Statistical Medical Pattern Recognition for Body Composition Data Using Bioelectrical Impedance Analyzer

: Identifying patterns, recognition systems, prediction methods, and detection methods is a major challenge in solving different medical issues. Few categories of devices for personal and professional assessment of body composition are available. Bioelectrical impedance analyzer is a simple, safe, affordable, mobile, non-invasive, and less expensive alternative device for body composition assessment. Identifying the body composition pattern of different groups with varying age and gender is a major challenge in defining an optimal level because of the body shape, body mass, energy requirements, physical fitness, health status, and metabolic profile. Thus, this research aims to identify the statistical medical pattern recognition of body composition data by using a bioelectrical impedance analyzer. In previous studies, a pattern was identified for four indicators that concern body composition (e.g., body mass index (BMI), body fat, muscle mass, and total body water). The novelty of our study is the fact that we identified a recognition pattern by using medical statistical methods for a body composition that contains seven indicators (e.g., body fat, visceral fat, BMI, muscle mass, skeletal muscle mass, sarcopenic index, and total body water). The youth that exhibited the body composition pattern identified in our study could be considered healthy. Every deviation of one or more parameters outside the margins of the pattern for body composition could be associated with health issues, and more medical investigations would be needed for a diagnosis. BIA is considered a valid and reliable device to assess body composition along with medical statistical methods to identify a pattern for body composition according to the age, gender, and other relevant parameters.


Introduction
In recent years, people became more interested in assessing and diagnosing their health status. Nowadays, many medical devices are available, thereby allowing the assessment and screening of one's health. Identifying patterns, recognition systems [1], prediction methods [2], detection methods [3][4][5][6], and benchmark methods [7] to solve different medical issues is a major challenge [8,9]. For body composition there are few categories of devices for personal and professional assessment: dual-energy X-ray absorptiometry (DXA), magnetic resonance imaging (MRI), and bioelectrical impedance analyzer (BIA).
DXA is used to assess the bone mineral density and body composition. The body measurement must be taken by a licensed radiological technician, and a complete scan lasts for 5 minutes [10][11][12]. MRI is a non-invasive technology that produces three-dimensional (3D) images for soft body tissues. This technology does not use ionizing radiation and allows the detection of changes in the protons found in the water of the human body. Special software processes the image pixels. MRI is considered a reference method for body composition assessment along with DXA [12][13][14].
BIA is considered a valid method for the assessment of body composition, and its reliability could be influenced by several factors, such as device, operator, subject, and environment [15]. Furthermore, it is a simpler, safer, more affordable, mobile, non-invasive, and less expensive alternative than other devices or methods used for body composition assessment [16]. BIA allows the selection of standard or athletic mode, gender, age, and height. Nevertheless, statistically significant differences are observed among BIAs because of calibration, different electric current frequencies, and different numbers of electrodes [12,17].
The novelty of this research is that we will determine the pattern recognition for body composition data by using BIA and statistical medical methods for at least seven parameters (e.g., body fat, visceral fat, body mass index (BMI), muscle mass, skeletal muscle mass, sarcopenic index, and total body water). The medical technology evolves rapidly, and new functions are available for use. However, each person is unique because although their age, height, and weight are the same, their body shapes, body composition, energy requirements, physical fitness, health status, and metabolic profiles are different. Establishing a pattern for body composition related to the age and gender is necessary.
A synaptic overview of the different devices used to assess the body composition in the analyzed studies is presented in Tab. 1.  Many of the previously analyzed studies used Pearson correlation and Bland-Altman analysis, but some of them used the analysis of variance (ANOVA) and paired t-test [12,20,24].
The statistical expression for ANOVA is:  Finally, with previous data, determining the value of F is possible: MS-Mean square.
A synthetic view for all this operation to determine F by using analysis of variance (ANOVA) is shown in Tab. 2. The statistical significance of at least p < 0.05 is required for ANOVA and paired t-test according to the number of cases/subjects.
The paired t-test statistic value is calculated by using the following formula: m-Mean differences n-Sample size s-Standard deviation.
Also, these studies allow us to identify a pattern because analyzed data concerning body composition are presented in Tab. 3.
According to the data from other studies, we identified the patterns for youth by using the statistical data for body composition in Tab. 4.
Identifying the body composition pattern of individuals with different age and gender is a major challenge in defining an optimal level because of the differences in their body shape, body mass, energy requirements, physical fitness, health status, and metabolic profile. These patterns allow the identification of possible health issues or disorders very easily and encourage people to have a healthy lifestyle along with a very good quality of life.
Studies on different age categories with various numbers of subjects and genders were identified by reviewing literature concerning the topic, and their components were analyzed using different kinds of devices for body mass composition. For youth, we identified a pattern for four body composition parameters in previous studies. By using a BIA device and statistical medical methods, we aim to establish a pattern for body composition for youth that comprises more parameters to identify the healthy profile particular for that age.  This research is organized as follows. Section 2 presents the methodology. Section 3 presents the results collected by BIA and analyzed by statistical medical methods. Section 4 analyzes the results and identifies the pattern recognition for the body composition of males and females. Section 5 presents the conclusions and prospects for future work.

