Anne Tjønndal
Stian Røsten- Department of Leadership and Innovation, Faculty of Social Sciences, Nord University, Bodø, Norway
Sports injury prevention is an important part of the athlete welfare and safeguarding research field. In sports injury prevention, sport-related concussion (SRC) has proved to be one of the most difficult and complex injuries to manage in terms of prevention, diagnosis, classification, treatment and rehabilitation. SRC can cause long-term health issues and is a commonly reported injury in both adult and youth athletes around the world. Despite increased knowledge of the prevalence of SRC, very few tools are available for diagnosing SRC in athletic settings. Recent technological innovations have resulted in different machine learning and deep learning methodologies being tested to improve the management of this complex sports injury. The purpose of this article is to summarize and map the existing research literature on the use of machine learning in the management of SRC, ascertain where there are gaps in the existing research and identify recommendations for future research. This is explored through a scoping review. A systematic search in the three electronic databases SPORTDiscus, PubMed and Scopus identified an initial 522 studies, of which 24 were included in the final review, the majority of which focused on machine learning for the prediction and prevention of SRC (N = 10), or machine learning for the diagnosis and classification of SRC (N = 11). Only 3 studies explored machine learning approaches for the treatment and rehabilitation of SRC. A main finding is that current research highlights promising practical uses (e.g., more accurate and rapid injury assessment or return-to-sport participation criteria) of machine learning in the management of SRC. The review also revealed a narrow research focus in the existing literature. As current research is primarily conducted on male adolescents or adults from team sports in North America there is an urgent need to include wider demographics in more diverse samples and sports contexts in the machine learning algorithms. If research datasets continue to be based on narrow samples of athletes, the development of any new diagnostic and predictive tools for SRC emerging from this research will be at risk. Today, these risks appear to mainly affect the health and safety of female athletes.
Introduction
Safeguarding and athlete welfare in sport have received increased public and academic attention in the last decade (Nery et al., 2021). Together, these terms encompass health and safety issues related to topics such as gendered violence (Barker-Ruchti et al., 2021), abusive coaching practices (McMahon et al., 2020), racism (Manning, 2021), bullying (Parent et al., 2021), the legitimization and stigmatization of disordered eating (Papathomas, 2019), psychological abuse (Krahn, 2021) and the risk of sustaining sports injuries (White, 2021). In the context of sports injuries, sports-related concussion (SRC) represents an important subject in that it can result in long-term health issues for athletes (Gagnon and Ptito, 2017; Sindelar and Bailes, 2017). The Concussion in Sport Group (CISG) defines SRC as: “a traumatic brain injury that is defined as a complex pathophysiological process affecting the brain, induced by biomechanical forces with several common features that help define its nature” (McCrory et al., 2017, p. 877).
For coaches and athletes, SRCs are notoriously difficult injuries to manage, as the changes in the neurological function associated with concussion are often rapid and resolved spontaneously (Malcolm, 2020). Due to this, many SRCs can go unreported and unrecognized by the athletes themselves and their coaches (Ventresca and McDonald, 2020). The difficulty in detecting this type of injury is part of the reason why SRC has been called “a silent epidemic” (Carroll and David, 2012) and “a contemporary crisis in sport” (Malcolm, 2020). There are also cases where SRC is purposely left unreported, as contracting a concussion will lead to athletes being suspended for set periods of time in many sports. Therefore, coaches and athletes may avoid reporting head injuries if they suspect a concussion, fearing that it will result in the athlete being omitted from important competitions. These complexities showcase how all studies on the topic of SRC concern important aspects of athlete safety and health in both adult and youth sport (Tjønndal et al., 2022).
SRC is a commonly reported injury among adult and youth athletes around the globe and represents a safeguarding issue in contact and non-contact sports at the elite and recreational levels (Tjønndal and Wågan, 2021). Ferris et al. (2021, p. 1040–1041) state that “up to an estimated 3.8 million sport-related [...] concussions occur annually in the United States” alone. Clinically, a concussion is known to affect an athlete's memory, reaction time and balance. A previous concussion may also lead to athletes being prone to subsequent concussions, which could result in cognitive impairment, depression or chronic traumatic encephalopathy (Ventresca and McDonald, 2020). As a research field, SRC has expanded significantly over the last 20 years. This is reflected in the growth of systematic reviews of different sports and age groups (e.g., Langlois et al., 2006; Clay et al., 2013; Pfister et al., 2016), as well as sport-specific reviews (Gardner et al., 2015; Mooney et al., 2020; Tjønndal et al., 2022), reviews on gender differences (Dick, 2009), risk factors (Abrahams et al., 2014), the effect of protective sports equipment (Benson et al., 2009) and the use of wearable technologies in the detection of SRC (Schmid et al., 2021). These studies and others reflect an increased research interest in the management of SRC. In this context, we use the term management to describe all aspects of SRC, including the prediction, prevention, diagnosis, classification, treatment and rehabilitation of the injury. Arguably, the growth of research on SRC is a response to increased public awareness of the risks of SRC and is a part of a wider movement of securing athlete welfare in sport (Graham et al., 2014).
While the incidence of SRC may vary between different sports, it is difficult to assess exactly how common SRC is in sport and among different age groups of athletes (McCrory, 2013). For instance, in a systematic review of head injuries and concussion in boxing, the studies included in the review reported between 10 and 70% head injuries of all the recorded injuries in the athlete populations (Tjønndal et al., 2022). In another recent systematic review of the prevalence of SRC across 12 youth sports, Pfister et al. (2016) estimated that the overall incidence rate of SRC was 0.23 per 1000 athlete exposures (AEs), varying from the highest in rugby (4.18) and ice hockey (1.20) and the lowest in volleyball (0.03) and baseball (0.07). These reviews showcase that SRC is a complicated area of study in the field of athlete welfare and safeguarding.
Despite the increased knowledge of the prevalence of SRC, very few tools are available to immediately diagnose or measure SRC in athletic settings. A commonly used diagnosis protocol for SRC is the Sideline Concussion Assessment Tool (SCAT), which combines both subjective and objective measurements and takes between 10 and 20 minutes to conduct (Yengo-Kahn et al., 2016; Echemendia et al., 2017). However, recent technological innovations have resulted in different machine learning and deep learning methodologies being tested to improve injury prediction, prevention and diagnosis in sports (Jauhiainen et al., 2021; Van Eetvelde et al., 2021).
