Predictive Modeling of Hamstring Strain Injuries in Elite Australian Footballers

Machine learning approaches to injury risk prediction in sport: a scoping review with evidence synthesis

OBJECTIVE

This study reviewed the current state of machine learning (ML) research for the prediction of sports-related injuries. It aimed to chart the various approaches used and assess their efficacy, considering factors such as data heterogeneity, model specificity and contextual factors when developing predictive models.

DESIGN

Scoping review.

DATA SOURCES

PubMed, EMBASE, SportDiscus and IEEEXplore.

RESULTS

In total, 1241 studies were identified, 58 full texts were screened, and 38 relevant studies were reviewed and charted. Football (soccer) was the most commonly investigated sport. Area under the curve (AUC) was the most common means of model evaluation; it was reported in 71% of studies. In 60% of studies, tree-based solutions provided the highest statistical predictive performance. Random Forest and Extreme Gradient Boosting (XGBoost) were found to provide the highest performance for injury risk prediction. Logistic regression outperformed ML methods in 4 out of 12 studies. Three studies reported model performance of AUC>0.9, yet the clinical relevance is questionable.

CONCLUSIONS

A variety of different ML models have been applied to the prediction of sports-related injuries. While several studies report strong predictive performance, their clinical utility can be limited, with wide prediction windows or broad definitions of injury. The efficacy of ML is hampered by small datasets and numerous methodological heterogeneities (cohort sizes, definition of injury and dependent variables), which were common across the reviewed studies.

An Educational Review on Machine Learning: A SWOT Analysis for Implementing Machine Learning Techniques in Football

PURPOSE

The abundance of data in football presents both opportunities and challenges for decision making. Consequently, this review has 2 primary objectives: first, to provide practitioners with a concise overview of the characteristics of machine-learning (ML) analysis, and, second, to conduct a strengths, weaknesses, opportunities, and threats (SWOT) analysis regarding the implementation of ML techniques in professional football clubs. This review explains the difference between artificial intelligence and ML and the difference between ML and statistical analysis. Moreover, we summarize and explain the characteristics of ML learning approaches, such as supervised learning, unsupervised learning, and reinforcement learning. Finally, we present an example of a SWOT analysis that suggests some actions to be considered in applying ML techniques by medical and sport science staff working in football. Specifically, 4 dimensions are presented: the use of strengths to create opportunities and make the most of them, the use of strengths to avoid threats, working on weaknesses to take advantage of opportunities, and upgrading weaknesses to avoid threats.

CONCLUSION

ML analysis can be an invaluable tool for football clubs and sport-science and medical departments due to its ability to analyze vast amounts of data and extract meaningful insights. Moreover, ML can enhance performance by assessing the risk of injury, physiological parameters, and physical fitness, as well as optimizing training, recommending strategies based on opponent analysis, and identifying talent and assessing player suitability.

Prediction of football injuries using GPS-based data in Iranian professional football players: a machine learning approach

Introduction

The study aims to assess and compare the predictive effectiveness of football-related injuries using external load data and a decision tree classification algorithm by unidimensional approach.

Methods

The sample consisted of 25 players from one of the 16 teams participating in the Persian Gulf Pro League during the 2022--2023 season. Player injury data and raw GPS data from all training and competition sessions throughout the football league season were gathered (214 training sessions and 34 competition sessions). The acute-tochronic workload ratio was calculated separately for each variable using a ratio of 1:3 weeks. Finally, the decision tree algorithm with machine learning was utilised to assess the predictive power of injury occurrence based on the acute-to-chronic workload ratio.

Results

The results showed that the variable of the number of decelerations had the highest predictive power compared to other variables [area under the curve (AUC) = 0.91, recall = 87.5%, precision = 58.3%, accuracy = 94.7%].

Conclusion

Although none of the selected external load variables in this study had high predictive power (AUC > 0.95), due to the high predictive power of injury of the number of deceleration variables compared with other variables, the necessity of attention and management of this variable as a risk factor for injury occurrence is essential for preventing future injuries.

Predicting noncontact injuries of professional football players using machine learning

Noncontact injuries are prevalent among professional football players. Yet, most research on this topic is retrospective, focusing solely on statistical correlations between Global Positioning System (GPS) metrics and injury occurrence, overlooking the multifactorial nature of injuries. This study introduces an automated injury identification and prediction approach using machine learning, leveraging GPS data and player-specific parameters. A sample of 34 male professional players from a Portuguese first-division team was analyzed, combining GPS data from Catapult receivers with descriptive variables for machine learning models-Support Vector Machines (SVMs), Feedforward Neural Networks (FNNs), and Adaptive Boosting (AdaBoost)-to predict injuries. These models, particularly the SVMs with cost-sensitive learning, showed high accuracy in detecting injury events, achieving a sensitivity of 71.43%, specificity of 74.19%, and overall accuracy of 74.22%. Key predictive factors included the player's position, session type, player load, velocity and acceleration. The developed models are notable for their balanced sensitivity and specificity, efficiency without extensive manual data collection, and capability to predict injuries for short time frames. These advancements will aid coaching staff in identifying high-risk players, optimizing team performance, and reducing rehabilitation costs.

Non-contact lower limb injuries in Rugby Union: A two-year pattern recognition analysis of injury risk factors

The cause of sport injuries are multifactorial and necessitate sophisticated statistical approaches for accurate identification of risk factors predisposing athletes to injury. Pattern recognition analyses have been adopted across sporting disciplines due to their ability to account for repeated measures and non-linear interactions of datasets, however there are limited examples of their use in injury risk prediction. This study incorporated two-years of rigorous monitoring of athletes with 1740 individual weekly data points across domains of training load, performance testing, musculoskeletal screening, and injury history parameters, to be one of the first to employ a pattern recognition approach to predict the risk factors of specific non-contact lower limb injuries in Rugby Union. Predictive models (injured vs. non-injured) were generated for non-contact lower limb, non-contact ankle, and severe non-contact injuries using Bayesian pattern recognition from a pool of 36 Senior Academy Rugby Union athletes. Predictors for non-contact lower limb injuries included dorsiflexion angle, adductor strength, and previous injury history (area under the receiver operating characteristic (ROC) = 0.70) Dorsiflexion angle parameters were also predictive of non-contact ankle injuries, along with slower sprint times, greater body mass, previous concussion, and previous ankle injury (ROC = 0.76). Predictors of severe non-contact lower limb injuries included greater differences in mean training load, slower sprint times, reduced hamstring and adductor strength, reduced dorsiflexion angle, greater perceived muscle soreness, and playing as a forward (ROC = 0.72). The identification of specific injury risk factors and useable thresholds for non-contact injury risk detection in sport holds great potential for coaches and medical staff to modify training prescriptions and inform injury prevention strategies, ultimately increasing player availability, a key indicator of team success.

Machine learning in knee injury sequelae detection: Unravelling the role of psychological factors and preventing long-term sequelae

Purpose

This study evaluated the performance of three machine learning (ML) algorithms-decision tree (DT), multilayer perceptron (MLP) and extreme gradient boosting (XGB)-in identifying regular athletes who suffered a knee injury several months to years prior. In addition, the contribution of psychological variables in addition to biomechanical ones in the classification performance of the ML algorithms was assessed, to better identify factors to get back to competitive sport with the lowest possible risk of new knee injury.

Methods

A cohort of 96 athletes, 36 with prior knee injuries, practicing an average of 5.7 ± 2.4 h per week, participated in a horizontal force-velocity test on a ballistic ergometer providing data of force, velocity and power from each lower limb. They also completed a psychological questionnaire, which included components from the Knee Injury and Osteoarthritis Outcome Score (KOOS) and the Sport Anxiety Scale (SAS). The three ML algorithms were trained on a thousand different train-test sets. Also, Shapley values were calculated for each input variable of a data set to highlight its contribution to the prediction from an ML model.

Results

Over a thousand cross-validations, higher area under the curve (AUC) values were obtained when accounted for the psychological attributes (p < 0.001). Also, higher AUC values were obtained from MLP compared to XGB or DT (p < 0.001). XGB exhibited higher AUC values than DT (p < 0.001).

Conclusions

Our results suggested that psychological factors play a more important role in recognition than biomechanical factors, with KOOS and SAS scores ranking high in the list of influential factors. Additionally, the computing stability of MLP could be recommended for classification tasks in the context of knee injuries.

Level of Evidence

Level III.

Artificial intelligence applications in the football codes: A systematic review

Artificial Intelligence (AI) is increasingly being adopted across many domains such as transport, healthcare, defence and sport, with football codes no exception. Though there is a range of potential benefits of AI, concern has also been expressed regarding potential risks. An important first step in ensuring that AI applications in football are usable, beneficial, safe and ethical is to understand the current range of applications, the AI models adopted and their proposed functions. This systematic review aimed to identify different applications of AI across football codes to synthesise current knowledge and determine whether potential risks are being considered. The systematic review included 190 peer-reviewed articles. Nine areas of application were found ranging from athlete evaluation and event detection to match outcome prediction and injury detection and prediction. In total, 27 different AI models were identified, with artificial neural networks the most frequently applied. Five AI assessment metrics were identified including specificity, recall, precision, accuracy and F1-score. Four potential risks were identified, concerning data security, usability, data biases and inappropriate athlete load management. It is concluded that, though a wide range of AI applications currently exist, further work is required to develop AI for football and identify and manage potential risks.

