Skip to main content
Download PDF

The role of various physiological and bioelectrical parameters for estimating the weight status in infants and juveniles cohort from the Southern Cuba region: a machine learning study

BMC Pediatrics volume 24, Article number: 313 (2024) Cite this article

Abstract

Objective

The search for other indicators to assess the weight status of individuals is important as it may provide more accurate information and assist in personalized medicine.This work is aimed to develop a machine learning predictions of weigh status derived from bioimpedance measurements and other physical parameters of healthy infant juvenile cohort from the Southern Cuba Region, Santiago de Cuba.

Methods

The volunteers were selected between 2002 and 2008, ranging in age between 2 and 18 years old. In total, 393 female and male infant and juvenile individuals are studied. The bioimpedance parameters are obtained by measuring standard tetrapolar whole-body configuration. A classification model are performed, followed by a prediction of other bioparameters influencing the weight status.

Results

The results obtained from the classification model indicate that fat-free mass, reactance, and corrected resistance primarily influence the weight status of the studied population. Specifically, the regression model demonstrates that other bioparameters derived from impedance measurements can be highly accurate in estimating weight status.

Conclusion

The classification and regression predictive models developed in this work are of the great importance for accessing to the weigh status with high accuracy of younger individuals at the Oncological Hospital in Santiago de Cuba, Cuba.

Peer Review reports

Introduction

According to World Health Organization (WHO), the body mass index (BMI) quantifies the weight status of individuals. There are various classes differentiating the weight status: underweight (BMI < 18.49 Kg/m2), normal weight (BMI 18.50–24.99 Kg/m2), overweight (BMI 25.0–29.99 Kg/m2), class I obesity (BMI 30.0–34.99 Kg/m2), class II obesity (BMI 35.0–39.99 Kg/m2) and class III obesity (BMI > 40.0 Kg/m2) [1, 2]. Besides, this associations vary across global regions and for younger and elder groups [1, 2].

Bioimpedance is a technique used to measure the electrical impedance or resistance of biological tissues or fluids. It involves the application of an electrical current, usually through the skin, and the measurement of the resulting voltage [3,4,5]. Bioimpedance is a non-invasive technique that can provide useful information on various physiological parameters, making it a valuable tool in healthcare and fitness settings. Several physiological parameters can be measured by bioimpedance, including body composition, hydration status, cell membrane integrity, tissue health among other.

The quantification of total body water (TBW), which includes both extracellular water (ECW) and intracellular water (ICW), plays a crucial role in diagnosing various health conditions [6,7,8,9,10,11]. For instance, dehydration can be identified through separate measurements of TBW and fat-free mass (FFM), whereas overhydration indicates the presence of oedema in individuals with heart disease or lymphoedema and mastectomy.

In the context of renal patients undergoing haemodialysis, a retention of fluid between treatments is observed. Assessing the volume of this excess fluid is vital to adjust the ultrafiltration process and understand how the fluid loss is divided between ECW and ICW [10, 11]. In patients on maintenance haemodialysis, recent studies observed an increase in the ECM/ICM index related with risk of sarcopenia. Bioimpedance technique is a non-invasive method to measure all of these anthropometric, metabolic and bioelectric parameters, offering a promising alternative to traditional techniques [3,4,5,6,7,8,9,10,11,12].

In recent years, the issue of overweight children and adolescents in Latin America and the Caribbean has become increasingly prevalent. Currently, an estimated 3 in 10 children and adolescents between 5 and 19 years old have overweight in the region [13, 14]. In 2020, UNICEF, The World Bank and WHO estimated that in Latin America and the Caribbean, a 7.5% of children under 5 years old, representing about 4 million children, are classified as overweight [13, 14]. This is higher than the global average of 5.7% [13, 14]. The origin of overweight and obesity statuses in childhood are the consumption of sugary drinks, ultra-processed foods and the lack of physical activity [13, 14].

In order to solve this issue, UNICEF has been promoting initiatives, in collaboration with governments from across the region, to improve the nutritional status of the population, guide families and communities, and contribute to regulatory actions to change food environments [13,14,15,16,17]. In addition, to avoid overweight, UNICEF supports nutritional campaigns in several countries and stimulates actions for the promotion, protection, and support of breastfeeding from birth to two years of age [13, 14]. Furthermore, scientists have created a US-Latin American research agenda on child obesity, finding evidence of anthropological factors influencing on the child obesity problem [13,14,15,16,17].