Machine learning is a field of computer science that automates analytical model building to reveal patterns in data to help humans make better decisions (Cenek et al., 2018). Similarly, deep learning is described as a class of artificial neural networks that learn in an unsupervised manner. In health care, these approaches are used to analyse medical data in disease prevention, diagnosis, patient monitoring and the development of new protocols (Patel et al., 2021). Machine learning approaches can recognize patterns, trends and associations and are critical for understanding SRC (Agoston, 2019). The emergence of machine learning algorithms in medical practice and their capacity to integrate thousands of biomarkers to build objective and automated in vivo detection tools to support clinical diagnosis of SRC is a promising development (Jordan and Mitchell, 2015; Obermeyer and Emanuel, 2016). Put simply, the use of machine learning techniques in sports injury prevention could provide faster and more accurate ways of managing SRC in athletes in a wide range of sports.
From this starting point, the purpose of the review is to: (1) summarize and map the existing research literature on the use of machine learning in the management of SRC and (2) ascertain where there are gaps in the existing research and identify recommendations for future research. We investigate these research purposes through a scoping review of research literature using the following three relevant academic databases: SPORTDiscus, PubMed and Scopus.
Method: Scoping Review
If we use the analogy of an ocean to express the current body of knowledge on machine learning in the management of SRC, at present there are no academic reviews to tell us: (a) the size or shape of the ocean (i.e., prevalence), (b) in which direction the wind blows and the waves travel (i.e., potentials for use) or (c) which species (i.e., sports) are present in the ocean. As machine learning in the management of SRC is a new scholarly field, a scoping review is valuable for gaining an overview of the research that has been conducted on this topic (Arksey and O'Malley, 2005). While scoping reviews do not typically include evidence-based methodological processes [e.g., assessment of bias, rating of level of evidence (LoE)] as recommended in the statement guidelines for the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Tricco et al., 2018), they are particularly helpful for understanding complex topics by appraising literature with a wide range of different study designs and methods (Peters et al., 2020). As SRC is often described as one of the most complex injuries of the most complex organ (Agoston, 2019), we find the scoping review to be a fruitful methodological approach for our examination of the field.
According to Arksey and O'Malley (2005), a scoping review is ideal for investigating the character and extent of research in a particular domain or on a specific topic. The present review was designed to summarize the current body of knowledge on how machine learning is used in the context of managing SRC. We chose a scoping review approach because the use of machine learning for the management of SRC is a relatively new topic in the sports injury field. As Bergeron et al. (2019, p. 1363) state: “the advent of using machine learning in addressing the complexity of various human health challenges is only recently underway.” Therefore, a scoping review is regarded as an appropriate approach for our examination, in that it is a review method that often precedes systematic literature reviews and is frequently used to map novel research fields to identify research gaps requiring further attention (Munn et al., 2018; Peters et al., 2020). In the following, we present our search strategy, the procedure of the scoping review and the study's limitations.
Search Strategy
The literature search was guided by the PRISMA-ScR guidelines (Tricco et al., 2018) and was conducted in October 2021 in the three academic databases of SPORTDiscus, PubMed and Scopus. These databases were chosen because they collectively provide a comprehensive academic profile of machine learning in the management of SRC in a wide variety of scholarly disciplines (e.g., sports science, medicine, biomedical, neurology, engineering, sociology, pedagogy, psychology).
The studies to be considered for inclusion in the review were identified using the following search term: “machine learning” or “ml” or “artificial intelligence” or “ai” or “learning algorithms” or “learning systems” AND “concussion” or “head injury” or “mild traumatic brain injury” or “mild tbi” or “mtbi” AND “sports” or “athletics” or “athletes” or “players” or “sport” or “competition.” No restrictions on publications dates were made in the search.
Although the intention was to include a wide profile of research on machine learning in the management of SRC, some eligibility criteria were applied. For inclusion in the review, studies needed to (1) focus on machine learning approaches to the management of SRC, (2) include data from athletes and/or be original empirical studies conducted in the context of sport, (3) be reported in peer-reviewed journals and (4) be written in English.
Procedure of the Review
The initial search identified 522 articles. Two hundred forty articles were identified from Scopus, 250 via PubMed and 32 through SPORTDiscus (Figure 1). The process for identifying articles for inclusion in the final review can be divided into two main phases. In phase 1, each individual study's author(s) and year of publication were recorded schematically. From this initial list of entries, the second author screened the title and abstract to identify relevant studies against the inclusion criteria. This preliminary phase revealed that many of the identified articles did not match the review's aim or fulfill the inclusion criteria. For example, a large number of hits were related to contexts other than sport (e.g., hospital, traffic injuries), utilized animal samples (e.g., mice, rats, piglets), or investigated the “phenomenon” of SRC but did not consider the use of machine learning approaches in its management. In those instances where the second author was uncertain about including an article in the potential final sample, the first author was consulted. In phase 1, duplicates between the three databases were removed.
Figure 1. Flow chart of the scoping review process, adapted from Page et al. (2021).
After the initial screening of titles and abstracts, 426 studies were excluded, leaving a total of 96 studies for further analysis. From this refined list, a more comprehensive assessment took place in phase 2, where both authors performed the full text screening of all the 96 articles. When assessing the full texts for eligibility, it was apparent that several articles only mentioned machine learning briefly but did not directly apply machine learning to analyse SRC or present the findings from such an approach. These articles were therefore excluded as they did not meet the inclusion criteria. In addition, several studies were considered according to the study sample or the use of non-sporting datasets (e.g., to test/simulate effects of different head impacts). For instance, in cases with both athletes (SRC) and other concussion patients, at least 50% of subjects had to have sustained their concussion in a sporting context in order for the study to be included in the review. Any uncertainties or issues that occurred during the process were resolved in discussions between the two authors. After the full text screening, 24 studies were ultimately included in the final review. The process of the scoping review is illustrated in Figure 1.
Once the final list had been completed, we summarized the following information about the included articles: year of publication, title of the journal, participants (sex and age), sample size, sports, geographical region the study was conducted in, study design and methods, name of the machine learning algorithm, key findings and how the machine learning approaches were used in the management of SRC (three categorisations: (1) diagnosis and classification, (2) prediction and prevention and (3) treatment and rehabilitation). The first author reviewed and amalgamated the work, while any disagreements were discussed. This clarified our understanding of the studies and was valuable to ensure consistency.
Limitations
Although this review provides insight into how machine learning can be used to manage SRC in different ways, there are limitations in our methodological approach. An important limitation is that both authors are social scientists and that this topic is dominated by research on medicine and engineering. When discussing our interpretations of the different studies, working in a dyad rather than a larger group of researchers may also have been a limitation. For instance, being able to discuss the studies included in the review with researchers in the field of sports medicine may have been more productive. However, as this is not a systematic review (and therefore does not aim to evaluate LoE), we do not consider our social scientific backgrounds to be detrimental to our analysis and findings. Finally, as the review was restricted to peer-reviewed studies written in English we excluded research reported in other languages that may have been relevant to the topic. However, this was a pragmatic necessity, given that the authors are only familiar with English and the Scandinavian languages.