Hamstrings on focus: Are 72 hours sufficient for recovery after a football (soccer) match? A multidisciplinary approach based on hamstring injury risk factors and histology

This study aimed to assess acute and residual changes in sprint-related hamstring injury (HSI) risk factors after a football (soccer) match, focusing on recovery within the commonly observed 72-h timeframe between elite football matches. We used a multifactorial approach within a football context, incorporating optical and ultrastructural microscopic analysis of BFlh (biceps femoris long head) muscle fibres, along with an examination of BFlh fibre composition. Changes in sprint performance-related factors and HSI modifiable risk factors were examined until 3 days after the match (MD +3) in 20 football players. BFlh biopsy specimens were obtained before and at MD +3 in 10 players. The findings indicated that at MD +3, sprint-related performance and HSI risk factors had not fully recovered, with notable increases in localized BFlh fibre disruptions. Interestingly, match load (both external and internal) did not correlate with changes in sprint performance or HSI risk factors nor with BFlh fibre disruption. Furthermore, our study revealed a balanced distribution of ATPase-based fibre types in BFlh, with type-II fibres associated with sprint performance. Overall, the results suggest that a 72-h recovery period may not be adequate for hamstring muscles in terms of both HSI risk factors and BFlh fibre structure following a football match.

The Trade Secret Taboo: Open Science Methods are Required to Improve Prediction Models in Sports Medicine and Performance

Clinical prediction models in sports medicine that utilize regression or machine learning techniques have become more widely published, used, and disseminated. However, these models are typically characterized by poor methodology and incomplete reporting, and an inadequate evaluation of performance, leading to unreliable predictions and weak clinical utility within their intended sport population. Before implementation in practice, models require a thorough evaluation. Strong replicable methods and transparency reporting allow practitioners and researchers to make independent judgments as to the model's validity, performance, clinical usefulness, and confidence it will do no harm. However, this is not reflected in the sports medicine literature. As shown in a recent systematic review of models for predicting sports injury models, most were typically characterized by poor methodology, incomplete reporting, and inadequate performance evaluation. Because of constraints imposed by data from individual teams, the development of accurate, reliable, and useful models is highly reliant on external validation. However, a barrier to collaboration is a desire to maintain a competitive advantage; a team's proprietary information is often perceived as high value, and so these 'trade secrets' are frequently guarded. These 'trade secrets' also apply to commercially available models, as developers are unwilling to share proprietary (and potentially profitable) development and validation information. In this Current Opinion, we: (1) argue that open science is essential for improving sport prediction models and (2) critically examine sport prediction models for open science practices.

Predicting Injury and Illness with Machine Learning in Elite Youth Soccer: A Comprehensive Monitoring Approach over 3 Months

The search for monitoring tools that provide early indication of injury and illness could contribute to better player protection. The aim of the present study was to i) determine the feasibility of and adherence to our monitoring approach, and ii) identify variables associated with up-coming illness and injury. We incorporated a comprehensive set of monitoring tools consisting of external load and physical fitness data, questionnaires, blood, neuromuscular-, hamstring, hip abductor and hip adductor performance tests performed over a three-month period in elite under-18 academy soccer players. Twenty-five players (age: 16.6 ± 0.9 years, height: 178 ± 7 cm, weight: 74 ± 7 kg, VO2max: 59 ± 4 ml/min/kg) took part in the study. In addition to evaluating adherence to the monitoring approach, data were analyzed using a linear support vector machine (SVM) to predict illness and injuries. The approach was feasible, with no injuries or dropouts due to the monitoring process. Questionnaire adherence was high at the beginning and decreased steadily towards the end of the study. An SVM resulted in the best classification results for three classification tasks, i.e., illness prediction, illness determination and injury prediction. For injury prediction, one of four injuries present in the test data set was detected, with 96.3% of all data points (i.e., injuries and non-injuries) correctly detected. For both illness prediction and determination, there was only one illness in the test data set that was detected by the linear SVM. However, the model showed low precision for injury and illness prediction with a considerable number of false-positives. The results demonstrate the feasibility of a holistic monitoring approach with the possibility of predicting illness and injury. Additional data points are needed to improve the prediction models. In practical application, this may lead to overcautious recommendations on when players should be protected from injury and illness.

An Overview of Machine Learning Applications in Sports Injury Prediction

Use injuries, i.e., injuries caused by repetitive strain on the body, represent a serious problem in athletics that has traditionally relied on historic datasets and human experience for prevention. Existing methodologies have been frustratingly slow at developing higher precision prevention practices. Technological advancements have permitted the emergence of artificial intelligence and machine learning (ML) as promising toolsets to enhance both injury mitigation and rehabilitation protocols. This article provides a comprehensive overview of recent advances in ML techniques as they have been applied to sports injury prediction and prevention. A comprehensive literature review was conducted searching PubMed/Medline, Institute of Electrical and Electronics Engineers (IEEE)/Institute of Engineering and Technology (IET), and ScienceDirect. Ovid Discovery and Google Scholar were used to provide additional aggregate results and a grey literature search. A focus was placed on papers published from 2017 to 2022. Algorithms of interest were limited to K-Nearest Neighbor (KNN), K-means, decision tree, random forest, gradient boosting and AdaBoost, and neural networks. A total of 42 original research papers were included, and their results were summarized. We conclude that given the current lack of open source, uniform data sets, as well as a reliance on dated regression models, no strong conclusions about the real-world efficacy of ML as it applies to sports injury prediction can be made. However, it is suggested that addressing these two issues will allow powerful, novel ML architectures to be deployed, thus rapidly advancing the state of this field, and providing validated clinical tools.

Common High-Speed Running Thresholds Likely Do Not Correspond to High-Speed Running in Field Sports

ABSTRACT

Freeman, BW, Talpey, SW, James, LP, Opar, DA, and Young, WB. Common high-speed running thresholds likely do not correspond to high-speed running in field sports. J Strength Cond Res 37(7): 1411-1418, 2023-The purpose of this study was to clarify what percentage of maximum speed is associated with various running gaits. Fifteen amateur field sport athletes (age = 23 ± 3.6 years) participated in a series of 55-meter running trials. The speed of each trial was determined by instructions relating to 5 previously identified gait patterns (jog, run, stride, near maximum sprint, and sprint). Each trial was filmed in slow motion (240 fps), whereas running speed was obtained using Global Positioning Systems. Contact time, stride angle, and midstance free-leg knee angle were determined from video footage. Running gaits corresponded with the following running speeds, jogging = 4.51 m·s -1 , 56%Vmax, running = 5.41 m·s -1 , 66%Vmax , striding = 6.37 m·s -1 , 78%Vmax, near maximum sprinting = 7.08 m·s -1 , 87%Vmax, and sprinting = 8.15 m·s -1 , 100%Vmax. Significant ( p < 0.05) increases in stride angle were observed as running speed increased. Significant ( p < 0.05) decreases were observed in contact time and midstance free-leg knee angle as running speed increased. These findings suggest currently used thresholds for high-speed running (HSR) and sprinting most likely correspond with jogging and striding, which likely underestimates the true HSR demands. Therefore, a higher relative speed could be used to describe HSR and sprinting more accurately in field sports.

Relationship between a daily injury risk estimation feedback (I-REF) based on machine learning techniques and actual injury risk in athletics (track and field): protocol for a prospective cohort study over an athletics season

INTRODUCTION

Two-thirds of athletes (65%) have at least one injury complaint leading to participation restriction (ICPR) in athletics (track and field) during one season. The emerging practice of medicine and public health supported by electronic processes and communication in sports medicine represents an opportunity for developing new injury risk reduction strategies. Modelling and predicting the risk of injury in real-time through artificial intelligence using machine learning techniques might represent an innovative injury risk reduction strategy. Thus, the primary aim of this study will be to analyse the relationship between the level of Injury Risk Estimation Feedback (I-REF) use (average score of athletes' self-declared level of I-REF consideration for their athletics activity) and the ICPR burden during an athletics season.

METHOD AND ANALYSIS

We will conduct a prospective cohort study, called Injury Prediction with Artificial Intelligence (IPredict-AI), over one 38-week athletics season (from September 2022 to July 2023) involving competitive athletics athletes licensed with the French Federation of Athletics. All athletes will be asked to complete daily questionnaires on their athletics activity, their psychological state, their sleep, the level of I-REF use and any ICPR. I-REF will present a daily estimation of the ICPR risk ranging from 0% (no risk for injury) to 100% (maximal risk for injury) for the following day. All athletes will be free to see I-REF and to adapt their athletics activity according to I-REF. The primary outcome will be the ICPR burden over the follow-up (over an athletics season), defined as the number of days lost from training and/or competition due to ICPR per 1000 hours of athletics activity. The relationship between ICPR burden and the level of I-REF use will be explored by using linear regression models.

