFPS group, which may be attributed to measurement errors or differences in running technique among affected individuals.\u003c/p\u003e \u003cp\u003eMaximum Hip Abduction Velocity (HABV) showed the highest correlation with other features, suggesting its strong representational power. It also demonstrated a significant correlation (0.61) with the target, highlighting its critical role in prediction. Maximum Hip Internal Rotation Acceleration (HIRA) displayed the highest correlation (0.64) with the target variable, indicating its direct impact on the prediction outcome. Feature importance analysis using the Neighborhood Component Analysis (NCA) method identified Maximum Hip Adduction Angle (HADD) as the most important feature. Both DT and RF, as hierarchical evaluators of feature importance, also identified HADD as the most significant factor. In contrast, HIRA showed zero importance in these models, likely because correlation is a simple linear measure and does not account for interactions between features. Random Forest, Decision Tree, and NCA methods can capture these non-linear interactions, which enhances their ability to identify important features. The study demonstrated that HADD is the most crucial feature for classifying individuals with PFPS and could be considered as a key diagnostic marker. Earlier research has demonstrated that individuals with PFPS tend to show increased hip adduction angles while running, in contrast to healthy counterparts \u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e, reinforcing the significance of HADD as a key differentiating factor.\u003c/p\u003e \u003cp\u003eA secondary goal of this research was to determine the most effective ML model for differentiating PFPS patients from healthy individuals. Six models, including SVM, RF, DT, KNN, ANN, and NB, were tested and compared for performance. The models KNN, ANN, RF, and SVM achieved 99% accuracy in classifying the samples. The NB model, which is simpler and does not require specific hyperparameter adjustments, performed the poorest with an accuracy of 82%. This lower performance is likely due to the complexity of the feature relationships, which NB is less capable of modeling. Beyond accuracy, the impact of Type I (false positive) and Type II (false negative) errors was taken into account when selecting the optimal model.\u003c/p\u003e \u003cp\u003eAmong the models achieving 99% accuracy, SVM and ANN exhibited the least total error (Type I and Type II errors combined). SVM misclassified one healthy sample as affected by PFPS (Type I error), which could lead to unnecessary stress and testing. On the other hand, ANN misclassified one PFPS patient as healthy (Type II error), which is more concerning as untreated patients may experience worsening conditions. Therefore, SVM is considered the best model for distinguishing between healthy and PFPS samples.\u003c/p\u003e \u003cp\u003eNo previous study has directly compared multiple machine learning models with kinematic hip joint features for classifying healthy and PFPS samples during running. A prior study employed three joint angles (hip flexion, knee flexion, and ankle dorsiflexion) alongside EMG data as inputs for an Extreme Learning Machine (ELM) model to predict PFPS. The model\u0026rsquo;s performance was compared in terms of accuracy and computational time with other models such as KNN, SVM, RDF, NB, and MLP \u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eTo our knowledge, this is the first study utilizing kinematic features of the hip joint, collected via IMU sensors, as input to machine learning models for distinguishing individuals with PFPS from healthy participants during running. Despite the small sample size, the models SVM, RF, KNN, and ANN achieved an accuracy of 99% in classification. Based on the confusion matrix comparison, SVM demonstrated the best performance. Additionally, NCA, DT, and RF identified HADD as the most important kinematic feature for PFPS detection. While these findings are important for identifying patients without the need for expensive equipment and in real-life activity settings, further studies with larger sample sizes, including both male and female participants and a greater number of sensors, are necessary. The findings of this research indicate that integrating IMU sensors with machine learning approaches offers a precise means for detecting PFPS while running in non-laboratory or non-clinical settings. Moreover, frontal plane hip kinematics appear to play a more significant role in identifying this condition.