Results
The presentation of the results is guided by the first purpose of this scoping review, which is to summarize and map the existing research literature on the use of machine learning in the management of SRC. The scoping review of the databases SPORTDiscus, PubMed and Scopus identified an initial 522 peer-reviewed studies reported in academic journals (see Figure 1). After applying the inclusion criteria to the titles and abstracts, 96 potentially relevant studies were included for full text examination by both authors. From these, 72 studies were excluded because they did not meet the inclusion criteria. There were two main reasons for exclusion at this stage: (1) no athletes were included in the sample (e.g., patients from non-sport related concussions) and (2) no machine learning techniques were applied in the study. This resulted in a final sample of 24 articles being included in our scoping review. The results of the scoping review are summarized in Table 1. Complete references of the included studies can be found in the reference list of this article.
Table 1. Results of the scoping review.
Machine Learning for the Prediction and Prevention of SRC
Of the 24 studies included in the review, 10 focus on the use of machine learning in predicting and preventing SRC (see Appendix 1). The studies in this management phase of SRC are diverse in terms of methodological approach, sample and aim. Three of the 10 studies focus on the role of biomarkers in the prediction of SRC (Helfer et al., 2014; Seeger et al., 2020; Fedorchak et al., 2021). For instance, Seeger et al. (2020) studied youth ice hockey participants (aged 12–17 years) to explore cytokine alterations following pediatric SRC. Their aim was to determine whether a specific cytokine profile could predict symptom burden and time to return to sports (RTS). Using saliva cytokine samples, they determined that there is evidence of alterations in saliva cytokine profiles that are associated with increased symptom burden following SRC. Seeger et al. (2020) therefore provide support for the feasibility of exploring alterations in salivary cytokines in adolescent athletes' post-SRC and the potential to associate increased symptom burden and recovery.
Other studies attempt to use machine learning to develop new prediction models for SRC (Bergeron et al., 2019; Gabler et al., 2020; Castellanos et al., 2021). For example, in their study, Castellanos et al. (2021) carried out baseline assessments of 15,682 participants and monitored them for SRC during the subsequent season. This data was then used to develop a risk-stratification tool capable of identifying athletes with an elevated risk of SRC. The tool is based on a predictive algorithm for SRC. Castellanos et al. (2021) contribution is clinically important because it represents the first successful attempt to predict SRC using baseline data from athletes.
Another notable exploratory contribution in this category of the reviewed literature is that of Zhuang et al. (2021), who use machine learning and simulation of force impacts to develop a smart helmet prototype with real-time sensors that record blunt-force impacts to the head. In their study, Zhuang et al. (2021) found that the smart helmet prototype successfully achieved a real-time sensing of concussive events and concluded that the novel smart helmet system could serve as an early-stage intervention strategy during a concussive event.
Overall, only one of the 10 studies categorized as prediction and prevention apply a deep learning approach (Domel et al., 2021), while the remaining 9 are conducted using machine learning algorithms, such as RReliefF (Seeger et al., 2020), XGBoost (DiCesare et al., 2020), Support Vector Machine (SVM) (Helfer et al., 2014; Reynolds et al., 2018) and Adaboost (Gabler et al., 2020).
Machine Learning and the Diagnosis and Classification of SRC
Eleven of the 24 studies included in the review examine how machine learning can be used in the diagnosis and classification of SRC (see Appendix 1). Several studies are devoted to using machine learning to create tools that classify SRC more accurately than the current approaches (Cai et al., 2018; Visscher et al., 2019; Bazarian et al., 2021; Ferris et al., 2021). Some studies also attempt to develop novel diagnostic tools to rapidly and objectively detect SRC (Tremblay et al., 2017; McNerny et al., 2019; Shim et al., 2020). One example is the study conducted by McNerny et al. (2019), which uses a unique combination of simple EEG measurements paired with advanced quantitative analysis in order to develop a highly accurate injury assessment tool. Studies such as those by Tremblay et al. (2017); McNerny et al. (2019) and Shim et al. (2020) represent valuable contributions to the field, in that developing tools that facilitate an accurate diagnosis of SRC is of the utmost importance, especially as many athletes and coaches under-report symptoms and “believe it is fine to play through a concussion even with background knowledge of the potential dangers” (McNerny et al., 2019, p. 6). The metabolic crisis that occurs in the brain following a concussion makes the brain vulnerable to subsequent injuries that could result in chronic brain disease. As McNerny et al. (2019, p. 6) highlight: “the period of vulnerability is unknown as it depends on many factors, such as severity, but it is clear that a subsequent injury occurring prior to full physiological recovery can be detrimental to the brain.” Therefore, it is important to create a way of measuring the initial injury and the recovery of that injury.
Two of the studies on the diagnosis and classification of SRC engage in deep learning approaches (Wu et al., 2019; Thanjavur et al., 2021). In their study, Wu et al. (2019) developed a trained neural network that is capable of assimilating impact-response samples to iteratively improve accuracy. Together with sensors that measure the impact of kinematics on head collisions, Wu et al. (2019) technique may enable a sophisticated head injury model to produce region-specific brain responses. Wu et al. (2019) and Thanjavur et al. (2021) studies showcase novel techniques that may offer clinical diagnostic values to SRC, and in Wu et al. (2019) case, even facilitate head impact sensors in concussion detection via a mobile device. This could be a valuable way of mitigating the millions of SRC incidents worldwide every year. The remaining 9 studies [excluding those of Wu et al. (2019) and Thanjavur et al. (2021)] are based on machine learning algorithms, such as TotalBoost (McNerny et al., 2019), Support Vector Machine (SVM) (Falcone et al., 2013; Tremblay et al., 2017; Raji et al., 2020), or Kohenn's Self-Organizing Map (SOM) (Visscher et al., 2019).
Machine Learning in the Treatment and Rehabilitation of SRC
Only 3 of the 24 analyzed studies focus on how machine learning can contribute to the treatment and rehabilitation of SRC (Cao et al., 2008; Goswami et al., 2016; Bohsra et al., 2019). Both Cao et al. (2008) and Goswami et al. (2016) investigated brain abnormalities following concussive episodes and applied a SVM classifier to train and test multichannel EEG (Cao et al., 2008) or rs-fMRI (Goswami et al., 2016) data. Their findings, which are promising, could help clinicians with more accurate return-to-sport participation criteria (Cao et al., 2008) or to predict recovery following a SRC event (Goswami et al., 2016).