ETHICS AND DISSEMINATION

This prospective cohort study was reviewed and approved by the Saint-Etienne University Hospital Ethical Committee (Institutional Review Board: IORG0007394, IRBN1062022/CHUSTE). Results of the study will be disseminated in peer-reviewed journals and in international scientific congresses, as well as to the included participants.

The impact of sport-specific physical fitness change patterns on lower limb non-contact injury risk in youth female basketball players: a pilot study based on field testing and machine learning

Background: In recent years, identifying players with injury risk through physical fitness assessment has become a hot topic in sports science research. Although practitioners have conducted many studies on the relationship between physical fitness and the likelihood of injury, the relationship between the two remains indeterminate. Consequently, this study utilized machine learning to preliminary investigate the relationship between individual physical fitness tests and injury risk, aiming to identify whether patterns of physical fitness change have an impact on injury risk. Methods: This study conducted a retrospective analysis by extracting the records of 17 young female basketball players from the sport-specific physical fitness monitoring and injury registration database in Fujian Province. Sports-specific physical fitness tests included physical performance, physiological, biochemical, and subjective perceived responses. The data for each player was standardized individually using Z-scores. Synthetic minority over-sampling techniques and edited nearest neighbor algorithms were used to sample the training set to address the negative impact of class imbalance on model performance. Feature extraction was performed on the dataset using linear discriminant analysis, and the prediction model was constructed using the cost-sensitive neural network. Results: The 10 replicate 5-fold stratified cross-validation showed that the lower limb non-contact injury prediction model based on the cost-sensitive neural network had achieved good discrimination and calibration (average Precision: 0.6360; average Recall: 0.8700; average F2-Score: 0.7980; average AUC: 0.8590; average Brier-score: 0.1020), which could be well applied in training practice. According to the attribution analysis, agility and speed were important physical attributes that affect youth female basketball players' non-contact lower limb injury risk. Specifically, there was enhance in the performance of the 1-min double under, accompanied by an increase in urinary ketone and urinary blood levels following the agility test. The 3/4 basketball court sprint performance improved, while urinary protein and RPE levels decreased after the speed test. Conclusion: The sport-specific physical fitness change pattern can impact the lower limb non-contact injury risk of young female basketball players in Fujian Province, specifically in terms of agility and speed. These findings will provide valuable insights for planning athletes' physical training programs, managing fatigue, and preventing injuries.

Predictive utility of commercial grade technologies for assessing musculoskeletal injury risk in US Marine Corps Officer candidates

Recently, commercial grade technologies have provided black box algorithms potentially relating to musculoskeletal injury (MSKI) risk and functional movement deficits, in which may add value to a high-performance model. Thus, the purpose of this manuscript was to evaluate composite and component scores from commercial grade technologies associations to MSKI risk in Marine Officer Candidates. 689 candidates (Male candidates = 566, Female candidates = 123) performed counter movement jumps on SPARTA™ force plates and functional movements (squats, jumps, lunges) in DARI™ markerless motion capture at the start of Officer Candidates School (OCS). De-identified MSKI data was acquired from internal OCS reports for those who presented to the Physical Therapy department for MSKI treatment during the 10 weeks of training. Logistic regression analyses were conducted to validate the utility of the composite scores and supervised machine learning algorithms were deployed to create a population specific model on the normalized component variables in SPARTA™ and DARI™. Common MSKI risk factors (cMSKI) such as older age, slower run times, and females were associated with greater MSKI risk. Composite scores were significantly associated with MSKI, although the area under the curve (AUC) demonstrated poor discrimination (AUC = .55-.57). When supervised machine learning algorithms were trained on the normalized component variables and cMSKI variables, the overall training models performed well, but when the training models were tested on the testing data the models classified MSKI "by chance" (testing AUC avg = .55-.57) across all models. Composite scores and component population specific models were poor predictors of MSKI in candidates. While cMSKI, SPARTA™, and DARI™ models performed similarly, this study does not dismiss the use of commercial technologies but questions the utility of a singular screening task to predict MSKI over 10 weeks. Further investigations should evaluate occupation specific screening, serial measurements, and/or load exposure for creating MSKI risk models.

Just How Confident Can We Be in Predicting Sports Injuries? A Systematic Review of the Methodological Conduct and Performance of Existing Musculoskeletal Injury Prediction Models in Sport

BACKGROUND

An increasing number of musculoskeletal injury prediction models are being developed and implemented in sports medicine. Prediction model quality needs to be evaluated so clinicians can be informed of their potential usefulness.

OBJECTIVE

To evaluate the methodological conduct and completeness of reporting of musculoskeletal injury prediction models in sport.

METHODS

A systematic review was performed from inception to June 2021. Studies were included if they: (1) predicted sport injury; (2) used regression, machine learning, or deep learning models; (3) were written in English; (4) were peer reviewed.

RESULTS

Thirty studies (204 models) were included; 60% of studies utilized only regression methods, 13% only machine learning, and 27% both regression and machine learning approaches. All studies developed a prediction model and no studies externally validated a prediction model. Two percent of models (7% of studies) were low risk of bias and 98% of models (93% of studies) were high or unclear risk of bias. Three studies (10%) performed an a priori sample size calculation; 14 (47%) performed internal validation. Nineteen studies (63%) reported discrimination and two (7%) reported calibration. Four studies (13%) reported model equations for statistical predictions and no machine learning studies reported code or hyperparameters.

CONCLUSION

Existing sport musculoskeletal injury prediction models were poorly developed and have a high risk of bias. No models could be recommended for use in practice. The majority of models were developed with small sample sizes, had inadequate assessment of model performance, and were poorly reported. To create clinically useful sports musculoskeletal injury prediction models, considerable improvements in methodology and reporting are urgently required.

A novel lower extremity non-contact injury risk prediction model based on multimodal fusion and interpretable machine learning

The application of machine learning algorithms in studying injury assessment methods based on data analysis has recently provided a new research insight for sports injury prevention. However, the data used in these studies are primarily multi-source and multimodal (i.e., longitudinal repeated-measures data and cross-sectional data), resulting in the models not fully utilising the information in the data to reveal specific injury risk patterns. Therefore, this study proposed an injury risk prediction model based on a multi-modal strategy and machine learning algorithms to handle multi-source data better and predict injury risk. This study retrospectively analysed the routine monitoring data of sixteen young female basketball players. These data included training load, perceived well-being status, physiological response, physical performance and lower extremity non-contact injury registration. This study partitions the original dataset based on the frequency of data collection. Extreme gradient boosting (XGBoost) was used to construct unimodal submodels to obtain decision scores for each category of indicators. Ultimately, the decision scores from each submodel were fused using the random forest (RF) to generate a lower extremity non-contact injury risk prediction model at the decision-level. The 10-fold cross-validation results showed that the fusion model was effective in classifying non-injured (mean Precision: 0.9932, mean Recall: 0.9976, mean F2-score: 0.9967), minimal lower extremity non-contact injuries risk (mean Precision: 0.9317, mean Recall: 0.9167, mean F2-score: 0.9171), and mild lower extremity non-contact injuries risk (mean Precision: 0.9000, mean Recall: 0.9000, mean F2-score: 0.9000). The model performed significantly more optimal than the submodel. Comparing the fusion model proposed with a traditional data integration scheme, the average Precision and Recall improved by 8.2 and 20.3%, respectively. The decision curves analysis showed that the proposed fusion model provided a higher net benefit to athletes with potential lower extremity non-contact injury risk. The validity, feasibility and practicality of the proposed model have been confirmed. In addition, the shapley additive explanation (SHAP) and network visualisation revealed differences in lower extremity non-contact injury risk patterns across severity levels. The model proposed in this study provided a fresh perspective on injury prevention in future research.

Predicting ACL Injury Using Machine Learning on Data From an Extensive Screening Test Battery of 880 Female Elite Athletes

BACKGROUND

Injury risk prediction is an emerging field in which more research is needed to recognize the best practices for accurate injury risk assessment. Important issues related to predictive machine learning need to be considered, for example, to avoid overinterpreting the observed prediction performance.

PURPOSE

To carefully investigate the predictive potential of multiple predictive machine learning methods on a large set of risk factor data for anterior cruciate ligament (ACL) injury; the proposed approach takes into account the effect of chance and random variations in prediction performance.

STUDY DESIGN

Case-control study; Level of evidence, 3.

METHODS

The authors used 3-dimensional motion analysis and physical data collected from 791 female elite handball and soccer players. Four common classifiers were used to predict ACL injuries (n = 60). Area under the receiver operating characteristic curve (AUC-ROC) averaged across 100 cross-validation runs (mean AUC-ROC) was used as a performance metric. Results were confirmed with repeated permutation tests (paired Wilcoxon signed-rank-test; P < .05). Additionally, the effect of the most common class imbalance handling techniques was evaluated.