Alternatively, machine learning is a subfield of artificial intelligence that involves the development of algorithms and statistical models that enable computers to learn from data and make predictions or decisions without being directly programmed [18,19,20]. In medicine, machine learning has the potential to make more accurate diagnoses, decision making and personalized treatment plans [18,19,20,21]. Despite the development of bioimpedance method, there is no direct association between the bioelectrical parameters with weight status estimation. As the WHO system have the lack of variation between global regions and age for estimating the weight status, further studies are required to solve these problems. In this context, the aim of this work is to provide a predictive model, based on classification and regression learner methods, as a complementary approach to weigh status evaluation of younger volunteers from the main Oncological Hospital of Santiago de Cuba, Cuba.

Methodology

We conducted a pilot random study at the Oncological Hospital in South Cuba, specifically in Santiago de Cuba. This is a public sector hospital specialized in cancer diseases. The study involved volunteers who were selected between 2002 and 2008 and aged between 2 and 18 years old. In total, 776 female and male volunteers were studied. This research followed the code of ethics, good medical and clinical practices established by the Health General Law of the Ministry of Public Health of Republic of Cuba (Number 41, 13 July 1983 and updated in 2010).

The research was evaluated and approved by the ethics committees and scientific councils of Provincial Blood Bank Renato Guitart, Oncological Hospital Conrado Benítez, Pediatric Hospital Juan Martinez Maceira and Pediatric Hospital Antonio María Béguez César. All relevant national regulations, institutional policies, and the Regional Committees for Medical, Health Research Ethics, and scientific council were in accordance with the tenets of the Helsinki Declaration. Additionally, parents of children and adult participants signed informed consent before starting the study.

The database collected in this study is not publicly available because it is still under study to extract more information that can provide valuable data for a better understanding of the behavior of bioelectrical parameters in healthy and diseased patients. However, the datasets are available from the corresponding author upon reasonable request.

Bioimpedance parameters were obtained by measuring using the standard tetrapolar whole-body configuration. The Bioimpedance analyzer used was the BioScan 98® model (Biológica Tecnología Médica S.L., Barcelona, Spain. URL: http://www.bl-biologica.es). Healthy volunteers participated in a previous fast for at least 3 h, with an empty bladder, and had not exercised or consumed alcohol in the previous 12 h. A 50 kHz frequency was used for the measurements. Adults used disposable pre-gelled Ag/AgCl electrodes model 3 M Red Dot 2560 (3 M, Ontario, Canada), while the paediatric sample used the 3 M 2248-50 Red Dot.

The study was performed in the morning by trained personnel in an air-conditioned room at 23 °C with an ambient humidity of 60–65%. To perform the measurement, subjects were placed in a supine position without clothing, without a pillow under their heads, with their arms separated 30° from the chest and their legs separated at an angle of 45° without contact between them, on a non-conductive surface. The electrodes were placed after cleaning the skin with 70% alcohol. The injector electrodes were placed medial to the dorsal surfaces of the hands and feet, close to the third metacarpal and metatarsophalangeal joints. Detector electrodes were placed between the distal epiphyses of the ulna and radius, at the level of the pisiform eminence, as well as at the midpoint between both malleolus respectively. The distance between the injector and detector electrodes was 5 cm.

To perform a serious AI study, machine learning involves problem formulation, dataset quality analysis, feature selection, and model generalization in the real world [18,19,20,21,22,23,24]. In this study, weight status is considered as the response, and bioelectrical and biological parameters of each volunteer constitute the features. With a large dataset of 393 individuals, we used a 95% cross-validation for training and 5% for validation to avoid the overfitting problem [23,24,25]. Initially, a feature selection of 24 features is made, followed by a down-selection under the simplest model premise for describing weight status. Various machine learning methods were used for classification and prediction of relevant bioelectrical and bioparameters. Further details will be provided along next sections.