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003ePFPS \u0026nbsp; \u0026nbsp; \u0026nbsp; Patellofemoral Pain Syndrome\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eML \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Machine learning\u003c/p\u003e\n\u003cp\u003eIMU \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Inertial measurement unit\u003c/p\u003e\n\u003cp\u003eHABD \u0026nbsp; \u0026nbsp; Hip abduction\u003c/p\u003e\n\u003cp\u003eHADD \u0026nbsp; \u0026nbsp; Hip adduction\u003c/p\u003e\n\u003cp\u003eHADV \u0026nbsp; \u0026nbsp; Hip adduction velocity\u003c/p\u003e\n\u003cp\u003eHABV \u0026nbsp; \u0026nbsp; \u0026nbsp;Hip abduction velocity\u003c/p\u003e\n\u003cp\u003eHADA \u0026nbsp; \u0026nbsp; \u0026nbsp;Hip adduction acceleration\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHABA \u0026nbsp; \u0026nbsp; \u0026nbsp;Hip abduction acceleration\u003c/p\u003e\n\u003cp\u003eHERD \u0026nbsp; \u0026nbsp; \u0026nbsp;Hip external rotation\u003c/p\u003e\n\u003cp\u003eHIRD \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Hip internal rotation\u003c/p\u003e\n\u003cp\u003eHERV \u0026nbsp; \u0026nbsp; \u0026nbsp; Hip external rotation velocity\u003c/p\u003e\n\u003cp\u003eHIRA \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Hip internal rotation acceleration\u003c/p\u003e\n\u003cp\u003eHED \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Hip extension\u003c/p\u003e\n\u003cp\u003eHFD \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Hip flexion\u003c/p\u003e\n\u003cp\u003eHEV \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Hip extension velocity \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHFV \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Hip flexion velocity\u003c/p\u003e\n\u003cp\u003eHEA \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Hip extension acceleration\u003c/p\u003e\n\u003cp\u003eHFA \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Hip flexion acceleration\u003c/p\u003e\n\u003cp\u003eKNN \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;K-Nearest Neighbors\u003c/p\u003e\n\u003cp\u003eSVM \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Support Vector Machines\u003c/p\u003e\n\u003cp\u003eANN \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Artificial Neural Networks\u003c/p\u003e\n\u003cp\u003eRF \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Random Forests\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eConflict of interest\u003c/h2\u003e \u003cp\u003eThe researchers confirm that there are no conflicts of interest.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eFinancial support for this study was not provided by any public or non-profit funding organizations.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eData Collection: Mahbubeh Keivan; Conceptualization: Mahbubeh Keivan, Mehrdad Anbarian; Methodology: Mahbubeh Keivan, Mehrdad Anbarian, Hassan Khotanlou; Software: Hassan Khotanlou; Formal Analysis: Hassan Khotanlou, Mahbubeh Keivan, Mehrdad Anbarian; Writing\u0026mdash;Review and Editing: Mahbubeh Keivan, Mehrdad Anbarian; Supervision: Mehrdad Anbarian\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThis research is extracted from PhD dissertation of Sports Biomechanics at Bu-Ali Sina University. The authors would like to thank all subjects who participated in this study.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe data used and analyzed during the study are available from the corresponding author upon reasonable request. The data are not publicly available due to privacy on ethical issues.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eNeal, B. S., Barton, C. J., Birn-Jeffery, A. \u0026amp; Morrissey, D. Increased hip adduction during running is associated with patellofemoral pain and differs between males and females: A case-control study. \u003cem\u003eJ. Biomech.