RESULTS

For the best classifier (linear support vector machine), the mean AUC-ROC was 0.63. Regardless of the classifier, the results were significantly better than chance, confirming the predictive ability of the data and methods used. AUC-ROC values varied substantially across repetitions and methods (0.51-0.69). Class imbalance handling did not improve the results.

CONCLUSION

The authors' approach and data showed statistically significant predictive ability, indicating that there exists information in this prospective data set that may be valuable for understanding injury causation. However, the predictive ability remained low from the perspective of clinical assessment, suggesting that included variables cannot be used for ACL prediction in practice.

Including Modifiable and Nonmodifiable Factors Improves Injury Risk Assessment in Professional Baseball Pitchers

OBJECTIVES

To (1) evaluate an injury risk model that included modifiable and nonmodifiable factors into an arm injury risk prediction model in Minor League Baseball (MiLB) pitchers and (2) compare model performance separately for predicting the incidence of elbow and shoulder injuries.

DESIGN

Prospective cohort.

METHODS

A 10-year MiLB injury risk study was conducted. Pitchers were evaluated during preseason, and pitches and arm injuries were documented prospectively. Nonmodifiable variables included arm injury history, professional experience, arm dominance, year, and humeral torsion. Modifiable variables included BMI, pitch count, total range of motion, and horizontal adduction. We compared modifiable, nonmodifiable, and combined model performance by R2, calibration (best = 1.00), and discrimination (area under the curve [AUC]; higher number is better). Sensitivity analysis included only arm injuries sustained in the first 90 days.

RESULTS

In this study, 407 MiLB pitchers (141 arm injuries) were included. Arm injury incidence was 0.27 injuries per 1000 pitches. The arm injury model (calibration 1.05 [0.81-1.30]; AUC: 0.74 [0.69-0.80]) had improved performance compared to only using modifiable predictors (calibration: 0.91 [0.68-1.14]; AUC: 0.67 [0.62-0.73]) and only shoulder range of motion (calibration: 0.52 [0.29, 0.75]; AUC: 0.52 [0.46, 58]). Elbow injury model demonstrated improved performance (calibration: 1.03 [0.76-1.33]; AUC: 0.76 [0.69-0.83]) compared to the shoulder injury model (calibration: 0.46 [0.22-0.69]; AUC: 0.62 [95% CI: 0.55, 0.69]). The sensitivity analysis demonstrated improved model performance compared to the arm injury model.

CONCLUSION

Arm injury risk is influenced by modifiable and nonmodifiable risk factors. The most accurate way to identify professional pitchers who are at risk for arm injury is to use a model that includes modifiable and nonmodifiable risk factors. J Orthop Sports Phys Ther 2022;52(9):630-640. Epub: 9 July 2022. doi:10.2519/jospt.2022.11072.

Preseason Eccentric Strength Is Not Associated with Hamstring Strain Injury: A Prospective Study in Collegiate Athletes

INTRODUCTION

Established risk factors for hamstring strain injuries (HSI) include older age and prior HSI. However, these are nonmodifiable and have a limited role in injury prevention. Eccentric hamstring strength is a common component of HSI prevention programs, but its association with injury is less clear.

PURPOSE

This study aimed to determine if eccentric hamstring strength was prospectively associated with HSI among collegiate athletes, while controlling for sex, age, and prior HSI. We hypothesized that athletes with lower eccentric hamstring strength or greater between-limb strength asymmetry at preseason would have an increased risk of HSI.

METHODS

Hamstring eccentric strength measures, maximum total force ( FTotal ) and between-limb asymmetry in maximum force ( FAsym ), were measured at preseason on male and female athletes. HSIs were tracked over the subsequent 12 months. Generalized estimating equations were used to identify univariable and multivariable associations between athlete demographics, eccentric hamstring strength, and HSI risk.

RESULTS

Data for 326 athletes (85 female; 30 track, 43 basketball, 160 American football, 93 soccer) were included, and 64 HSIs were observed. Univariable associations between eccentric hamstring strength and subsequent HSI were nonsignificant ( FTotal : odds ratio [OR], 0.99 (95% confidence interval (CI), 0.93-1.05); P = 0.74; FAsym : OR, 1.35 (95% CI, 0.87-2.09); P = 0.23). No relationship between eccentric hamstring strength and HSI ( FAsym : OR, 1.32 (95% CI, 0.84-2.08); P = 0.23) was identified after adjusting for confounders including sex, age, and prior HSI.

CONCLUSIONS

No association between preseason eccentric hamstring strength and risk of subsequent HSI was identified after controlling for known risk factors and sex among collegiate athletes. Eccentric hamstring strengthening may continue to serve as a preventative approach to HSI, but it does not provide additional insight into HSI risk beyond factors such as age and prior HSI.

The development of a HAMstring InjuRy (HAMIR) index to mitigate injury risk through innovative imaging, biomechanics, and data analytics: protocol for an observational cohort study

Background

The etiology of hamstring strain injury (HSI) in American football is multi-factorial and understanding these risk factors is paramount to developing predictive models and guiding prevention and rehabilitation strategies. Many player-games are lost due to the lack of a clear understanding of risk factors and the absence of effective methods to minimize re-injury. This paper describes the protocol that will be followed to develop the HAMstring InjuRy (HAMIR) index risk prediction models for HSI and re-injury based on morphological, architectural, biomechanical and clinical factors in National Collegiate Athletic Association Division I collegiate football players.

Methods

A 3-year, prospective study will be conducted involving collegiate football student-athletes at four institutions. Enrolled participants will complete preseason assessments of eccentric hamstring strength, on-field sprinting biomechanics and muscle–tendon volumes using magnetic-resonance imaging (MRI). Athletic trainers will monitor injuries and exposure for the duration of the study. Participants who sustain an HSI will undergo a clinical assessment at the time of injury along with MRI examinations. Following completion of structured rehabilitation and return to unrestricted sport participation, clinical assessments, MRI examinations and sprinting biomechanics will be repeated. Injury recurrence will be monitored through a 6-month follow-up period. HAMIR index prediction models for index HSI injury and re-injury will be constructed.

Discussion

The most appropriate strategies for reducing risk of HSI are likely multi-factorial and depend on risk factors unique to each athlete. This study will be the largest-of-its-kind (1200 player-years) to gather detailed information on index and recurrent HSI, and will be the first study to simultaneously investigate the effect of morphological, biomechanical and clinical variables on risk of HSI in collegiate football athletes. The quantitative HAMIR index will be formulated to identify an athlete’s propensity for HSI, and more importantly, identify targets for injury mitigation, thereby reducing the global burden of HSI in high-level American football players.

Trial Registration The trial is prospectively registered on ClinicalTrials.gov (NCT05343052; April 22, 2022).

Monitoring Variables Influence on Random Forest Models to Forecast Injuries in Short-Track Speed Skating

Injuries limit the athletes' ability to participate fully in their training and competitive process. They are detrimental to performance, affecting the athletes psychologically while limiting physiological adaptations and long-term development. This study aims to present a framework for developing random forest classifier models, forecasting injuries in the upcoming 1 to 7 days, to assist the performance support staff in reducing injuries and maximizing performance within the Canadian National Female Short-Track Speed Skating Program. Forty different variables monitored daily over two seasons (2018-2019 and 2019-2020) were used to develop two sets of forecasting models. One includes only training load variables (TL), and a second (ALL) combines a wide array of monitored variables (neuromuscular function, heart rate variability, training load, psychological wellbeing, past injury type, and location). The sensitivity (ALL: 0.35 ± 0.19, TL: 0.23 ± 0.03), specificity (ALL: 0.81 ± 0.05, TL: 0.74 ± 0.03) and Matthews Correlation Coefficients (MCC) (ALL: 0.13 ± 0.05, TL: -0.02 ± 0.02) were computed. Paired T-test on the MCC revealed statistically significant (p < 0.01) and large positive effects (Cohen d > 1) for the ALL forecasting models' MCC over every forecasting window (1 to 7 days). These models were highly determined by the athletes' training completion, lower limb and trunk/lumbar injury history, as well as sFatigue, a training load marker. The TL forecasting models' MCC suggests they do not bring any added value to forecast injuries. Combining a wide array of monitored variables and quantifying the injury etiology conceptual components significantly improve the injury forecasting performance of random forest models. The ALL forecasting models' performances are promising, especially on one time windows of one or two days, with sensitivities and specificities being respectively above 0.5 and 0.7. They could add value to the decision-making process for the support staff in order to assist the Canadian National Female Team Short-Track Speed Skating program in reducing the number of incomplete training days, which could potentially increase performance. On longer forecasting time windows, ALL forecasting models' sensitivity and MCC decrease gradually. Further work is needed to determine if such models could be useful for forecasting injuries over three days or longer.