Results and discussion

Weight status classification model using machine learning techniques

After revising the data, the cohort consisted of 393 infants and juveniles, including 184 females and 209 males. Within the cohort, 133 children are classified with normal weight, 128 overweight and 132 underweight based on age and sex. In this sense, the cohort is balanced according to sex and weight status. Other weight status classes were not encountered due to the age range. It is well-know that accessing to the weight status in preadolescents and early childhood is a problem that must deal with care due to the difference and the metabolic changes occurring during the growing [1, 2]. In this sense, we conducts a study in three age groups chronologically sored, namely 2 to 18 year old (2–18 age group compiling the entire cohort), 2 to 11 year old (2–11 age group) and12 to 18 year old (12–18 age group).

The weight status with classed normal weight, overweight and underweight is considered as the response for the classification model. Features considered for training include anthropometric parameters such as sex, height, weight, age, body surface area (BSA), body mass index (BMI), fat mass (MGC) according to Nhanes, total body water volume according to Kotler (TBWKotl), skeletal muscle mass (smmbia), extracellular mass (ECM), body density (Densbia), fat-free mass according to Siri (FFMS) [26],. In addition, other bioelectrical parameters such as resistance (r), impedance (Z), capacitive reactance (Xc), phase angle (phase), corrected resistance (rc), corrected capacitance (xc) and specific resistivity (\({\rho }_{00}\)). To complete the features, other metabolic parameters are considered intra (ICWKotl) and extracellular (ECWKotl) water according to Kotler, extra-intracellular index (ECWKotl/ ICWKotl), basal energetic cost (GBE), and basal metabolic index (IMB).

The best classification model is the bagged trees ensemble for the three age group studied. For the entire cohort (i.e., 2–18 age group), the accuracy of the classification model is 98.50% and 5.00 total misclassification cost. For the 2–11 age group the accuracy is 93.80% with a total misclassification cost of 5.00, while for the 12–18 age group is 98.70% and 4.00 misclassification cost. The receiver operating characteristic curve (ROC) provides information regarding the true positive versus the false positive rate for a selected trained classifier [27,28,29]. The red point represents the values of the false negative rate (FNR) and the true positive rate (TPR) for the classifier. Furthermore, the area under the curve (AUC) provide information of the accuracy of the classifier model [27,28,29]. A larger AUC value indicates better classifier accuracy and predictions [27,28,29]. Figure 1 displays the ROC for each weight class and year group. From Figure a), b) and c) the AUC = 1.00, assigning low false positive rate observations incorrectly to the positive class in concern. For the 2–11 age group (Figure d, e and f), the classification model assigns 2.00% of the false positive rate observations incorrectly to the positive normal weight and overweight classes with an AUC of 0.98. Analogously, 6.00% of the false positive rate observations are wrongly assigned to the positive underweight class with AUC = 0.98. Analogously, for the 12–18 age group (Fig. 1g, h and i), the model predicts accurately the weigh status with AUC = 1.00 and low false positive rate of 1.00% for underweight individuals.

Fig. 1
figure 1

Receiver operating characteristic curve (ROC) for each positive class considered: a), b) and c) corresponds to the complete cohort, d), e) and f) to 2–11, while g), h) and i) to 12–18 age group, respectively

Full size image

Confusion matrix provides detailed information about the accuracy predictions by comparing within specific classes [30, 31]. Figure 2 show the confusion matrix of each age group. For the entire age group (Fig. 2a) the model predict the normal weight status with an accuracy of 99.20%, the overweight and underweight statuses with a 97.70% and 98.50% of accuracy, respectively. In addition, the model misclassifies 0.80% underweight individuals as a normal weight, 2.30% of normal weight to overweight and 1.50% of overweight to the true class of underweight.

Fig. 2
figure 2

Confusion matrix for (a) 2–18, (b) 2–11 and (c) 12–18 age group, respectively

Full size image

Analogously, for the younger age range (Fig. 2b), the model predicts the normal weight status with an accuracy of 96.20%, while the overweight and underweight with 91.30 and 93.81% of accuracy. However, the model assigns wrongly 3.80% to the normal weight class of underweight individual, 8.70% of underweight to overweight and 6.30% of normal and overweight to underweight class. From Fig. 2c, the overweight class is predicted with the highest accuracy. In addition, the normal weight and underweight classes are also predicted with a 98.00% of accuracy assigning 2.00% wrongly weight status classes to normal and underweight class. These findings can be explained considering that some individuals can be on a transition between weight statuses.