Bohsra et al. (2019) demonstrate the use of machine learning as a methodological approach to utilizing the neurophysiological markers found in group-level studies relating to the detection of SRC in individual athletes. They identify a set of features that can help to accurately detect past concussion in up to 81% of retired athletes. Through this machine learning approach, Bohsra et al., 2019 were able to correctly and accurately classify past concussion even though the concussed athlete received their last injury up to “45 years ago, which is far beyond the suggested time frame of a few months for symptom resolution” (Bohsra et al., 2019 p. 1498). Bohsra et al. (2019) represents a first report of machine learning based analysis for the assessment of SRC history decades after injury, which is a significant step in the identification of chronic effects of concussion in athletes and an important factor in the rehabilitation and treatment of long-term health issues following SRC.
Both Bohsra et al. (2019) and Goswami et al. (2016) findings have potential applications for treatment and rehabilitation, as well as for the diagnosis and classification of SRC. As Bohsra et al., 2019 ( p. 1498) state, their results: “can be used by health professionals if machine learning tools like the one described here are adopted as diagnostic aids in the future.” Goswami et al. (2016, p. 10) also highlight that their work suggests “the potential utility of machine learning for the diagnosis of brain injury following concussions.”
Discussion
The results of this scoping review suggest that although, the use of machine learning for athlete injury prevention includes a plethora of studies (Juhiainen et al., 2021; Van Eetvelde et al., 2021), machine learning and the management of SRC is a novel and somewhat underdeveloped research field. Most of the studies in the review focus on predicting and preventing SRC (n = 10) or diagnosing and classifying SRC (n = 11), while only 3 studies explore the treatment and rehabilitation of SRC (see Table 1 and Appendix 1). From an athlete welfare perspective this makes sense, as one would prefer to prevent SRC from ever happening, especially in youth sports. The focus on developing accurate tools that coaches and athletes can use to diagnose SRC appears to be especially important, as several studies highlight the lack of objective tools to measure and classify SRC symptoms rapidly (Shim et al., 2020; Thanjavur et al., 2021).
The discussion is guided by the second purpose presented for this scoping review in the introduction, which is to ascertain where there are gaps in the existing research and identify recommendations for future research. Therefore, we first present an overview of the empirical study characteristics of the reviewed literature (who is the research on machine learning for the management of SRC based on?), and secondly identify knowledge gaps and the need for future research.
Study Context and Sample: Who Is the Research on Machine Learning in the Management of SRC Based on?
Current research on machine learning and the management of SRC is mainly conduced on male, adolescent or adult athletes from team sports in North America. As Table 2 demonstrates, only one article has a female only sample. This is DiCesare et al. (2020) study of female soccer players in North America. While some studies have gender mixed samples, many of these report a predominately male sample. For instance, Seeger et al. (2020) report a sample of 29 male athletes and 7 female athletes. The absence of studies on female athletes in research on machine learning in the management of SRC reflects the sex data gap in sport and exercise science research field. In a recent meta-study of medical and physiological sport and exercise research, Cowley et al. (2021) found that out of 5,261 publications and 12,511,386 participants, 63% of publications included both male and female participants, 31% included males only, and 6% included females only. In these studies there was a total of 8,253,236 (66%) male participants and 4,254,445 (34%) female participants, across all journals (Cowley et al., 2021). In other words, while women are included as participants in sport science research generally, and SRC injury research specifically, the numbers of female participants are not equal to the quantity of male participants. The sex data gap does, however, appear to be more extensive in studies of machine learning for the management of SRC than in the sport and exercise field generally (Cowley et al., 2021). This may be explained by the relatively small number of machine learning studies in the SRC research field (n = 24).
Table 2. Study characteristics of the reviewed literature.
In terms of age, none of the studies in this scoping review report a sample of children (age 3–5) or youth (6–11) athletes. Five studies use data from adolescent athletes (age 12–18) and four studies report on adult athlete samples (see Tables 1, 2). However, 10 studies do not report the age of the athletes in the sample at all (e.g., Falcone et al., 2013) and/or provide insufficient information about age. For instance, Visscher et al. (2019) report the median age, while Ferris et al. (2021) present the mean age in their samples. This variability in the quality of reporting an important variable such as age can create problems for the generalisability of the findings.
Team sports dominate the data of the analyzed studies. These sports include American football, ice hockey, soccer, basketball, lacrosse, rugby, volleyball, water polo and handball. The most prominent team sport is American football, with seven studies dedicated to this sport (see Table 2), as well as being included in 9 of 11 studies with mixed sports (Cao et al., 2008; Helfer et al., 2014; Tremblay et al., 2017; Reynolds et al., 2018; Bergeron et al., 2019; Wu et al., 2019; Raji et al., 2020; Bazarian et al., 2021; Castellanos et al., 2021). While many individual sports are included in the mixed samples, such as track and field (Helfer et al., 2014; Castellanos et al., 2021), mixed martial arts (Wu et al., 2019), snowboard (Visscher et al., 2019), skiing (Raji et al., 2020), tennis (Castellanos et al., 2021) and wrestling (Bergeron et al., 2019), there is only one individual sport with its own study: boxing (Falcone et al., 2013). In the studies with data from a mix of sports, it is often uncertain how many athletes from each sport are included.
Twenty-two of 24 studies included in the review have been conducted in North America (see Table 2). It is not uncommon for research fields to be predominately based in the Global North. Hence, we were not surprised to find that this is also the case for the machine learning and SRC field. Still, we were perhaps expecting more contributions from Europe. Only one study in our review is conducted in a European sports context (Visscher et al., 2019). Lastly, most of the studies in the review report relatively small samples of athletes. Two of the 24 studies report a sample of more than 501 athletes (see Table 2), and although no studies report <20 athletes, 10 studies have samples of between 20 and 50 athletes. This seems somewhat paradoxical, as the advantage of machine learning is that it can analyse large, complex and unstructured data and reveal patterns, trends and associations that traditional methodological approaches cannot (Agoston, 2019).
Publication Frequency and Scholarly Discipline
There has been a dramatic increase in articles published on the potential of machine learning for the management of SRC. As seen in Table 3, there is a large increase in the number of articles published from 2016 and onwards. Only one article was published before 2011 (Cao et al., 2008) and only two were published between 2012 and 2015 (Falcone et al., 2013; Helfer et al., 2014). Looking at the scholarly field represented in these studies, Table 3 shows that research on machine learning and SRC is mainly published in engineering and medical journals. Overall, six articles are published in engineering journals, seven articles in general academic journals, and the remaining 11 articles in medical journals.