Forecasting football injuries by combining screening, monitoring and machine learning

Identifying players or circumstances associated with an increased risk of injury is fundamental for successful risk management in football. So far, time-constant and volatile risk factors are generally considered separately in either a screening (constant) or a monitoring (volatile) approach each resulting in a restricted set of explanatory variables. Consequently, improvements in predictive accuracy may be expected when screening and monitoring data are combined, especially when analysed with current machine learning (ML) techniques. This trial was designed as a prospective observational cohort study aiming to forecast non-contact time-loss injuries in male professional football (soccer). Injuries were registered according to the Fuller consensus. Gradient boosting with ROSE upsampling within a leave-one-out cross-validation was used for data analysis. The hierarchical data structure was considered throughout. Different splits of the original dataset were used to probe the robustness of results. Data of 88 players from 4 teams and 51 injuries could be analysed. The cross-validated performance of the gradient boosted model (ROC area under the curve 0.61) was promising and higher compared to models without integration of screening data. Importantly, holdout test set performance was similar (ROC area under the curve 0.62) indicating prospect of generalizability to new cases. However, the variation of predictive accuracy and feature importance with different splits of the original dataset reflects the relatively low number of events. It is concluded that ML-based injury forecasting based on the integration of screening and monitoring data is promising. However, external prospective verification and continued model development are required.

Blood sample profile helps to injury forecasting in elite soccer players

Purpose

By analyzing external workloads with machine learning models (ML), it is now possible to predict injuries, but with a moderate accuracy. The increment of the prediction ability is nowadays mandatory to reduce the high number of false positives. The aim of this study was to investigate if players’ blood sample profiles could increase the predictive ability of the models trained only on external training workloads.

Method

Eighteen elite soccer players competing in Italian league (Serie B) during the seasons 2017/2018 and 2018/2019 took part in this study. Players’ blood samples parameters (i.e., Hematocrit, Hemoglobin, number of red blood cells, ferritin, and sideremia) were recorded through the two soccer seasons to group them into two main groups using a non-supervised ML algorithm (k-means). Additionally to external workloads data recorded every training or match day using a GPS device (K-GPS 10 Hz, K-Sport International, Italy), this grouping was used as a predictor for injury risk. The goodness of ML models trained were tested to assess the influence of blood sample profile to injury prediction.

Results

Hematocrit, Hemoglobin, number of red blood cells, testosterone, and ferritin were the most important features that allowed to profile players and to analyze the response to external workloads for each type of player profile. Players’ blood samples’ characteristics permitted to personalize the decision-making rules of the ML models based on external workloads reaching an accuracy of 63%. This approach increased the injury prediction ability of about 15% compared to models that take into consideration only training workloads’ features. The influence of each external workload varied in accordance with the players’ blood sample characteristics and the physiological demands of a specific period of the season.

Conclusion

Field experts should hence not only monitor the external workloads to assess the status of the players, but additional information derived from individuals’ characteristics permits to have a more complete overview of the players well-being. In this way, coaches could better personalize the training program maximizing the training effect and minimizing the injury risk.

Analysis of developments and hotspots of international research on sports AI

In this paper, 1,538 papers retrieved with the keywords “sports artificial intelligence (AI)” on the Web of Science database since 2007 were taken as the data source, and the Cite Space V software was used to visualize and analyze them. A visual knowledge graph was used to streamline the countries, institutions and authors conducting sports AI research, discipline distribution, research hotspots and development trends in the past 15 years. Subsequently, its development direction and research progress were discussed. Sports AI was widely distributed, with the US, China and the UK leading the way. The most prolific authors and teams in research on sports AI were concentrated in American universities. Their main research direction is to develop and improve smart wearable devices based on machine learning and deep learning technologies for different groups of people. Research on sports AI involved multiple disciplines, which mainly applied and referred to research methodologies and theories on engineering, computer science and sports science. It could be seen from the frequency and centrality of keywords that in the current field of sports AI, machine learning is the main direction, artificial neural networks is the main algorithm, and practical and empirical research based on data mining is the focus. The research hotspots were divided into three major clusters: physical health promotion, sports injury prevention and control, and athletic performance enhancement. How to introduce intelligent technology into sports for a perfect integration still has an arduous and long way to go. Future development requires joint efforts and participation of scientific researchers, professionals and common people.

Predicting Hamstring Strains in Soccer Players Based on ROM: An Analysis From a Gender Perspective

Purpose: The aim of this study was to explore if specific hip and knee range-of-movement (ROM) tests can predict a risk factor for hamstring strain (HS) injury in male and female soccer players. Methods: One hundred amateur soccer players (56 men and 44 women) performed six tests to determine hip and knee ROM: straight leg raise test (SLR), modified Thomas test (TT), hip internal and external rotation (ER), hip abduction and adduction, Nachlas test and Ridge test. A logistic regression analysis was carried out to create a predictive model for HS injuries. Results: The percentage of HS injury was 20.45% and 30.35%, for female and male players. The logistic regression showed a significant model for both genders on the logit of suffering an HS injury with active-SLR and TT variables for females (R²_CS = 0.491; R²_N = 0.771) and active SLR and ER variables for males (R²_CS = 0.623; R²_N = 0.882). The predictive models correctly classify 95.5% and 94.6% of cases presenting good sensitivity (77.8% and 88.2%) and full (100%) and high (97.4%) specificity respectively. Furthermore, female players showed a greater ROM than males (p ≤ 0.01). Conclusion: Both female and male soccer players that suffered a HS injury had lower ROM in SLR, NT and RT and higher ROM in the TT that non-injured players. The tests that most likely predict HS injury are SLR and TT in females and SLR and ER in males. Thus, it is suggested to including specific exercises in amateur soccer players training programs to improve hip and knee ROM for injury prevention.

Impact of Gender and Feature Set on Machine-Learning-Based Prediction of Lower-Limb Overuse Injuries Using a Single Trunk-Mounted Accelerometer

Even though practicing sports has great health benefits, it also entails a risk of developing overuse injuries, which can elicit a negative impact on physical, mental, and financial health. Being able to predict the risk of an overuse injury arising is of widespread interest because this may play a vital role in preventing its occurrence. In this paper, we present a machine learning model trained to predict the occurrence of a lower-limb overuse injury (LLOI). This model was trained and evaluated using data from a three-dimensional accelerometer on the lower back, collected during a Cooper test performed by 161 first-year undergraduate students of a movement science program. In this study, gender-specific models performed better than mixed-gender models. The estimated area under the receiving operating characteristic curve of the best-performing male- and female-specific models, trained according to the presented approach, was, respectively, 0.615 and 0.645. In addition, the best-performing models were achieved by combining statistical and sports-specific features. Overall, the results demonstrated that a machine learning injury prediction model is a promising, yet challenging approach.

Poor Reporting of Exercise Interventions for Hamstring Strain Injury Rehabilitation: A Scoping Review of Reporting Quality and Content in Contemporary Applied Research

OBJECTIVE

To review the quality of reporting and identify the content of exercise interventions prescribed for hamstring strain injury (HSI) rehabilitation in the scientific literature from 2010 to 2020.

DESIGN

Scoping review.

LITERATURE SEARCH

We searched the bibliometric databases Web of Science, CINAHL, SPORTDiscus, Scopus, Cochrane Library, MEDLINE, and Embase.

STUDY SELECTION CRITERIA

Original research articles (randomized controlled trials and cohort studies) published from 2010 to 2020 that described an exercise rehabilitation intervention for participants with acute HSIs were included. Injuries must have been confirmed within 7 days of occurrence via clinical assessment and/or diagnostic imaging.

DATA SYNTHESIS

The quality of reporting, in terms of completeness of exercise intervention description, was evaluated using the Consensus on Exercise Reporting Template (CERT), and the content of interventions was categorized into exercise types.

RESULTS

Fourteen studies were included; exercise intervention quality of reporting was moderate in 3 studies and low in 11 studies. Using the 19-item CERT, an average of 8.8 items (range, 4-14) were reported across all studies. Two studies reported sufficient exercise content and progression information to allow replication. Exercises categorized as hamstring flexibility, hamstring strength, running related, and non-hamstring specific were prescribed in 13, 11, 10, and 10 studies, respectively. Half of the included studies incorporated all 4 exercise types in their exercise interventions.

CONCLUSION

There is a wide variety of exercise interventions applied in published research that has addressed HSI rehabilitation. Researchers must improve reporting quality to support other professionals in replicating exercise interventions and help practitioners to effectively implement research in practice. J Orthop Sports Phys Ther 2022;52(3):130-141. Epub 21 Sep 2021. doi:10.2519/jospt.2022.10641.

Prediction of Injuries in CrossFit Training: A Machine Learning Perspective

CrossFit has gained recognition and interest among physically active populations being one of the most popular and rapidly growing exercise regimens worldwide. Due to the intense and repetitive nature of CrossFit, concerns have been raised over the potential injury risks that are associated with its training including rhabdomyolysis and musculoskeletal injuries. However, identification of risk factors for predicting injuries in CrossFit athletes has been limited by the absence of relevant big epidemiological studies. The main purpose of this paper is the identification of risk factors and the development of machine learning-based models using ensemble learning that can predict CrossFit injuries. To accomplish the aforementioned targets, a survey-based epidemiological study was conducted in Greece to collect data on musculoskeletal injuries in CrossFit practitioners. A Machine Learning (ML) pipeline was then implemented that involved data pre-processing, feature selection and well-known ML models. The performance of the proposed ML models was assessed using a comprehensive cross validation mechanism whereas a discussion on the nature of the selected features is also provided. An area under the curve (AUC) of 77.93% was achieved by the best ML model using ensemble learning (Adaboost) on the group of six selected risk factors. The effectiveness of the proposed approach was evaluated in a comparative analysis with respect to numerous performance metrics including accuracy, sensitivity, specificity, AUC and confusion matrices to confirm its clinical relevance. The results are the basis for the development of reliable tools for the prediction of injuries in CrossFit.