Feature importance analysis is a critical step in the development and optimization of classification models, as it helps to identify the most relevant features and ensure the accuracy and reliability of the model, avoiding the common over fitting problems in regression learners [22,23,24]. Figure 3 displays the feature importance for each age groups. As it is shown in the Fig. 3a, the most important features of the entire age group are the fat-free mass (FFMS), reactance (Xc) and the corrected resistivity (rc). Among age groups, again the FFMS and Xc prevails as the most important characteristics. Considering the other insignificant characteristics, the main difference is that the thirds important characteristic in the 2–11 age group (Fig. 3b) is ICWKotl and rc in the 12–18 age group (Fig. 3c). Note that in the three cohorts the FFMS, Xc and rc play an important role accessing to the weight status of individuals.

The main results derived from the features importance analysis indicate that there are other characteristics derived from bioimpedance measurements that can be used alternatively for weight status predictions. For instance, the phase angle and the body density have more contribution to determine the weight status than the anthropometric body mass index (BMI). Reports concerning the failure of bioimpedance spectroscopy to determine the weight status have been described [32,33,34]. However, bioimpedance measurements are not superior to BMI as a predictor of overall adiposity in a general population [32, 33]. A study to determine obesity of 200 Taiwanese women with breast cancer by combining BIA and BMI revealed the underestimation of WHO criterion to state the cut-off of females with breast cancer [34]. On the other hand, both BIA and BMI methods can similarly detect normal and obese female individuals and are less accurate in determining underweight [32,33,34]. Our results suggest that there are other anthropometric and bioelectrical parameters (FFMS, Xc, rc) that can be used to support the diagnosis of weight status of pre- and adolescents individuals.

Fig. 3
figure 3

Feature importance of: (a) 2 to 18, (b) 2 to 11 and (c) 12 to 18 age group, respectively. The highest bar represents the most important characteristic of the classification model

Full size image

It is well-known FFM results in an important change of the human energy control of obese patients [35, 36]. Research shows that fat-free mass plays both an active and passive role in the body energy intake and requirements [13]. Increasing your fat-free body mass can be helpful for weight management [13]. The results supports those findings.

The role of fundamental characteristics for accessing the weight status

Given that FFMS, Xc, and rc are the primary characteristics that determine weight status in the studied cohort, this section focuses on developing regression models to predict these characteristics and assess their robustness in determining weight status. For the selected characteristic as the response, the data is divided into two parts: 95% for training and 5% for cross-validation. Figure 4a displays the predicted versus true response for the entire cohort, whereas Fig. 4b and c correspond to the 2–11 and 12–18 age groups, respectively. The observations are well-replicated by the chosen models, as indicated by all points lying near the control straight line. These results indicate good generalizability and high relevance of the identified features in explaining relative trends of the selected responses (FFMS, Xc, rc and \({\rho }_{00}\)).

Fig. 4
figure 4

Response vs. predicted plot of the selected responses of each cohort: (a) 2–18, (b) 2–11 and (c) 12–18 age groups

Full size image

Table 1 collects the results of the model for each response and their respective accuracy values. The accuracy is taken by considering the root mean square error (RMSE), R-squared (R2), mean square error (MSE) and the mean absolute error (MAE). As it is shown in Table 1, the main model describing these characteristic is the linear support vector machine (Linear SVM), except for \({\rho }_{00}\) where the best model is the rational quadratic (GPR) and quadratic support vector machine (Quadratic SVM) in the 2–11 and 12–18 age groups, respectively.

Table 1 Accuracy parameters of each model and age group categorized chronologically
Full size table

Figure 5a shows the trend of response parameters in relation to the weight status of the entire cohort, while Fig. 5b and c display the results for the 2–11 and 12–18 age groups, respectively. Along the cohorts, underweight status have the lowest FFMS and highest Xc and rc, while the overweight have the lowest Xc and highest \({\rho }_{00}\). In the case of normal weight, the parameters lies between overweight and underweight statuses. These results agree with the affirmation that underweight individuals generally have larger phase angle because Xc is directly proportional to the phase angle, while overweight individuals have lower phase angle associated to the body fluid imbalance [35]. In our study the mean value of the phase angle are 7.07 ± 0.74º, 7.06 ± 0.45º and 6.70 ± 0.58º for underweight, normal and overweight status, respectively, of the entire cohort. In addition, the true and predicted values are close each other, reaffirming the accuracy of the models. In general, the predictor models reproduces well the observed parameters in all age groups. This finding supports the affirmation that by controlling the fat-free mass one can monitoring the fluid unbalance of individuals and the weight status as a consequence [36, 37].