Table 3. Reviewed literature by journal and year of publication.
Research Agenda and Current Knowledge Gaps
The use of machine learning for the management of SRC is a relatively new area of research. The surge in publication frequency on this topic since 2016 (see Table 3) indicates that this is a growing yet still under-explored research field. The potential of machine learning for the prevention, classification and treatment of SRC is still unclear. A main finding from our scoping review is that new methods of machine learning analysis for the management of SRC are now being actively explored in engineering and medical research. The small number of published papers on the use of machine learning in the management of SRC (n = 24) could also suggest that its value is not yet apparent to the research community on safeguarding in sports. From our review, it would seem that the potential value of employing machine learning in the management of SRC is still inconclusive.
As the field is novel and there is limited research on the topic, pointing to clear directions for future research is challenging. Nevertheless, we have three recommendations to make based on our analysis of the 24 articles included in this review. These are: (1) more investigations into the potential of deep learning, (2) more diverse data sets to be analyzed by machine learning algorithms and (3) a need for social scientific perspectives on the implementation of novel technologies and tools created through machine learning approaches.
The review demonstrates that there are almost no studies on the treatment and rehabilitation of SRC and that there are only a few select studies on the deep learning and management of SRC. As Cai et al. (2018, p. 16) note: “the superior performances of deep learning in concussion prediction and suggest its promise for future applications in biomechanical investigations of traumatic brain injury.” Still, it is worth noting that our search string has been specifically focused on machine learning and that there may be other research contributions on deep learning that have not been identified in this scoping review. Furthermore, the number of studies on evidence-based treatment and rehabilitation for SRC remains sparse, which might explain why there are so few studies on machine learning in the management of SRC that addressing this topic.
Secondly, there is a lack of diversity in the samples in the reviewed literature. A prominent limitation in current research is the lack of empirical samples from sporting contexts outside North America. As far as we can ascertain, no studies have been conducted with data from athletes in Asia, Africa, South America or Oceania, and there are almost no studies with data from Europe. This does not mean that there is no research on machine learning and the management of SRC from these continents, but could simply mean that it is not currently reported in English. Furthermore, there is a need for more investigations with data from female athletes. The lack of studies on female athletes is an issue that has been identified in other reviews of the SRC field (Tjønndal et al., 2022). Additionally, many of the studies focus on team sports, so there is less knowledge about machine learning and SRC in individual sports. In other words, there is a need for diversity in the athlete data analyzed by the machine learning algorithms. Otherwise, the findings from this research field will lose credibility among athletes and coaches over time.
Finally, the review has revealed that there are no social scientific investigations into machine learning and the management of SRC. As knowledge about the potential applications of machine learning and deep learning for the prevention, diagnosis and treatment of SRC progresses, there will be a salient need for social science perspectives in this research field. For instance, many of the studies included in the review focus on the development of practical tools for the different management phases of SRC. For instance, Wu et al. (2019) estimate that the practical implication of their machine learning technique may offer clinical diagnostic values that could facilitate head impact sensors in concussion detection via a mobile device. Zhuang et al. (2021) hope to develop a new type of sporting equipment - a smart helmet. Both these examples require athletes and coaches to embrace the use of novel technologies and to manage them effectively. However, current social scientific studies of athletes' and coaches' experiences of new sport technologies has demonstrated that athletes and coaches are often dissatisfied with the usability of new sports technologies and feel that such technologies do not meet their needs (Trabal, 2008; Luczak et al., 2020). Machine learning approaches to the management of SRC might also entail new ethical issues related to coaching athletes. While machine learning could bring new advances in terms of predicting the risk of athletes contracting SRC, when should coaches draw the line and take athletes off the field? Based on predictions from the machine learning algorithms of previous SRC events, a coach is likely to know when their athletes are at risk of serious injury. However, as coaches only keep their job if they can provide results and win competitions, the threshold for taking a star athlete out of an important game is likely to be high. With potential issues like these, social scientific investigations will be needed in the future, in order to understand what athletes and coaches need from novel procedures, equipment and tools based on machine learning approaches for the management of SRC.
Conclusion
The purpose of this review has been 2-fold: (1) to summarize and map the existing research literature on the use of machine learning in the management of SRC and (2) to ascertain where there are knowledge gaps in the existing research. This review concludes that most studies on machine learning and the management of SRC have either focused on the diagnosis and classification of SRC, or on the prediction and prevention of SRC. Very few studies on the treatment and rehabilitation of SRC have been identified in this review. The limited number of studies on machine learning in the treatment and rehabilitation of SRC might be explained by the broader lack of evidence-based studies on treatment and rehabilitation for SRC.
In the prediction and prevention category, studies are divided in two overarching categories. Some studies explore the role of biomarkers in predicting and preventing SRC, while others work toward the development of new prediction models for SRC based on various machine learning approaches and datasets. Contributions to the diagnosis and classification of SRC are mainly devoted to developing novel diagnostic tools to detect and classify SRC rapidly and more accurately than the currently available diagnostic approaches.
Most of the reviewed studies employ machine learning algorithms to analyse athlete injury data derived from male athletes playing team sports in North America. As Zhang et al. (2020, p. 1956) argue: “There is urgency for sex and gender sensitive concussion research, as well as a need to develop better and more accurate care strategies for wider demographics.” From our review, it would appear that this also applies to future research on the potential of machine learning for the management of SRC. Research on machine learning in the management of SRC mirrors the current sex data gap in sports and exercise research (Cowley et al., 2021). However, such narrow empirical data, at least in some cases, makes us question the transferability and usability of these findings. There is also a need for social science investigations of the usability and implementation of new tools and sports equipment aimed at athletes and coaches, given that previous studies indicate that new sports technology is often met with resistance and skepticism by practitioners (Trabal, 2008; Luczak et al., 2020).
Further studies using more diverse athlete data for analysis through machine learning algorithms may have the potential to fill important knowledge gaps in safeguarding and preventing injury. Addressing these gaps may also improve athletes' welfare by reducing the risk of sustaining SRC, as well as provide better treatment and rehabilitation from SRC. This review indicates that there could be potential in exploring the use of machine learning in the management of SRC. However, if the datasets used in this research continue to be based on narrow samples of athletes, there are considerable risks associated with the development of any new diagnostic and predictive tools for SRC emerging from this research. Today these risks appear to mainly affect the health and safety of female athletes of all ages. More diverse athlete data is thus needed to ensure that the development of such tools does not have unintended negative consequences for the health and safety of female athletes around the world.
Author Contributions
All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.
Funding
Nord University funded the open access publication of this article.