Characteristics of Complex Systems in Sports Injury Rehabilitation: Examples and Implications for Practice

Complex systems are open systems consisting of many components that can interact among themselves and the environment. New forms of behaviours and patterns often emerge as a result. There is a growing recognition that most sporting environments are complex adaptive systems. This acknowledgement extends to sports injury and is reflected in the individual responses of athletes to both injury and rehabilitation protocols. Consequently, practitioners involved in return to sport decision making (RTS) are encouraged to view return to sport decisions through the complex systems lens to improve decision-making in rehabilitation. It is important to clarify the characteristics of this theoretical framework and provide concrete examples to which practitioners can easily relate. This review builds on previous literature by providing an overview of the hallmark features of complex systems and their relevance to RTS research and daily practice. An example of how characteristics of complex systems are exhibited is provided through a case of anterior cruciate ligament injury rehabilitation. Alternative forms of scientific inquiry, such as the use of computational and simulation-based techniques, are also discussed-to move the complex systems approach from the theoretical to the practical level.

Screening Hamstring Injury Risk Factors Multiple Times in a Season Does Not Improve the Identification of Future Injury Risk

PURPOSE

To determine if eccentric knee flexor strength and biceps femoris long head (BFlh) fascicle length were associated with prospective hamstring strain injury (HSI) in professional Australian Football players, and if more frequent assessments of these variables altered the association with injury risk.

METHODS

Across two competitive seasons, 311 Australian Football players (455 player seasons) had their eccentric knee flexor strength during the Nordic hamstring exercise and BFlh architecture assessed at the start and end of preseason and in the middle of the competitive season. Player age and injury history were also collected in preseason. Prospective HSIs were recorded by team medical staff.

RESULTS

Seventy-four player seasons (16%) sustained an index HSI. Shorter BFlh fascicles (<10.42 cm) increased HSI risk when assessed at multiple time points only (relative risk [RR], 1.9; 95% confidence interval [CI], 1.2-3.0). Neither absolute (N) nor relative (N·kg-1) eccentric knee flexor strength was associated with HSI risk, regardless of measurement frequency (RR range, 1.0-1.1); however, between-limb imbalance (>9%), when measured at multiple time points, was (RR, 1.8; 95% CI, 1.1-3.1). Prior HSI had the strongest univariable association with prospective HSI (RR, 2.9; 95% CI, 1.9-4.3). Multivariable logistic regression models identified a combination of prior HSI, BFlh architectural variables and between-limb imbalance in eccentric knee flexor strength as optimal input variables; however, their predictive performance did not improve with increased measurement frequency (area under the curve, 0.681-0.726).

CONCLUSIONS

More frequent measures of eccentric knee flexor strength and BFlh architecture across a season did not improve the ability to identify which players would sustain an HSI.

Predictive Analytic Techniques to Identify Hidden Relationships between Training Load, Fatigue and Muscle Strains in Young Soccer Players

This study aimed to analyze different predictive analytic techniques to forecast the risk of muscle strain injuries (MSI) in youth soccer based on training load data. Twenty-two young soccer players (age: 13.5 ± 0.3 years) were recruited, and an injury surveillance system was applied to record all MSI during the season. Anthropometric data, predicted age at peak height velocity, and skeletal age were collected. The session-RPE method was daily employed to quantify internal training/match load, and monotony, strain, and cumulative load over the weeks were calculated. A countermovement jump (CMJ) test was submitted before and after each training/match to quantify players' neuromuscular fatigue. All these data were used to predict the risk of MSI through different data mining models: Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM). Among them, SVM showed the best predictive ability (area under the curve = 0.84 ± 0.05). Then, Decision tree (DT) algorithm was employed to understand the interactions identified by the SVM model. The rules extracted by DT revealed how the risk of injury could change according to players' maturity status, neuromuscular fatigue, anthropometric factors, higher workloads, and low recovery status. This approach allowed to identify MSI and the underlying risk factors.

A Narrative Review for a Machine Learning Application in Sports: An Example Based on Injury Forecasting in Soccer

In the last decade, the number of studies about machine learning algorithms applied to sports, e.g., injury forecasting and athlete performance prediction, have rapidly increased. Due to the number of works and experiments already present in the state-of-the-art regarding machine-learning techniques in sport science, the aim of this narrative review is to provide a guideline describing a correct approach for training, validating, and testing machine learning models to predict events in sports science. The main contribution of this narrative review is to highlight any possible strengths and limitations during all the stages of model development, i.e., training, validation, testing, and interpretation, in order to limit possible errors that could induce misleading results. In particular, this paper shows an example about injury forecaster that provides a description of all the features that could be used to predict injuries, all the possible pre-processing approaches for time series analysis, how to correctly split the dataset to train and test the predictive models, and the importance to explain the decision-making approach of the white and black box models.

Prediction of Hamstring Injuries in Australian Football Using Biceps Femoris Architectural Risk Factors Derived From Soccer

BACKGROUND

Hamstring strain injuries are the most common injuries in team sports. Biceps femoris long head architecture is associated with the risk of hamstring injury in soccer. To assess the overall predictive ability of architectural variables, risk factors need to be applied to and validated across different cohorts.

PURPOSE

To assess the generalizability of previously established risk factors for a hamstring strain injury (HSI), including demographics, injury history, and biceps femoris long head (BFlh) architecture to predict HSIs in a cohort of elite Australian football players.

STUDY DESIGN

Cohort study; Level of evidence, 3.

METHODS

Demographic, injury history, and BFlh architectural data were collected from elite soccer (n = 152) and Australian football (n = 169) players at the beginning of the preseason for their respective competitions. Any prospectively occurring HSIs were reported to the research team. Optimal cut points for continuous variables used to determine an association with the HSI risk were established from previously published data in soccer and subsequently applied to the Australian football cohort to derive the relative risk (RR) for these variables. Logistic regression models were built using data from the soccer cohort and utilized to estimate the probability of an injury in the Australian football cohort. The area under the curve (AUC) and Brier score were the primary outcome measures to assess the performance of the logistic regression models.

RESULTS

A total of 27 and 30 prospective HSIs occurred in the soccer and Australian football cohorts, respectively. When using cut points derived from the soccer cohort and applying these to the Australian football cohort, only older athletes (aged ≥25.4 years; RR, 2.7 [95% CI, 1.4-5.2]) and those with a prior HSI (RR, 2.5 [95% CI, 1.3-4.8]) were at an increased risk of HSIs. Using the same approach, height, weight, fascicle length, muscle thickness, pennation angle, and relative fascicle length were not significantly associated with an increased risk of HSIs in Australian football players. The logistic regression model constructed using age and prior HSIs performed the best (AUC = 0.67; Brier score = 0.14), with the worst performing model being the one that was constructed using pennation angle (AUC = 0.53; Brier score = 0.18).

CONCLUSION

Applying cut points derived from previously published data in soccer to a dataset from Australian football identified older age and prior HSIs, but none of the modifiable HSI risk factors, to be associated with an injury. The transference of HSI risk factor data between soccer and Australian football appears limited and suggests that cohort-specific cut points must be established.

Detecting Injury Risk Factors with Algorithmic Models in Elite Women's Pathway Cricket

This exploratory retrospective cohort analysis aimed to explore how algorithmic models may be able to identify important risk factors that may otherwise not have been apparent. Their association with injury was then assessed with more conventional data models. Participants were players registered on the England and Wales Cricket Board women's international development pathway (n=17) from April 2018 to August 2019 aged between 14-23 years (mean 18.2±1.9) at the start of the study period. Two supervised learning techniques (a decision tree and random forest with traditional and conditional algorithms) and generalised linear mixed effect models explored associations between risk factors and injury. The supervised learning models did not predict injury (decision tree and random forest area under the curve [AUC] of 0.66 and 0.72 for conditional algorithms) but did identify important risk factors. The best-fitting generalised linear mixed effect model for predicting injury (Akaike Information Criteria [AIC]=843.94, conditional r-squared=0.58) contained smoothed differential 7-day load (P<0.001), average broad jump scores (P<0.001) and 20 m speed (P<0.001). Algorithmic models identified novel injury risk factors in this population, which can guide practice and future confirmatory studies can now investigate.

Lower-limb injury in elite Australian football: A narrative review of kinanthropometric and physical risk factors

OBJECTIVE

This review aims to provide a succinct and critical analysis of the current physical and mechanical demands of elite Australian football while examining lower-limb injury and the associated physical and kinanthropometric risk factors.