Fig. 5
figure 5

True and predicted anthropometric and bioelectrical parameters by weight status of each: (a) 2–18, (b) 2–11 and (c) 12–18 age group

Full size image

Figure 6 discloses the dependence of the above-mentioned responses upon gender of each age groups. In the entire cohort (Fig. 6a), there is no significant difference of rc and Xc between sex, while boys have larger FFMS and lower \({\rho }_{00}\) than girls. In contrast, for the early stage group (Fig. 6b), the difference is appreciated in all responses; boys have larger Xc and rc than girls, larger FFMS is observed for infant girls. For the adolescent group (Fig. 6c) the result shows that boys have larger FFMS and lower rc, Xc and \({\rho }_{00}\).

The weight gain in children is primarily dependent on fat-free mass, fat mass in children tends to decrease during their development stages [38, 39]. Only after the onset of puberty and as sex differences in overall and regional body composition become more noticeable, there is a visible increase in body fat percentage [40, 41]. Once children hit puberty, sex hormones lead to changes in body composition. During puberty, young females typically experience an increase in body fat, particularly in the hips and breasts, while young males tend to see an increase in muscle mass [42].

Fig. 6
figure 6

True and predicted anthropometric and bioelectrical parameters by gender class of: (a) 2–18, (b) 2–11 and (c) 12–18 age group

Full size image

Conclusions

In this work, a predictive classification and regression learner model is used to study the association of weight status as a possible risk of disease of infant-juvenile cohort from Santiago de Cuba, Cuba. We used 24 characteristics derived from bioimpedance measurements, including other physical parameters. The classification model shows that there are other characteristics different than body mass index that can be used as a predictors of weight status [fat-free mass (FFMS), reactance (Xc), corrected resistance (rc) and specific resistivity (\({\rho }_{00}\))]. The regression learner model was trained with the data and the abovementioned characteristics, predicting with high accuracy the weight status of the volunteers. The results concerning the variation of the above-mentioned characteristics against weight status and sex along the cohort agree with those reported in the literature. These predictive models developed in this work are of the great importance for accessing to the weigh status with high accuracy of younger individuals at the Oncological Hospital in Santiago de Cuba, Cuba.

Data availability

The datasets used during the current study available from the corresponding author on reasonable request.

References

  1. Di Angelantonio E, Bhupathiraju SN, Wormser D, Gao P, Kaptoge S, de González AB, et al. Continents Lancet. 2016;388(10046):776–86. https://0-doi-org.brum.beds.ac.uk/10.1016/S0140-6736(16)30175-1. Body-Mass Index and All-Cause Mortality: Individual-Participant-Data Meta-Analysis of 239 Prospective Studies in Four.

  2. Hill JO, Wyatt HR, Peters JC. Energy balance and obesity. Circulation. 2012;126(1):126–32. https://0-doi-org.brum.beds.ac.uk/10.1161/CIRCULATIONAHA.111.087213.

    Article  PubMed  PubMed Central  Google Scholar 

  3. Kyle UG, Bosaeus I, De Lorenzo AD, Deurenberg P, Elia M, Gómez JM, et al. Bioelectrical Impedance Analysis - Part I: review of principles and methods. Clin Nutr. 2004;23(5):1226–43. https://0-doi-org.brum.beds.ac.uk/10.1016/j.clnu.2004.06.004.

    Article  PubMed  Google Scholar 

  4. Kyle UG, Bosaeus I, De Lorenzo AD, Deurenberg P, Elia M, Gómez JM, et al. Bioelectrical Impedance Analysis - Part II: utilization in clinical practice. Clin Nutr. 2004;23(6):1430–53. https://0-doi-org.brum.beds.ac.uk/10.1016/j.clnu.2004.09.012.

    Article  PubMed  Google Scholar 

  5. Shah C, Whitworth P, Valente S, Schwarz GS, Kruse M, Kohli M, et al. Bioimpedance spectroscopy for breast cancer-related lymphedema assessment: clinical practice guidelines. Breast Cancer Res Treat. 2023;198(1):1–9. https://0-doi-org.brum.beds.ac.uk/10.1007/s10549-022-06850-7.