Materials and Methods
This research aims to identify the statistical medical pattern recognition for body composition data by using BIA.

Participants
The research subjects include students who provided written informed consent prior to the research. The protocol was approved by the University Ethic Research Committee. The body characteristics and age averages by gender are presented in Tab. 5. The BMI is a parameter that allows determining the body composition very easily: The normal index must be between 18.5 and 25 points, the people who are in this range are considered to have a normal weight, and their health status is usually optimal. A point that falls below 18.5 are considered underweight, and an individual health status could be affected in this situation. Values over 25 points suggests an overweight level and over 30 corresponds to obesity, that is, the health status of an individual is deteriorating with poor effects at the physical and physiological level. Along with the health issues, the people who are in these categories (over 25 or 30) are also affected at the psychological level, because they cannot perform their daily tasks. The BMI is moderately correlated to the level of body fat [27,28].

Materials
The subjects followed four hours of physical activities according to their curricula and additional 4 to 8 hours of extracurricular physical activities weekly. The assessments were made throughout the first and in the last week of the research during the same day and at the same hour.

Procedure
In this research, we used a BIA (Tanita MC-780 MA) to assess body composition with high-frequency current (50 Khz, 90 µA) and eight electrodes that allow the current to flow into the upper and lower limbs (tetrapolar). All subjects in the standard mode were selected. The assessment protocol by BIA is shown in Fig. 1. The descriptive statistics, including Pearson correlation and Bland-Altman analysis, were calculated on the basis of the collected data by using IBM SPSS version 26. To achieve statistical significance, the value was set at p < 0.05.
Pearson correlation formula was applied to determine the concordance between data sets. The value of r ranges between +1 and −1. If the correlation coefficient is −1, then a strong negative connection exists, and it is a perfect negative connection between the variables. If the correlation coefficient is 0, then there is no connection. If the correlation coefficient is 1, then a strong positive connection exists. A level of confidence of 95% was established by the Bland-Altman analysis, and we obtained a bias and lower and upper limits of agreement (LOA).

Results
In our research, data concerning body composition (body fat, visceral fat, BMI, muscle mass, skeletal muscle mass, sarcopenic index, and total body water) were collected. The collected data were analyzed separately for each gender because differences concerning body composition are observed, as shown in Tabs. 6 and 7. By applying ANOVA and paired t-test at each gender and for each parameter of the body composition between pre and post-test, the statistical significance of 10 out of 14 indicators were not achieved (Tab. 8). These results confirm that BIA is a reliable device in assessing body composition, and its results could be used at the benchmark to determine pattern recognition.
The average values for each body composition parameters did not show any significant differences between pre and post-test at males (Fig. 2) and females (Fig. 3). These aspects confirm that BIA had an excellent accuracy for assessing body composition.