\u003c/em\u003e \u003cb\u003e91\u003c/b\u003e, 133\u0026ndash;139 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrossley, K. M. et al. Patellofemoral pain consensus statement from the 4th international patellofemoral pain. \u003cem\u003eBr. J. Sports Med.\u003c/em\u003e \u003cb\u003e50\u003c/b\u003e (14), 839\u0026ndash;843 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDutton, R. A., Khadavi, M. J. \u0026amp; Fredericson, M. Patellofemoral pain. \u003cem\u003ePhys. Med. Rehabil Clin. N Am.\u003c/em\u003e \u003cb\u003e27\u003c/b\u003e (1), 31\u0026ndash;52. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.pmr.2015.08.008\u003c/span\u003e\u003cspan address=\"10.1016/j.pmr.2015.08.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetersen, W. et al. Patellofemoral pain syndrome. \u003cem\u003eKnee Surg. Sports Traumatol. Arthrosc.\u003c/em\u003e \u003cb\u003e22\u003c/b\u003e (10), 2264\u0026ndash;2274. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00167-014-2863-1\u003c/span\u003e\u003cspan address=\"10.1007/s00167-014-2863-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmith, B. E. et al. Incidence and prevalence of patellofemoral pain: A systematic review and meta-analysis. \u003cem\u003ePLoS One\u003c/em\u003e. \u003cb\u003e13\u003c/b\u003e (1), e0190892. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0190892\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0190892\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu, H. et al. Effects of neuromuscular training on pain intensity and self-reported functionality for patellofemoral pain syndrome in runners: Study protocol for a randomized controlled clinical trial. \u003cem\u003eTrials\u003c/em\u003e \u003cb\u003e20\u003c/b\u003e (1), 409. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s13063-019-3454-2\u003c/span\u003e\u003cspan address=\"10.1186/s13063-019-3454-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetersen, W. et al. Patellofemoral pain syndrome. \u003cem\u003eKnee Surg. Sports Traumatol. Arthrosc.\u003c/em\u003e \u003cb\u003e22\u003c/b\u003e (10), 2264\u0026ndash;2274. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00167-013-2759-6\u003c/span\u003e\u003cspan address=\"10.1007/s00167-013-2759-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNunes, G. S. et al. Clinical test for diagnosis of patellofemoral pain syndrome: Systematic review with meta-analysis. \u003cem\u003ePhysiotherapy Sport\u003c/em\u003e. \u003cb\u003e14\u003c/b\u003e (1), 54\u0026ndash;59. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ptsp.2012.07.003\u003c/span\u003e\u003cspan address=\"10.1016/j.ptsp.2012.07.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCook, C. et al. Best tests/clinical findings for screening and diagnosis of patellofemoral pain syndrome: A systematic review. \u003cem\u003ePhysiotherapy\u003c/em\u003e \u003cb\u003e98\u003c/b\u003e (2), 93\u0026ndash;100 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiang, L. et al. Recent machine learning progress in lower limb running biomechanics with wearable technology: A systematic review. \u003cem\u003eFront. Neurorobot\u003c/em\u003e. \u003cb\u003e16\u003c/b\u003e, 913052. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fnbot.2022.913052\u003c/span\u003e\u003cspan address=\"10.3389/fnbot.2022.913052\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMason, R. et al. Wearables for running gait analysis: A systematic review. \u003cem\u003eSports Med.\u003c/em\u003e \u003cb\u003e53\u003c/b\u003e (1), 241\u0026ndash;268. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40279-022-01760-6\u003c/span\u003e\u003cspan address=\"10.1007/s40279-022-01760-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDelgado-Garc\u0026iacute;a, G. et al.. IMU gyroscopes are a valid alternative to 3D optical motion capture system for angular kinematics analysis in tennis. \u003cem\u003eProc. Inst. Mech. Eng. P J. Sports Eng. Technol.\u003c/em\u003e \u003cb\u003e235\u003c/b\u003e (1), 3\u0026ndash;12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1754337120965444\u003c/span\u003e\u003cspan address=\"10.1177/1754337120965444\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohansen, S. K. et al. Exploring patients' and physiotherapists' visions on modelling treatments and optimising self-management strategies for patellofemoral pain: A future workshop approach. \u003cem\u003eMusculoskelet. Sci. Pract.