    Article  PubMed  Google Scholar 

  6. Stupin DD, Kuzina EA, Abelit AA, Emelyanov AK, Nikolaev DM, Ryazantsev MN, et al. Bioimpedance spectroscopy: basics and applications. ACS Biomaterials Sci Eng. 2021;7(6):1962–86. https://0-doi-org.brum.beds.ac.uk/10.1021/acsbiomaterials.0c01570.

    Article  CAS  Google Scholar 

  7. Carobbio ALC, Cheng Z, Gianiorio T, Missale F, Africano S, Ascoli A, et al. Electric Bioimpedance sensing for the detection of Head and Neck squamous cell carcinoma. Diagnostics. 2023;13(14):2453. https://0-doi-org.brum.beds.ac.uk/10.3390/diagnostics13142453.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Ward LC, Brantlov S. Bioimpedance basics and phase angle fundamentals. Reviews Endocr Metabolic Disorders. 2023;24(3):381–91. https://0-doi-org.brum.beds.ac.uk/10.1007/s11154-022-09780-3.

    Article  Google Scholar 

  9. Ge YZ, Ruan GT, Zhang Q, Dong WJ, Zhang X, Song MM, et al. Extracellular Water to total body water ratio predicts survival in Cancer patients with Sarcopenia: a Multi-center Cohort Study. Nutr Metab. 2022;19(1):1–9. https://0-doi-org.brum.beds.ac.uk/10.1186/s12986-022-00667-3.

    Article  CAS  Google Scholar 

  10. Jaffrin MY, Morel H. Body fluid volumes measurements by Impedance: a review of Bioimpedance Spectroscopy (BIS) and Bioimpedance Analysis (BIA) methods. Med Eng Phys. 2008;30(10):1257–69. https://0-doi-org.brum.beds.ac.uk/10.1016/j.medengphy.2008.06.009.

    Article  PubMed  Google Scholar 

  11. Zhou C, Lin X, Ma G, Yuan J, Zha Y. Increased Predialysis Extracellular to Intracellular Water ratio is Associated with Sarcopenia in Hemodialysis patients. J Ren Nutr. 2023;33(1):157–64. https://0-doi-org.brum.beds.ac.uk/10.1053/j.jrn.2022.03.004.

    Article  CAS  PubMed  Google Scholar 

  12. Pérez-Morales R, Donate-Correa J, Martín-Núñez E, Pérez-Delgado N, Ferri C, López-Montes A, et al. Extracellular water/total body water ratio as predictor of mortality in hemodialysis patients. Ren Fail. 2021;43(1):821–9. https://0-doi-org.brum.beds.ac.uk/10.1080/0886022X.2021.1922442.

    Article  PubMed  PubMed Central  Google Scholar 

  13. King AC, Perez-Escamilla R, Vorkoper S, Anand N, Rivera J. Childhood obesity prevention across borders: the promise of U.S.–Latin American research collaboration. Obes Rev. 2021;22:e13238. https://0-doi-org.brum.beds.ac.uk/10.1111/obr.13238.

    Article  PubMed  PubMed Central  Google Scholar 

  14. United Nations International Children’s Emergency Fund. 3 in 10 children and adolescents in Latin America and the Caribbean have overweight. Https://WwwUnicefOrg/Lac/En/Press-Releases/3-in-10-Children-and-Adolescents-in-Latin-America-and-the-Caribbean-Have-Overweight 2021. https://www.unicef.org/lac/en/press-releases/3-in-10-children-and-adolescents-in-latin-america-and-the-caribbean-have-overweight (accessed November 4, 2023).

  15. Pérez L, Vizcarra M, Hughes SO, Papaioannou MA. Food Parenting Practices and Feeding Styles and their relations with weight status in children in Latin America and the Caribbean. Int J Environ Res Public Health. 2022;19(4):2027. https://0-doi-org.brum.beds.ac.uk/10.3390/IJERPH19042027/S1.