Conflict of Interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Publisher's Note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary Material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fspor.2022.837643/full#supplementary-material
References
Abrahams, S., Mc Fie, S., Patricios, J., Posthumus, M., and September, A. V. (2014). Risk factors for sports concussion: an evidence-based systematic review. Br. J. Sports Med. 48, 91–97. doi: 10.1136/bjsports-2013-092734
Agoston, D. V.. (2019). “Big data, artificial intelligence, and machine learning in neurotrauma,” in Leveraging Biomedical and Healthcare Data. Semantics, Analytics and Knowledge, eds F. Kobeissy, A. Alawieh, F. A. Zaraket, and K. Wang (Kidlington: Elsevier Inc.) 53–75. doi: 10.1016/B978-0-12-809556-0.00004-6
Arksey, H., and O'Malley, L. (2005). Scoping studies: Towards a methodological framework. Int. J. Soc. Res. Methodol. 8, 19–32. doi: 10.1080/1364557032000119616
Barker-Ruchti, N., Schubring, A., and Stewart, C. (2021). “Gendered violence in women's artistic gymnastics: a sociological analysis,” in Routledge Handbook of Athlete Welfare, ed M. Lang (Abingdon: Routledge). doi: 10.4324/9780429201745-8
Bazarian, J. J., Elbin, R. J., Casa, D. J., Hotz, G. A., Neville, C., Lopez, R. M., et al. (2021). Validation of a machine learning brain electrical activity–based index to aid in diagnosing concussion among athletes. JAMA Network Open 4:2. doi: 10.1001/jamanetworkopen.2020.37349
Benson, B. W., Hamilton, G. M., Meeuwisse, W. H., McCrory, P., and Dvorak, J. (2009). Is protective equipment useful in preventing concussion? A systematic review of the literature. Br. J. Sports Med. 43, 56–67. doi: 10.1136/bjsm.2009.058271
Bergeron, M. F., Landset, S., Maugans, T. A., Williams, V. B., Collins, C. L, Wasserman, E. B., et al. (2019). Machine learning in modeling high school sport concussion symptom resolve. Med. Sci. Sports Exerc. 51, 1362–1371. doi: 10.1249/MSS.0000000000001903
Bohsra, R., Dhindsa, K., Boursaile, O., Ruiter, K. I., Sonnadara, R., Samavi, R., et al. (2019). From group-level statistics to single-subject prediction: machine learning detection of concussion in retired athletes. IEEE Transact. Neural Syst. Rehabilit. Eng. 27, 1492–1501. doi: 10.1109/TNSRE.2019.2922553
Cai, Y., Wu, S., Zhao, W., Li, Z., Wu, Z., and Ji, S. (2018). Concussion classification via deep learning using whole-brain white matter fiber strains. PLoS ONE 13, e0197992. doi: 10.1371/journal.pone.0197992
Cao, C., Tutwiler, R. L., and Slobounov, S. (2008). Automatic classification of athletes with residual functional deficits following concussion by means of EEG signal using support vector machine. IEEE Transact. Neural Syst. Rehabil. Eng. 16, 327–335. doi: 10.1109/TNSRE.2008.918422
Carroll, L., and David, R. (2012). The Concussion Crisis: Anatomy of a Silent Epidemic. New York, NY: Simon & Schuster.
Castellanos, J., Phoo, C. P., Eckner, J. T., Franco, L., Broglio, S. P., McCrea, M., et al. (2021). Predicting risk of sport-related concussion in collegiate athletes and military cadets: A machine learning approach using baseline data from the CARE Consortium Study. Sports Med. 51, 567–579. doi: 10.1007/s40279-020-01390-w
Cenek, M., Hu, M., York, G., and Dahl, S. (2018). Survey of image processomg techniques for brain pathology diagnosis: challenges and opportunities. Front. Robotics AI 5, 120. doi: 10.3389/frobt.2018.00120
Clay, M. B., Glover, K. L., and Lowe, D. T. (2013). Epidemiology of concussion in sport: a literature review. J. Chiropr. Med. 12, 230–251. doi: 10.1016/j.jcm.2012.11.005
Cowley, E. S., Olenick, A. A., McNulty, K. L., and Ross, E. Z. (2021). Invisible sportswomen: the sex data gap in sport and exercise science research. Women Sport Phys. Activity Journal 29, 146–151. doi: 10.1123/wspaj.2021-0028
DiCesare, C. A., Green, B., Yuan, W., Diekfuss, J. A., Foss, K. D. B., Dudley, J., et al. (2020). Machine learning classification of verified head impact exposure strengthens associations with brain changes. Ann. Biomed. Eng. 48, 2772–2782. doi: 10.1007/s10439-020-02662-2
Dick, R. W.. (2009). Is there a gender difference in concussion incidence and outcomes? Br. J. Sports Med. 43, 46–50. doi: 10.1136/bjsm.2009.058172
Domel, A. G., Raymond, S. J., Giordano, C., Liu, Y., Yousefsani, S. A., Fanton, M., et al. (2021). A new open-access platform for measuring and sharing mTBI data. Sci. Rep. 11:87085. doi: 10.1038/s41598-021-87085-2
Echemendia, R. J., Meeuwisse, W., McCrory, P., Davis, G. A., Putukian, M., Leddy, J., et al. (2017). The sport concussion assessment tool 5th edition (SCAT5): Background and rationale. Br. J. Sports Med. 51, 848–850. doi: 10.1136/bjsports-2017-097506
Falcone, M., Yadav, N., Poellabauer, C., and Flynn, P. (2013). Using isolated vowel sounds for classification of mild traumatic brain injury. IEEE Int. Conf. Acoustics Speech Sig. Process. 2013:e6639136. doi: 10.1109/ICASSP.2013.6639136
Fedorchak, G., Ragnekar, A., Onks, C., Loeffert, A. C., Loeffert, J., Olympia, R. P., et al. (2021). Saliva RNA biomarkers predict concussion duration and detect symptom recovery: a comparison with balance and cognitive testing. J. Neurol. 268, 4349–4361. doi: 10.1007/s00415-021-10566-x
Ferris, L. M., Kontos, A. P., Eagle, S. R., Elbin, R. J., Collins, M. W., Mucha, A., et al. (2021). Predictive accuracy of the sport concussion assessment tool 3 and vestibular/ocular-motor screening, individually and in combination: a national collegiate athletic association-department of defense concussion assessment, research and education consortium analysis. Am. J. Sports Med. 49, 1040–1048. doi: 10.1177/0363546520988098
Gabler, L. F., Huddleston, S. H., Dau, N. Z., Lessley, D. J., Arbogast, K. B., Thompson, X., et al. (2020). On-field performance of an instrumented mouthguard for detecting head impacts in American football. Ann. Biomed. Eng. 48, 2599–2612. doi: 10.1007/s10439-020-02654-2
Gagnon, I., and Ptito, A. (2017). Sports Concussion. A Complete Guide to Recovery and Management. Boca Raton, FL: CRC Press. doi: 10.1201/9781315119328
Gardner, A., Iverson, G. L., Levi, C. R., Schofield, P. W., Kay-Lambkin, F., Kohler, R. M. N., et al. (2015). A systematic review of concussion in rugby league. Br. J. Sports Med. 49, 495–498. doi: 10.1136/bjsports-2013-093102
Goswami, R., Dufort, P., Tartagloa, M. C., Green, R. E., Crawley, A., Tator, C. H., et al. (2016). Frontotemporal correlates of impulsivity and machine learning in retired professional athletes with a history of multiple concussions. Brain Struct. Funct. 221, 1911–1925. doi: 10.1007/s00429-015-1012-0
Graham, R., Rivara, F. P., Ford, M. A., and Mason, C. (2014). Sports-Related Concussions in Youth. Improving the Science, Changing the Culture. Washington, DC: National Academies Press.