METHODS

MEDLINE, PubMed, Web of Science and SPORTSDiscus electronic databases were searched for studies that investigated the playing demands, injury trends, and physical and kinanthropometric injury risk factors of elite Australian football. Articles from similar team sports including soccer and rugby (union and league) were also included.

RESULTS

While the physical demands of elite AF have steadied over the past decade, injury rates continue to rise with more than two-thirds of all injuries affecting the lower-limbs. Body composition and musculoskeletal morphological assessments are regularly adopted in many sporting settings with current research suggesting high and low body mass are both associated with heightened injury risk. However, more extensive investigations are required to determine whether the proportions of muscle and fat are linked. Repeated assessment of musculoskeletal morphology may also provide further insight into stress fracture rates.

CONCLUSIONS

While kinanthropometric and physical attributes are highly valued within elite sporting environments, establishing a deeper connection with injury may provide practitioners with more insight into current injury trends.

Low Horizontal Force Production Capacity during Sprinting as a Potential Risk Factor of Hamstring Injury in Football

Clear decreases in horizontal force production capacity during sprint acceleration have been reported after hamstring injuries (HI) in football players. We hypothesized that lower FH0 is associated with a higher HI occurrence in football players. We aimed to analyze the association between sprint running horizontal force production capacities at low (FH0) and high (V0) velocities, and HI occurrence in football. This prospective cohort study included 284 football players over one season. All players performed 30 m field sprints at the beginning and different times during the season. Sprint velocity data were used to compute sprint mechanical properties. Players' injury data were prospectively collected during the entire season. Cox regression analyses were performed using new HI as the outcome, and horizontal force production capacity (FH0 and V0) was used at the start of the season (model 1) and at each measurement time point within the season (model 2) as explanatory variables, adjusted for individual players' (model 2) age, geographical group of players, height, body mass, and previous HI, with cumulative hours of football practice as the time scale. A total of 47 new HI (20% of all injuries) were observed in 38 out of 284 players (13%). There were no associations between FH0 and/or V0 values at the start of the season and new HI occurrence during the season (model 1). During the season, a total of 801 measurements were performed, from one to six per player. Lower measured FH0 values were significantly associated with a higher risk of sustaining HI within the weeks following sprint measurement (HR = 2.67 (95% CI: 1.51 to 4.73), p < 0.001) (model 2). In conclusion, low horizontal force production capacities at low velocity during early sprint acceleration (FH0) may be considered as a potential additional factor associated with HI risk in a comprehensive, multifactorial, and individualized approach.

Lower limb joint position sense and prospective hamstring injury

BACKGROUND

The hamstrings remain the most injured muscle group within the Australian Football League (AFL).

OBJECTIVE

To investigate preseason measures of hip and knee joint position sense (JPS) and prospective hamstring injury in AFL players.

DESIGN

Prospective cohort study.

METHODS

A total of 116 AFL players were recruited for this study. JPS was assessed with 3-D sensors using mono-articular hip (45° flexion and 0°) and knee (90° and 45° flexion) joint reproduction tests conducted in the preseason. Hamstring injury data were collected prospectively in the following AFL season. Wilcoxon-signed rank tests were used to assess between the subsequently injured and uninjured limbs. Mann-Whitney U tests were used to assess between group differences and odds ratio (OR) were used to predict players at risk of hamstring injury.

RESULTS

Eight players with JPS data sustained a season hamstring injury and 108 players did not. JPS was not significantly different between the subsequently injured and uninjured limbs (all P values > 0.05). No significant differences for any JPS measure were found between the subsequently injured and uninjured players (all p's > 0.05). ORs did not achieve significance for AE (2.7, p = 0.21) or for RMSE (OR = 1.9, p = 0.44).

CONCLUSION

Lower limb JPS measures were not predictive of hamstring injury in AFL players.

A prospective study of risk factors for hamstring injury in Australian football league players

This study aimed to explore the association between hamstring strength, age and lower limb soft tissue injury history and subsequent hamstring injury among Australian Football League (AFL) players. This prospective cohort study recruited 125 players from three professional AFL teams. Eccentric knee flexor strength was assessed while performing the Nordic hamstring exercise in pre-season, and injury data were collected retrospectively (hamstring, groin, calf, quadriceps and knee), and prospectively (hamstring injuries) for one AFL playing season. Fourteen players (11%) sustained a hamstring injury in the subsequent playing season. Nordic strength was not significantly associated with future hamstring injury (Odds Ratio (OR) 1.9, p = 0.36), whereas player age greater than 25 years (OR = 2.9, p < 0.05), report of a hamstring injury within the previous year (OR = 3.7, p = 0.01), or greater than 1-year (OR = 3.6, p = 0.01), a previous groin (OR = 8.6, p < 0.01) or calf injury (OR = 4.6, p = 0.01) were factors significantly associated with subsequent hamstring injury. Based on these findings, increasing age and previous hamstring, groin and calf injury are all associated with an elevated risk of subsequent hamstring injury in AFL players.

Is Pre-season Eccentric Strength Testing During the Nordic Hamstring Exercise Associated with Future Hamstring Strain Injury? A Systematic Review and Meta-analysis

BACKGROUND

Interventions utilising the Nordic hamstring exercise (NHE) have resulted in reductions in the incidence of hamstring strain injury (HSI). Subsequently, quantifying eccentric knee flexor strength during performance of the NHE to identify an association with the occurrence of future HSI has become increasingly common; however, the data to date are equivocal.

OBJECTIVE

To systematically review the association between pre-season eccentric knee flexor strength quantified during performance of the NHE and the occurrence of future HSI.

DESIGN

Systematic review and meta-analysis.

DATA SOURCES

CINAHL, Cochrane Library, Medline Complete, Embase, Web of Science and SPORTDiscus databases were searched from January 2013 to January 10, 2020.

ELIGIBILITY CRITERIA FOR SELECTING STUDIES

Prospective cohort studies which assessed the association between pre-season eccentric knee flexor strength quantified during performance of the NHE and the occurrence of future HSI.

METHODS

Following database search, article retrieval and title and abstract screening, articles were assessed for eligibility against pre-defined criteria then assessed for risk of bias. Meta-analysis was used to pool data across studies, with meta-regression utilised where possible.

RESULTS

A total of six articles were included in the meta-analysis, encompassing 1100 participants. Comparison of eccentric knee flexor strength during performance of the NHE in 156 injured participants and the 944 uninjured participants revealed no significant differences, regardless of whether strength was expressed as absolute (N), relative to body mass (N kg^-1) or between-limb asymmetry (%). Meta-regression analysis revealed that the observed effect sizes were generally not moderated by age, mass, height, strength, or sport played.

CONCLUSION

Eccentric knee flexor strength quantified during performance of the NHE during pre-season provides limited information about the occurrence of a future HSI.

Sprinting, Strength, and Architectural Adaptations Following Hamstring Training in Australian Footballers

The aim of this study was to determine the sprinting, strength, and architectural adaptations following a hip-dominant flywheel (FLY) or Nordic hamstring exercise (NHE) intervention in Australian footballers. Twenty-seven male athletes were randomized to FLY (n = 13) or NHE (n = 14) training across a 39-week period (inclusive of pre-season and in-season). Biceps femoris long head (BFlh) architecture was assessed throughout. Eccentric hamstring strength and 40 m sprint times (with force-velocity profiling) were assessed at baseline, end of pre-season, and following the intervention. After the intervention, BFlh fascicle length was longer in both groups compared to baseline (FLY: 1.16 cm, 95%CI: 0.66 to 1.66 cm, d = 1.99, p < 0.001; NHE: 1.08 cm, 95%CI: 95%CI 0.54 to 1.61 cm, d = 1.73, p < 0.001). Both groups also increased their eccentric strength (FLY: mean change 82 N, 95%CI 12 to 152 N, d = 1.34, p = 0.026; NHE: mean change 97 N, 95%CI 47 to 146 N, d = 1.77, p = 0.001). After pre-season, the NHE group improved their 5 m sprint time by 3.5% (±1.2%) and were 3.7% (±1.4%) and 2.0% (±0.5%) faster than the FLY group across 5 m and 10 m, respectively. At the end of pre-season, the FLY group improved maximal velocity by 3.4% (±1.4%) and improved horizontal force production by 9.7% in-season (±2.2%). Both a FLY and NHE intervention increase BFlh fascicle length and eccentric strength in Australian Footballers. An NHE intervention led to enhanced acceleration capacity. A FLY intervention was suggested to improve maximal sprint velocity and horizontal force production, without changes in sprint times. These findings have implications for hamstring injury prevention but also programs aimed at improving sprint performance.

Machine learning methods in sport injury prediction and prevention: a systematic review

PURPOSE

Injuries are common in sports and can have significant physical, psychological and financial consequences. Machine learning (ML) methods could be used to improve injury prediction and allow proper approaches to injury prevention. The aim of our study was therefore to perform a systematic review of ML methods in sport injury prediction and prevention.