\u003c/em\u003e \u003cb\u003e60\u003c/b\u003e, 102567. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.msksp.2022.102567\u003c/span\u003e\u003cspan address=\"10.1016/j.msksp.2022.102567\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNoehren, B., Pohl, M. B., Sanchez, Z., Cunningham, T. \u0026amp; Lattermann, C. Proximal and distal kinematics in female runners with patellofemoral pain. \u003cem\u003eClin. Biomech.\u003c/em\u003e \u003cb\u003e26\u003c/b\u003e (10), 1049\u0026ndash;1055. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.clinbiomech.2011.10.005\u003c/span\u003e\u003cspan address=\"10.1016/j.clinbiomech.2011.10.005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLuedke, L. E., Heiderscheit, B. C., Williams, D. S. \u0026amp; Rauh, M. J. Association of isometric strength of hip and knee muscles with injury risk in high school cross-country runners. \u003cem\u003eJ. Orthop. Sports Phys. Ther.\u003c/em\u003e \u003cb\u003e46\u003c/b\u003e (8), 667\u0026ndash;676. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2519/jospt.2016.6405\u003c/span\u003e\u003cspan address=\"10.2519/jospt.2016.6405\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNoehren, B., Hamill, J. \u0026amp; Davis, I. Prospective evidence for a hip etiology in patellofemoral pain. \u003cem\u003eMed. Sci. Sports Exerc.\u003c/em\u003e \u003cb\u003e45\u003c/b\u003e (6), 1120\u0026ndash;1124. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1249/MSS.0b013e31828249d2\u003c/span\u003e\u003cspan address=\"10.1249/MSS.0b013e31828249d2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSouza, R. B. \u0026amp; Powers, C. M. Differences in hip kinematics, muscle strength, and muscle activation between subjects with and without patellofemoral pain. \u003cem\u003eJ. Orthop. Sports Phys. Ther.\u003c/em\u003e \u003cb\u003e39\u003c/b\u003e (1), 12\u0026ndash;19. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2519/jospt.2009.2885\u003c/span\u003e\u003cspan address=\"10.2519/jospt.2009.2885\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilly, R. W. \u0026amp; Davis, I. S. The effect of a hip-strengthening program on mechanics during running and during a single-leg squat. \u003cem\u003eJ. Orthop. Sports Phys. Ther.\u003c/em\u003e \u003cb\u003e41\u003c/b\u003e (9), 625\u0026ndash;632. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2519/jospt.2011.3470\u003c/span\u003e\u003cspan address=\"10.2519/jospt.2011.3470\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeckey, C. et al. Machine learning approaches to injury risk prediction in sport: A scoping review with evidence synthesis. \u003cem\u003eBr. J. Sports Med.\u003c/em\u003e \u003cb\u003e59\u003c/b\u003e (7), 491\u0026ndash;500. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1136/bjsports-2024-108576\u003c/span\u003e\u003cspan address=\"10.1136/bjsports-2024-108576\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHalilaj, E. et al. Machine learning in human movement biomechanics: Best practices, common pitfalls, and new opportunities. \u003cem\u003eJ. Biomech.\u003c/em\u003e \u003cb\u003e81\u003c/b\u003e, 1\u0026ndash;11. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jbiomech.2018.09.009\u003c/span\u003e\u003cspan address=\"10.1016/j.jbiomech.2018.09.009\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohansen, S. K. et al. Exploring patients' and physiotherapists' visions on modelling treatments and optimising self-management strategies for patellofemoral pain: A future workshop approach. \u003cem\u003eMusculoskelet. Sci. Pract.\u003c/em\u003e \u003cb\u003e60\u003c/b\u003e, 102567. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.msksp.2022.102567\u003c/span\u003e\u003cspan address=\"10.1016/j.msksp.2022.102567\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeidarimoghaddam, R., Hosseini, M. H., Ilbeigi, S., Anbarian, M. \u0026amp; Tapak, L. Kinematic analysis of head and trunk in individual and team one-handed carrying. \u003cem\u003eInt. J. Ind. Ergon.\u003c/em\u003e \u003cb\u003e94\u003c/b\u003e, 103422. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ergon.2023.103422\u003c/span\u003e\u003cspan address=\"10.1016/j.ergon.2023.103422\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu, H. Value evaluation of knee joint sports injury detection model-aided diagnosis based on machine learning. \u003cem\u003eFront. Phys.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e, 1166275. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fphy.2023.1166275\u003c/span\u003e\u003cspan address=\"10.3389/fphy.2023.1166275\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTayfur, B., Charuphongsa, C., Morrissey, D. \u0026amp; Miller, S. C. Neuromuscular function of the knee joint following knee injuries: Does it ever get back to normal? A systematic review with meta-analyses. \u003cem\u003eSports Med.\u003c/em\u003e \u003cb\u003e51\u003c/b\u003e (2), 321\u0026ndash;338. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40279-020-01386-6\u003c/span\u003e\u003cspan address=\"10.1007/s40279-020-01386-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIacobescu, P., Marina, V., Anghel, C. \u0026amp; Anghele, A. D. Evaluating binary classifiers for cardiovascular disease prediction: Enhancing early diagnostic capabilities. \u003cem\u003eJ. Cardiovasc. Dev. Dis.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e (12), 396. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/jcdd11120396\u003c/span\u003e\u003cspan address=\"10.3390/jcdd11120396\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhairuddin, M. Z. F. et al. Occupational injury risk mitigation: Machine learning approach and feature optimization for smart workplace surveillance. \u003cem\u003eInt. J. Environ. Res. Public. Health\u003c/em\u003e. \u003cb\u003e19\u003c/b\u003e (21), 13962. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/ijerph192113962\u003c/span\u003e\u003cspan address=\"10.3390/ijerph192113962\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRay, P., Reddy, S. S. \u0026amp; Banerjee, T. Various dimension reduction techniques for high dimensional data analysis: A review. \u003cem\u003eArtif. Intell. Rev.\u003c/em\u003e \u003cb\u003e54\u003c/b\u003e (3), 3473\u0026ndash;3485. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10462-020-09928-0\u003c/span\u003e\u003cspan address=\"10.1007/s10462-020-09928-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaghu, S. \u0026amp; Sriraam, N. Classification of focal and non-focal EEG signals using neighborhood component analysis and machine learning algorithms. \u003cem\u003eExpert Syst. Appl.\u003c/em\u003e \u003cb\u003e113\u003c/b\u003e, 18\u0026ndash;32. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.eswa.2018.06.042\u003c/span\u003e\u003cspan address=\"10.1016/j.eswa.2018.06.042\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMohammadi Moghadam Sh, Yeung, T. \u0026amp; Choisne, J. A comparison of machine learning models\u0026rsquo; accuracy in predicting lower-limb joints\u0026rsquo; kinematics, kinetics, and muscle forces from wearable sensors. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e, 5046. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-023-31906-z\u003c/span\u003e\u003cspan address=\"10.1038/s41598-023-31906-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDierks, T. A., Manal, K. T., Hamill, J. \u0026amp; Davis, I. S. Proximal and distal influences on hip and knee kinematics in runners with patellofemoral pain during a prolonged run. \u003cem\u003eJ. Orthop. Sports Phys. Ther.\u003c/em\u003e \u003cb\u003e38\u003c/b\u003e (8), 448\u0026ndash;456. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2519/jospt.2008.2490\u003c/span\u003e\u003cspan address=\"10.2519/jospt.2008.2490\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eM\u0026uuml;ndermann, A., Asay, J. L., M\u0026uuml;ndermann, L. \u0026amp; Andriacchi, T. P. Implications of increased medio-lateral trunk sway for ambulatory mechanics. \u003cem\u003eJ. Biomech.\u003c/em\u003e \u003cb\u003e41\u003c/b\u003e (1), 165\u0026ndash;170. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jbiomech.2007.09.020\u003c/span\u003e\u003cspan address=\"10.1016/j.jbiomech.2007.09.020\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin, C. C. et al. Machine learning-enhanced prediction of running-related injuries. \u003cem\u003ePLoS One\u003c/em\u003e. \u003cb\u003e18\u003c/b\u003e (2), e0281427. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0281427\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0281427\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMalinauskas, R. A., Ogami, T., Witzke, K. A. \u0026amp; Doyle, W. A. Development of a methodology for computing patellofemoral joint mechanics. \u003cem\u003eJ. Biomech. Eng.\u003c/em\u003e \u003cb\u003e142\u003c/b\u003e (4), 041010. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1115/1.4045753\u003c/span\u003e\u003cspan address=\"10.1115/1.4045753\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLavagnino, M. et al. Patellofemoral loading and knee flexion during squatting. \u003cem\u003eJ. Orthop. Res.\u003c/em\u003e \u003cb\u003e26\u003c/b\u003e (8), 1084\u0026ndash;1090. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/jor.20621\u003c/span\u003e\u003cspan address=\"10.1002/jor.20621\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDonnelly, C. J., Elliott, B. C., Doyle, T. L., Finch, C. F. \u0026amp; Dempsey, A. R. Lloyd D.G. Changes in knee joint biomechanics following balance and technique training and a season of Australian football. \u003cem\u003eBr. J. Sports Med.\u003c/em\u003e \u003cb\u003e46\u003c/b\u003e (13), 917\u0026ndash;922. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1136/bjsports-2012-090936\u003c/span\u003e\u003cspan address=\"10.1136/bjsports-2012-090936\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2012).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Hip Joint, Patellofemoral Pain Syndrome, Running, Machine Learning","lastPublishedDoi":"10.21203/rs.3.rs-8928002/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8928002/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAs a prevalent knee joint disorder, Patellofemoral Pain Syndrome (PFPS) is identified by musculoskeletal issues, often involving symptoms such as pain experienced around or posterior to the patella. Hip joint kinematics may play an essential role in PFPS. Although this condition has been studied from various perspectives, there is no definitive standard clinical method for diagnosis and functional classification. This study aimed to identify the most significant hip joint kinematic features for classifying PFPS patients during running using machine learning (ML) methods. Seven females with unilateral PFPS were paired with controls considering factors such as age, weight, height, and duration of physical activity. In total, 560 running cycles were captured utilizing a motion analysis system based on an inertial measurement unit (IMU). Hip joint kinematic variables, including three-dimensional angles, velocity, and acceleration, were measured. Nineteen features were used as inputs for ML algorithms. Four models\u0026mdash;SVM, KNN, ANN, and RF\u0026mdash;achieved 99% accuracy in classifying healthy and PFPS patients. Maximum hip adduction emerged as the most significant kinematic feature, and the SVM model performed best for PFPS classification. In conclusion, this study demonstrates that combining IMU sensors with machine learning techniques provides an accurate approach for diagnosing PFPS during running in non-laboratory and clinical environments. Moreover, frontal plane hip joint kinematics appears to be a critical factor in identifying this condition.\u003c/p\u003e","manuscriptTitle":"Classification of patients with patellofemoral pain syndrome during running based on hip joint kinematic feature using machine learning method","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-09 07:05:04","doi":"10.21203/rs.3.rs-8928002/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"b7aab285-22d2-4cb5-bbea-4a6cc107befe","owner":[],"postedDate":"March 9th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":64009479,"name":"Health sciences/Diseases"},{"id":64009480,"name":"Physical sciences/Engineering"},{"id":64009481,"name":"Health sciences/Health care"},{"id":64009482,"name":"Health sciences/Medical research"}],"tags":[],"updatedAt":"2026-03-09T09:43:20+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-09 07:05:04","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8928002","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8928002","identity":"rs-8928002","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}