Helfer, B. S., Quatieri, T., Williamson, J., Keyes, L., Evans, B., Greene, W. N., et al. (2014). Articulatory dynamics and coordination in classifying cognitive change with preclinical mTBI. ISCA Pro. Interspeech 2014, 485–489. doi: 10.21437/Interspeech.2014-121
Jordan, M. I., and Mitchell, T. M. (2015). Machine learning: Trends, perspectives, and prospects. Science 349, 255–260. doi: 10.1126/science.aaa8415
Jauhiainen, S., Kauppi, J.-P., Leppänen, M., Pasanen, K., Parkkari, J., Vasankari, T., et al. (2021). New machine learning approach for detection of injury risk factors in young team sport athletes. Int. J. Sports Med. 42, 175–182. doi: 10.1055/a-1231-5304
Krahn, A.. (2021). “Psychological abuse in competitive and high-performance women's volleyball,” in Routledge Handbook of Athlete Welfare, ed M. Lang (Abingdon: Routledge). doi: 10.4324/9780429201745-9
Langlois, J. A., Rutland-Brown, W., and Wald, M. M. (2006). The epidemiology and impact of traumatic brain injury: a brief overview. J. Head Trauma Rehabilitation 21, 375–378. doi: 10.1097/00001199-200609000-00001
Luczak, T., Burch, B., Lewis, E., Chander, H., and Ball, J. (2020). State-of-the-art review of athletic wearable technology: What 113 strength and conditioning coaches and athletic trainers from the USA said about technology in sports. Int. J. Sports Sci. Coach. 15, 26–40. doi: 10.1177/1747954119885244
Malcolm, D.. (2020). The Concussion Crisis in Sport. Abingdon: Routledge. doi: 10.4324/9780429292408
Manning, A.. (2021). “Racial and ethnic diversity and racist ruptures in US youth soccer,” in Routledge Handbook of Athlete Welfare, ed M. Lang (Abingdon: Routledge). doi: 10.4324/9780429201745-22
McCrory, P.. (2013). Concussion management in 2014. What must we learn from the Zurich consensus statement? Aspetar Sports Med. J. 4, 542–552. Available online at: https://www.aspetar.com/journal/viewarticle.aspx?id=93#.YjrnKefML-h
McCrory, P., Feddermann-Demont, N., Dvorák, J., Cassidy, J. D., McIntosh, A., Vos, P. E., et al. (2017). What is the definition of sports-related concussion: a systematic review. Br. J. Sports Med. 51:11. doi: 10.1136/bjsports-2016-097393
McMahon, J., Zehntner, C., McGannon, K. R., and Lang, M. (2020). The fast-tracking of one elite athlete swimmer into a swimming coaching role: a practice contributing to the perpetuation and recycling of abuse in sport? Eur. J. Sport Soc. 17, 265–284. doi: 10.1080/16138171.2020.1792076
McNerny, M. W., Hobday, T., Cole, B., Ganong, R., Winans, N., Matthews, D., et al. (2019). Objective classification of mTBI using machine learning on a combination of frontopolar electroencephalography measurements and self-reported symptoms. Sports Med. Open 5:1. doi: 10.1186/s40798-019-0187-y
Mooney, J., Self, M., ReFaey, K., Elsayed, G., Chagoya, G., Bernstock, et al. (2020). Concussion in soccer: a comprehensive review of the literature. Concussion 5:3. doi: 10.2217/cnc-2020-0004
Munn, Z., Peters, M. D. J., Stern, C., Tufanaru, C., McArthur, A., and Aromataris, E. (2018). Systematic review or scoping review? Guidance for authors when choosing between a systematic or scoping review approach. BMC Med. Res. Methodol. 18, 2119–2126. doi: 10.1186/s12874-018-0611-x
Nery, M., Neto, C., Rosado, A., and Smith, P. K. (2021). “Welfare and the protection of young athletes: an analysis of the nature and prevalence of bullying behaviours in sport settings,” in Routledge Handbook of Athlete Welfare, ed M. Lang (Abingdon, Oxon: Routledge) 105–118. doi: 10.4324/9780429201745-12
Obermeyer, Z., and Emanuel, E. J. (2016). Predicting the future—-big data, machine learning, and clinical medicine. N. Engl. J. Med. 375, 1216–1219. doi: 10.1056/NEJMp1606181
Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., et al. (2021). The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 372:n71. doi: 10.1136/bmj.n71
Papathomas, A.. (2019). “Disordered eating in sport: Legitimized and stigmatized,” in Sport, Mental Illness and Sociology, ed M. Atkinson (Wagon Lane: Emerald Publishing Limited). doi: 10.1108/S1476-285420180000011007
Parent, S., Kotiuga, J., Vertommen, T., and Fortier, K. (2021). “Exploring the links between self-directed violence and child maltreatment and bullying in sport,” in Routledge Handbook of Athlete Welfare, ed M. Lang (Abingdon: Routledge). doi: 10.4324/9780429201745-14
Patel, U. K., Anwar, A., Saleem, S., Malik, P., Rasul, B., Patel, K., et al. (2021). Artificial intelligence as an emerging technology in the current care of neurological disorders. J. Neurol. 268, 1623–1642. doi: 10.1007/s00415-019-09518-3
Peters, M. D., Marnie, C., Tricco, A. C., Pollock, D., Munn, Z., Alexander, L., et al. (2020). Updated methodological guidance for the conduct of scoping reviews. JBI Evid. Synth. 18, 2119–2126. doi: 10.11124/JBIES-20-00167
Pfister, T., Pfister, K., Hagel, B., Ghali, W. A., and Ronksley, P. E. (2016). The incidence of concussion in youth sports: a systematic review and meta-analysis. Br. J. Sports Med. 50, 292–297. doi: 10.1136/bjsports-2015-094978
Raji, C. A., Wang, M. B., Nguyen, N., Owen, J. P., Palacios, E. M., Yuh, E. L., et al. (2020). Connectome mapping with edge density imaging differentiates pediatric mild traumatic brain injury from typically developing controls: proof of concept. Pediatr. Radiol. 50, 1594–1601. doi: 10.1007/s00247-020-04743-9