METHODS

A search of the PubMed database was performed on March 24th 2020. Eligible articles included original studies investigating the role of ML for sport injury prediction and prevention. Two independent reviewers screened articles, assessed eligibility, risk of bias and extracted data. Methodological quality and risk of bias were determined by the Newcastle-Ottawa Scale. Study quality was evaluated using the GRADE working group methodology.

RESULTS

Eleven out of 249 studies met inclusion/exclusion criteria. Different ML methods were used (tree-based ensemble methods (n = 9), Support Vector Machines (n = 4), Artificial Neural Networks (n = 2)). The classification methods were facilitated by preprocessing steps (n = 5) and optimized using over- and undersampling methods (n = 6), hyperparameter tuning (n = 4), feature selection (n = 3) and dimensionality reduction (n = 1). Injury predictive performance ranged from poor (Accuracy = 52%, AUC = 0.52) to strong (AUC = 0.87, f1-score = 85%).

CONCLUSIONS

Current ML methods can be used to identify athletes at high injury risk and be helpful to detect the most important injury risk factors. Methodological quality of the analyses was sufficient in general, but could be further improved. More effort should be put in the interpretation of the ML models.

Can Injuries Be Predicted by Functional Movement Screen in Adolescents? The Application of Machine Learning

ABSTRACT

Karuc, J, Mišigoj-Duraković, M, Šarlija, M, Marković, G, Hadžić, V, Trošt-Bobić, T, and Sorić, M. Can injuries be predicted by functional movement screen in adolescents? The application of machine learning. J Strength Cond Res 35(4): 910-919, 2021-This study used machine learning (ML) to predict injuries among adolescents by functional movement testing. This research is a part of the CRO-PALS study conducted in a representative sample of adolescents and analyses for this study are based on nonathletic (n = 364) and athletic (n = 192) subgroups of the cohort (16-17 years). Sex, age, body mass index (BMI), body fatness, moderate-to-vigorous physical activity (MVPA), training hours per week, Functional Movement Screen (FMS), and socioeconomic status were assessed at baseline. A year later, data on injury occurrence were collected. The optimal cut-point of the total FMS score for predicting injury was calculated using receiver operating characteristic curve. These predictors were included in ML analyses with calculated metrics: area under the curve (AUC), sensitivity, specificity, and odds ratio (95% confidence interval [CI]). Receiver operating characteristic curve analyses with associated criterium of total FMS score >12 showed AUC of 0.54 (95% CI: 0.48-0.59) and 0.56 (95% CI: 0.47-0.63), for the nonathletic and athletic youth, respectively. However, in the nonathletic subgroup, ML showed that the Naïve Bayes exhibited highest AUC (0.58), whereas in the athletic group, logistic regression was demonstrated as the model with the best predictive accuracy (AUC: 0.62). In both subgroups, with given predictors: sex, age, BMI, body fat percentage, MVPA, training hours per week, socioeconomic status, and total FMS score, ML can give a more accurate prediction then FMS alone. Results indicate that nonathletic boys who have lower-body fat could be more prone to suffer from injury incidence, whereas among athletic subjects, boys who spend more time training are at a higher risk of being injured. Conclusively, total FMS cut-off scores for each subgroup did not successfully discriminate those who suffered from those who did not suffer from injury, and, therefore, our research does not support FMS as an injury prediction tool.

The Relationship Between the Contact Force at the Ankle Hook and the Hamstring Muscle Force During the Nordic Hamstring Exercise

A novel device has been developed to assess eccentric hamstring strength during the Nordic hamstring exercise (NHE) by measuring the contact force at the ankle hook (brace). The purpose of this study was to determine the correlation between the force measured at the ankle hook and the hamstring force estimated by a low extremity model. Thirteen male college sprinters were recruited to perform NHE on an instrumented device Nordbord (Vald Performance, Australia). Contact forces were measured at a sampling rate of 50 Hz at the hooks using the uniaxial load cells. 3D kinematics were measured simultaneously at a sampling rate of 200 Hz using a 16-camera motion analysis system (Vicon Motion Analysis, Oxford, United Kingdom) during the NHE. The data were processed with Visual 3D (C-Motion, Germantown, MD, United States) and OpenSim (NCSRR, Stanford, CA, United States) to calculate the knee joint center’s coordinates and hamstring moment arms during NHE. A static low extremity model was built to estimate the hamstring force during NHE. We have observed a significant but not very high correlation (r² = 0.58) between peak hamstring force and the peak contact force at the ankle hook. The peak contact force measured at the ankle hook can only explain a little more than half of the variations in peak hamstring muscle forces during NHE. Caution must be exercised when assessing the eccentric hamstring strength using the ankle contact force during NHE.

A Field-Based Approach to Determine Soft Tissue Injury Risk in Elite Futsal Using Novel Machine Learning Techniques

Lower extremity non-contact soft tissue (LE-ST) injuries are prevalent in elite futsal. The purpose of this study was to develop robust screening models based on pre-season measures obtained from questionnaires and field-based tests to prospectively predict LE-ST injuries after having applied a range of supervised Machine Learning techniques. One hundred and thirty-nine elite futsal players underwent a pre-season screening evaluation that included individual characteristics; measures related to sleep quality, athlete burnout, psychological characteristics related to sport performance and self-reported perception of chronic ankle instability. A number of neuromuscular performance measures obtained through three field-based tests [isometric hip strength, dynamic postural control (Y-Balance) and lower extremity joints range of motion (ROM-Sport battery)] were also recorded. Injury incidence was monitored over one competitive season. There were 25 LE-ST injuries. Only those groups of measures from two of the field-based tests (ROM-Sport battery and Y-Balance), as independent data sets, were able to build robust models [area under the receiver operating characteristic curve (AUC) score ≥0.7] to identify elite futsal players at risk of sustaining a LE-ST injury. Unlike the measures obtained from the five questionnaires selected, the neuromuscular performance measures did build robust prediction models (AUC score ≥0.7). The inclusion in the same data set of the measures recorded from all the questionnaires and field-based tests did not result in models with significantly higher performance scores. The model generated by the UnderBagging technique with a cost-sensitive SMO as the base classifier and using only four ROM measures reported the best prediction performance scores (AUC = 0.767, true positive rate = 65.9% and true negative rate = 62%). The models developed might help coaches, physical trainers and medical practitioners in the decision-making process for injury prevention in futsal.

New Machine Learning Approach for Detection of Injury Risk Factors in Young Team Sport Athletes

The purpose of this article is to present how predictive machine learning methods can be utilized for detecting sport injury risk factors in a data-driven manner. The approach can be used for finding new hypotheses for risk factors and confirming the predictive power of previously recognized ones. We used three-dimensional motion analysis and physical data from 314 young basketball and floorball players (48.4% males, 15.72±1.79 yr, 173.34±9.14 cm, 64.65±10.4 kg). Both linear (L1-regularized logistic regression) and non-linear methods (random forest) were used to predict moderate and severe knee and ankle injuries (N=57) during three-year follow-up. Results were confirmed with permutation tests and predictive risk factors detected with Wilcoxon signed-rank-test (p<0.01). Random forest suggested twelve consistent injury predictors and logistic regression twenty. Ten of these were suggested in both models; sex, body mass index, hamstring flexibility, knee joint laxity, medial knee displacement, height, ankle plantar flexion at initial contact, leg press one-repetition max, and knee valgus at initial contact. Cross-validated areas under receiver operating characteristic curve were 0.65 (logistic regression) and 0.63 (random forest). The results highlight the difficulty of predicting future injuries, but also show that even with models having relatively low predictive power, certain predictive injury risk factors can be consistently detected.

Using machine learning to improve our understanding of injury risk and prediction in elite male youth football players

OBJECTIVES

The purpose of this study was to examine whether the use of machine learning improved the ability of a neuromuscular screen to identify injury risk factors in elite male youth football players.

DESIGN

Prospective cohort study.

METHODS

355 elite youth football players aged 10-18 years old completed a prospective pre-season neuromuscular screen that included anthropometric measures of size, as well as single leg countermovement jump (SLCMJ), single leg hop for distance (SLHD), 75% hop distance and stick (75%Hop), Y-balance anterior reach and tuck jump assessment. Injury incidence was monitored over one competitive season. Risk profiling was assessed using traditional regression analyses and compared to supervised machine learning algorithms constructed using decision trees.

RESULTS

Using continuous data, multivariate logistic analysis identified SLCMJ asymmetry as the sole significant predictor of injury (OR 0.94, 0.92-0.97, p<0.001), with a specificity of 97.7% and sensitivity of 15.2% giving an AUC of 0.661. The best performing decision tree model provided a specificity of 74.2% and sensitivity of 55.6% with an AUC of 0.663. All variables contributed to the final machine model, with asymmetry in the SLCMJ, 75%Hop and Y-balance, plus tuck jump knee valgus and anthropometrics being the most frequent contributors.

CONCLUSIONS

Although both statistical methods reported similar accuracy, logistic regression provided very low sensitivity and only identified a single neuromuscular injury risk factor. The machine learning model provided much improved sensitivity to predict injury and identified interactions of asymmetry, knee valgus angle and body size as contributing factors to an injurious profile in youth football players.