An Overview of Multi Agent System for Sports and Healthcare Industry

Playersmore often engage in excessive physical activities during exercise session as well as in the game session because results of the games highly depend over the performance of participants that can be degraded due to various factors current health status, injury history, exercise types and duration, training and game experience. A Multi agent System can analyze all these factors and the overall performance of the participants can be improved using feedback. In this paper, the role of the Artificial Intelligence, Expert System, Machine/Deep Learning/Neural Networks in the sports and healthcare industry will be explored. CONTACT Naveen Dalal naveen.dalal@ggdsd.ac.in Goswami Ganesh DuttaSanatanDharam College, Chandigarh, India. © 2020 The Author(s). Published by Oriental Scientific Publishing Company This is an Open Access article licensed under a Creative Commons license: Attribution 4.0 International (CC-BY). Doi: http://dx.doi.org/10.13005/ojcst13.0203.07 Article History Received: 18 March 2021 Accepted: 24 March 2021


Introduction
Multi Agent System can be used to analyze the sports data and its feedback can improve the way of team selection, diagnosis process, and decision making etc. As per Figure 1, data related to different sports can be composed and sorted out for better planning and decision making by management.
A Multi Agent System can be developed for the following sport types

Contact Sports
It requires physical contact between participants i.e. Football, Rugby, Hockey, Wrestling, Boxing etc.

Non-Contact Sports
In this, no physical contact between participants is essential i.e. Racing/Riding/Cycling, Swimming, Snooker etc.
It can contribute to the following areas (as shown in Figure 2) • Training and performance analysis • Injury Categorization and Risk Assessment • Sports Medicines • Automated diagnosis support • Artificial decision support system Automation can be done using Multi Agent System applications and large scale data can be analyzed within a minimum time interval. The analyzed data may be utilized to form the training datasets for future use.

Training and Performance Analysis
Each sport type has unique requirements and the coach selects and prepares the team accordingly but in the absence of an automated feedback system, the performance of the players cannot be refined as desired and it affects the results of the games.
Multi Agent System can execute the gap analysis by associating the requirements of the sport with the current performance of the participants.

Injury Categorization and Risk Assessment
As per the sports types following are the injuries may happen to participants:

Contact/Direct Injuries
It is caused by collision with any object.

Non-Contact/In-Direct Injuries
It is caused by the internal forces of the body i.e. overstretching, muscle strain and ligament sprain etc. Frequency of above-discussed injuries can be reduced using Multi Agent System by collecting the data about various facts like health status (mental/ physical), sports/injury types, history of coaching/ exercise/training/total experience in sports etc.
Finally, feedback can be provided to the stack holders through risk assessment w.r.t. sports.

Contribution of Expert System in Sports
The computer-guided system can help the sports industry in different ways and day by day researchers are developing artificial intelligencebased applications. The following section describes the contribution made by them for the same: N. J. Cronin et al. 7 used a deep learning approach to analyze the sports Kinematic. It uses 2D samples collected from multiple sources/locations/sports to build a training set. The outcome of this study can be further utilized for the training/prediction/ development of 3D analysis of sports Kinematic.
G. Kakavas et al. 8 studied the role and applications of AI in the domain of sports injuries and found that AI-based prediction models can be used for risk assessment of different injuries related to various sports. The study found the correlation between the type of injury/ sport/performance and health etc. The outcomes of this study can be further utilized to enhance the accuracy of existing prediction schemes.
Elliot B et al. 9 did a survey of expert systems for the healthcare domain and categorized them as Rulebased and Machine Learning based systems. The accuracy of these systems depends on the input rules/samples used for training purpose. The study also explores the current issues related to sample data collection, potential challenges for service

Sports Medicines
Players must use the medicines recommended by the medical team but sometimes players may misuse the drugs so there must be an expert system to keep the track of drug consumption and health recovery over an interval and the doping results. According to the feedback, recommended drugs may be changed.
Automated Diagnosis Support Figure 6 shows the automated diagnosis supportsystem. It can collaborate the various facts for diagnosis purpose and these are related to patient feedback, current health status as compared to medical history, possible health recovery alternatives and by computing all these parameters, an expert system can recommend the diagnosis plan and estimated time of health recovery. Figure 7 shows the artificial decision support system that can analyze the large scale data generated by multiple resources and this data can be reused as training datasets to refine the overall performance of the existing system. 1-6