Reynolds, B. B., Stanton, A. N., Soldozy, S., Goodkin, H. P., Wintermark, M., and Druzgal, T. J. (2018). Investigating the effects of subconcussion on functional connectivity using mass-univariate and multivariate approaches. Brain Imaging Behav. 12, 1332–1345. doi: 10.1007/s11682-017-9790-z
Schmid, W., Fan, Y., Chi, T., Golanov, E., Regnier-Golanov, A. S., Austerman, R. J., et al. (2021). Review of wearable technologies and machine learning methodologies for systematic detection of mild traumatic brain injuries. J. Neural Eng. 18:4. doi: 10.1088/1741-2552/ac1982
Seeger, T. A., Tabor, J., Sick, S., Schneider, K. J., Jenne, C., La, P., et al. (2020). The association of saliva cytokines and pediatric sports-related concussion outcomes. J. Head Trauma Rehabil. 35, 354–362. doi: 10.1097/HTR.0000000000000605
Shim, V. B., Holdsworth, S., Champagne, A. A., Coverdale, N. S., Cook, D. J., Lee, T. R., et al. (2020). Rapid prediction of brain injury pattern in mTBI by combining FE analysis with a machine-learning based approach. IEEE Access 8:26350. doi: 10.1109/ACCESS.2020.3026350
Sindelar, B., and Bailes, J. E. (2017). Sports-Related Concussion: Diagnosis and Management (2nd ed.). Boca Raton, FL: CRC Press. doi: 10.1201/9781315153629
Thanjavur, K., Babul, A., Foran, B., Bielecki, M., Gilchrist, A., Hristopulos, D. T., et al. (2021). Recurrent neural network-based acute concussion classifier using raw resting state EEG data. Sci. Rep. 11:e91614. doi: 10.1038/s41598-021-91614-4
Tjønndal, A., Haudenhuyse, R., de Geus, B., and Buyse, L. (2022). Concussions, cuts and cracked bones: A systematic literature review on protective headgear and head injury prevention in Olympic boxing. Eur J Sport Sci. 22, 447–459. doi: 10.1080/17461391.2021.1872711
Tjønndal, A., and Wågan, F. (2021). Athletes' and coaches' attitudes toward protective headgear as concussion and head injury prevention: a scoping review. Front. Sports Active Living 3, e680773. doi: 10.3389/fspor.2021.680773
Trabal, P.. (2008). Resistance to technological innovation in elite sport. Int. Rev. Sociol. Sport 43, 313–330. doi: 10.1177/1012690208098255
Tremblay, S., Iturria-Medina, Y., Mateos-Pérez, J. M., Evans, A. C., and Beaumont, L. D. (2017). Defining a multimodal signature of remote sports concussions. Eur. J. Neurosci. 46, 1956–1967. doi: 10.1111/ejn.13583
Tricco, A. C., Lillie, E., Zarin, W., O'Brien, K. K., Colquhoun, H., Levac, D., et al. (2018). PRISMA extension for scoping reviews (PRISMA-ScR): Checklist and explanation. Ann. Intern. Med. 169:7. doi: 10.7326/M18-0850
Van Eetvelde, H., Mendonça, L. D., Ley, C., Seil, R., and Tischer, T. (2021). Machine learning methods in sport injury prediction and prevention: a systematic review. J. Experi. Orthopaedics 8:27. doi: 10.1186/s40634-021-00346-x
Ventresca, M., and McDonald, M. (2020). “Forces of impact: Critically examining sports' concussion crises,” in Sociocultural Examinations of Sports Concussions, eds M. Ventresca and M. McDonald (Abingdon: Routledge). doi: 10.4324/9780429028175
Visscher, R., Feddermann-Demont, N., Romano, F., Straumann, D., and Bertolini, G. (2019). Artificial intelligence for understanding concussion: Retrospective cluster analysis on the balance and vestibular diagnostic data of concussion patients. PLoS ONE 14, e214525. doi: 10.1371/journal.pone.0214525
White, A. J.. (2021). “The issue of athlete welfare and why tackling should be removed from physical education and school sport: #BanTackling”, in Routledge Handbook of Athlete Welfare, ed M. Lang (Abingdon: Routledge). doi: 10.4324/9780429201745-11
Wu, S., Zhao, W., Ghazi, K., and Ji, S. (2019). Convolutional neural network for efficient estimation of regional brain strains. Sci. Rep. 9:53551. doi: 10.1038/s41598-019-53551-1
Yengo-Kahn, A. M., Hale, A. T., Zalneraitis, B. H., Zuckerman, S. L., Sills, A. K., and Solomon, G. S. (2016). The sport concussion assessment tool: a systematic review. J. Neurosurg. 40:4. doi: 10.3171/2016.1.FOCUS15611
Zhang, J., Safar, K., Emami, Z., Ibrahim, G. M., Scratch, S. E., da Costa, L., et al. (2020). Local and large-scale beta oscillatory dysfunction in males with mild traumatic brain injury. J. Neurophysiol. 124, 1948–1956. doi: 10.1152/jn.00333.2020
Keywords: machine learning, sports-related concussion (SRC), deep learning, athlete welfare, sport injury prevention, sport technologies, sport and health
Citation: Tjønndal A and Røsten S (2022) Safeguarding Athletes Against Head Injuries Through Advances in Technology: A Scoping Review of the Uses of Machine Learning in the Management of Sports-Related Concussion. Front. Sports Act. Living 4:837643. doi: 10.3389/fspor.2022.837643
Received: 16 December 2021; Accepted: 14 March 2022;
Published: 20 April 2022.
Edited by:
Melanie Lang, Edge Hill University, United KingdomReviewed by:
Nathan D. Schilaty, University of South Florida, United StatesL. Syd Johnson, Upstate Medical University, United States
Copyright © 2022 Tjønndal and Røsten. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Anne Tjønndal, anne.tjonndal@nord.no; Stian Røsten, sitan.rosten@nord.no