Discussions
For Pearson correlation, at males, 6 out of 7 values were over 0.900 and that means a strong positive correlation and for one indicator the value was 0.890. For females, one indicator has a value of 0.755, and the six other indicators have values over 0.900, indicating a strong positive correlation. According to the data collected from BIA (Tanita MC 780 MA) and by applying Bland-Altman analysis (Figs. 2 and 3) and Pearson correlation, we determined the body composition pattern of the subjects involved in our study (males and females) and compared them with our previous studies. The body composition pattern of males is shown in Tab. 9, and the distribution of values is shown in Fig. 4.
The body composition pattern of females is shown in Tab. 10, and the distribution according to Bland Altman analysis is presented in Fig. 5.
For body fat (BF%), at males, there was obtained a value of 0.976 at Pearson correlation, that means a strong positive correlation; at Bland-Altman analysis the bias was −0.42, and the LOA were −2.06 (lower) and +1. 22 (upper). For females, a strong positive correlation was also recorded; the bias value was −0.71, and the LOA were −4.73 and +3.32. The visceral fat (VF%) for both genders exhibited a strong positive correlation (males-0.948, females-0.942). The Bland-Altman analysis for males shows that the lower and upper LOA were −1.44 and +1.54, respectively; whereas those for females were −0.57 and +0.44, respectively [17,24,25,[29][30][31][32]. Skeletal muscle mass of males    At BMI, the correlations were strong for both genders. The LOA for males and females at the Bland-Altman analysis was +0.46 and +1.24 and −0.98 and +0.84, respectively. The upper limit for the BMI of males is 26.08, which indicates a tendency for being overweight; the same situation is also observed in previous studies [26,32,33]. The correlations concerning the muscle mass and skeletal muscle mass for females were strong, and those for the muscle mass and skeletal muscle mass (0.890) of males are strong and moderate to strong, respectively. The LOA for the muscle mass and skeletal muscle mass of females were −3.08 and +4.53 and −2.86 and +2.97, respectively. For males The LOA for the muscle mass and skeletal muscle mass of males were +1.09 and 2.69 and +2.14 and +4.28, respectively [16,34,35].  At sarcopenic index, the bias for males and females was 0.25 and 0.10, respectively. The LOA for males and females was +0.32 and +0.88 and −0.55 and +0.75, respectively [36][37][38]. For total body water (TBW%), the Pearson correlation was strong for both genders, and the LOA for males and females was +1.93 and +3.97 and −2.01 and 2.34, respectively [18,[39][40][41].
This research is subjected to several limitations, including the number of participating subjects, the device settings, and the software algorithms used. The reason is that a wide range of BIA is available on the market, and differences are observed among them.

Conclusion
In previous studies, a pattern was identified for the four indicators that concern body composition (e.g., BMI, body fat, muscle mass, and total body water). The novelty of our study is the fact we identified a recognition pattern by using medical statistical methods for a body composition that contains seven indicators (e.g., body fat, visceral fat, BMI, muscle mass, skeletal muscle mass, sarcopenic index, and total body water). To statistically validate the pattern for the body composition obtained in our study for males and females, we used four different statistical methods (e.g., Bland-Altman analysis, paired t-test, ANOVA, and Pearson correlation). The limits of agreement allowed us to establish the margins for every analyzed indicator for each body composition. The statistical methods used confirmed the reliability of the device utilized in the research. Our results are in the margins of previous studies for both genders, except for the total body water (TBW), wherein the data were at a lower limit and below. The youth that possesses the body composition pattern identified in our study could be considered healthy. Every deviation of one or more parameters outside the margins of the pattern for body composition could be associated with health issues and more medical investigations would be needed for diagnosis. BIA is considered a valid and reliable device to assess body composition along with medical statistical methods to identify a pattern for body composition according to the age, gender, and other relevant parameters. Other studies that involve more subjects are needed to determine the pattern of body composition for different ages, weights, and physical activity levels.
Author Contributions: All authors contributed to writing, original draft preparation, conceptualization, designing, analysis, investigations, data analysis, review, and editing of the content. All authors read and agreed to the published version of the manuscript.

Funding Statement:
The authors received no specific funding for this study, and the APC was funded by "Ştefan cel Mare" University of Suceava, Romania. The authors gracefully thank "Ştefan cel Mare" University of Suceava for providing material and financial support.

Conflicts of Interest:
The authors declare that they have no conflicts of interest to report regarding the present study.