Fig. 7: Artificial Decision Support System
providers/end users and it can be further utilized to develop the decision support system for the healthcare industry.
M. Hatamzadeh et al. 10 introduced a machine learning method to recover from the knee injury. It subdivides the input samples into different categories i.e. healthy samples and injury samples. The machine learning process the variations in given samples to build a time-frequency strategy for diagnosis purposes. Improvement in health reduces the variations in healthy and injury samples. Analysis indicates its performance in terms of high accuracy of diagnosis/assessment cost as compared to traditional approaches and it can be further used to develop the health recovery models for different sports injuries.
P. Sardar et al. 11 investigated that AI can improve the accuracy of existing healthcare applications by analyzing large scale clinical data (including Text and Images). This data can be used for robotic assistance, decision making/training modules etc. The analysis found some barriers for AI-based healthcare systems and these are related to operational cost/complexity of the expert system, the security of clinical/patient data and benchmarking/ validation of training datasets.
H. Ma et al. 12 analyzed that issues related to medical data mining in the sports domain and developed an AI-based simulation model to process the large scale data associated with sports medicine/injuries/ diagnosis etc. The analysis shows that the accuracy of data mining can improve the performance of traditional AI methods and it can be used to analyze the (sports) medical image data/time series feature learning.
R. Li et al. 13 developed a framework to process large scale sports medical data. It can analyze the effect of health devices being used in sports health and can categorize the different risk levels associated with sports. The analysis shows that the performance and health level of the sports team can be improved using the proposed scheme.
P. Phan et al. 14 explored the limitations of existing perdition models used for injury recovery. The study found the deficiencies and lack of validation over the sample data (walking ability). Analysis data can be used to develop highly accurate prediction models using regression.
G. Lebedev et al. 15 developed a framework for sports health care services. It can provide the data related to the diet plans, current health issues and remedies etc. As per experiments, collected data can be utilized to guide the diagnosis process, as well as accuracy of existing medical support system,which can be enhanced.
C. E. Pulmano et al. 16 19 investigated the disorders related to walking patterns and found some constraints for the diagnosis system (i.e. lack of standardized gait data and acquisition methods, kinematics and reaction forces over gait etc.) and there is need to sort out all these factors to increase the accuracy of the prediction system. N. U. Ahamed et al. 26 introduced a fuzzy logic-based scheme to analyze the running patterns. It conducts the speed readings through a single sensor and these values are used as input for fuzzy logic to determine running conditions. Experimental results show its accuracy in terms of the detection of different running levels (High/Medium/Slow).
P. Paliyawan et al. 27 developed an intelligent subordinate that forms a health metric by analyzing the movements of body parts during game/ exercise sessions and finally, a fitness function is used for validation and injury risk is reduced by recommending the specific movements to the players. Experimental results show that its prediction accuracy can be amended through the active association of the participants.
S. Noordin 28 explored the various AI-based solutions for sports medicine, knee injuries diagnosis and decision making etc. The study found some factors that can degrade the accuracy of prediction i.e. quality/size of samples for training, sample collection strategy (linear/random) availability of the patients during real-time experiments and knee condition (before and after injury) etc. Facts collected during this survey can be used to develop expert systems for sports medicine and diagnosis.
A. Naglah et al. 29 introduced a machine learningbased sports injury model that uses different input parameters (sport type/mental and physical health conditions/exercise routine) for prediction w.r.t. different games and participants. K-means nearest neighbor method was deployed for experiments and results show that the performance of players can be improved using feedback data and its accuracy depends over the input samples and it can be further tested using other algorithms (Random Forest/ Decision Trees).

Conclusion
In this paper, AI-based techniques for the sports industry and healthcare services were investigated. It highlights the relationship between predictions, risk assessment, diagnosis, decision making, injury/ sports type and its impact on the patient's health.
Researchers considered the limitations of traditional diagnosis methods and introduced different frameworks that can act as an intelligent assistant for end users.
Expert systems can predict the performance of the individual teams by analyzing the large scale sport's data using deep/machine learning algorithms and integration of these techniques with clinical data can reduce the computational overhead. Health metrics and datasets can be formed for perdition models and risk assessment.
Expert systems can also be used for recommendations of sports medicine, doping test and the tracking of misuse of sports drugs.
In the future, a JADE framework based multi-agent diagnosis will be introduced for the diagnosis sports injuries.