    Article  PubMed  PubMed Central  Google Scholar 

  16. Rivera JÁ, De Cossío TG, Pedraza LS, Aburto TC, Sánchez TG, Martorell R. Childhood and adolescent overweight and obesity in Latin America: a systematic review. Lancet Diabetes Endocrinol. 2014;2(4):321–32. https://0-doi-org.brum.beds.ac.uk/10.1016/S2213-8587(13)70173-6.

    Article  PubMed  Google Scholar 

  17. Corvalán C, Garmendia ML, Jones-Smith J, Lutter CK, Miranda JJ, Pedraza LS, et al. Nutrition status of children in Latin America. Obes Rev. 2017;18:7–18. https://0-doi-org.brum.beds.ac.uk/10.1111/OBR.12571.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Tronstad C, Strand-Amundsen R. Possibilities in the application of machine learning on bioimpedance time-series. J Electr Bioimpedance. 2019;10(1):24–33. https://0-doi-org.brum.beds.ac.uk/10.2478/joeb-2019-0004.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Koh DM, Papanikolaou N, Bick U, Illing R, Kahn CE, Kalpathi-Cramer J, et al. Artificial intelligence and machine learning in cancer imaging. Commun Med. 2022;2(1):133. https://0-doi-org.brum.beds.ac.uk/10.1038/s43856-022-00199-0.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Wang B, Torriani M. Artificial Intelligence in the evaluation of body composition. Semin Musculoskelet Radiol. 2020;24(1):30–7. https://0-doi-org.brum.beds.ac.uk/10.1055/s-0039-3400267.

    Article  CAS  PubMed  Google Scholar 

  21. Trivizakis E, Papadakis GZ, Souglakos I, Papanikolaou N, Koumakis L, Spandidos DA, et al. Artificial intelligence radiogenomics for advancing precision and effectiveness in oncologic care. Int J Oncol. 2020;57(1):43–53. https://0-doi-org.brum.beds.ac.uk/10.3892/ijo.2020.5063.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Fernandez de Canete J, Roman M, De Santiago R. Artificial neural networks for adaptive control of profiled haemodialysis in patients with renal insufficiency. Expert Syst Appl. 2023;232:120775. https://0-doi-org.brum.beds.ac.uk/10.1016/j.eswa.2023.120775.

    Article  Google Scholar 

  23. Charilaou P, Battat R. Machine learning models and over-fitting considerations. World J Gastroenterol. 2022;28(5):605–7. https://0-doi-org.brum.beds.ac.uk/10.3748/wjg.v28.i5.605.

    Article  PubMed  PubMed Central  Google Scholar 

  24. Chakraborty P, Rafiammal SS, Tharini C, Jamal DN. Influence of Bias and Variance in Selection of Machine Learning Classifiers for Biomedical Applications. In Smart Data Intelligence: Proceedings of ICSMDI 2022 (pp. 459–472). Singapore: Springer Nature Singapore. https://0-doi-org.brum.beds.ac.uk/10.1007/978-981-19-3311-0_39.

  25. Bu C, Zhang Z. Research on overfitting problem and correction in machine learning. In Journal of Physics: Conference Series IOP Publishing. 2020;1693(1):012100. https://0-doi-org.brum.beds.ac.uk/10.1088/1742-6596/1693/1/012100.

  26. Siri WE. Body composition from fluid spaces and density: Analysis of methods Prospective Overview. 1956. https://scholarship.org/uc/item/6mh9f4nf.

  27. Bradley AP. The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern Recognit. 1997;30(7):1145–59. https://0-doi-org.brum.beds.ac.uk/10.1016/S0031-3203(96)00142-2.

    Article  Google Scholar 

  28. Lee J, van Es N, Takada T, Klok FA, Geersing GJ, Blume J, et al. Covariate-specific ROC curve analysis can accommodate differences between covariate subgroups in the evaluation of diagnostic accuracy. J Clin Epidemiol. 2023;160:14–23. https://0-doi-org.brum.beds.ac.uk/10.1016/j.jclinepi.2023.06.001.

    Article  PubMed  Google Scholar 

  29. Cho S, Kim YJ, Lee M, Woo JH, Lee HJ. Cut-off points between pain intensities of the postoperative pain using receiver operating characteristic (ROC) curves. BMC Anesthesiol. 2021;21:1–8. https://0-doi-org.brum.beds.ac.uk/10.1186/s12871-021-01245-5.

    Article  Google Scholar 

  30. Li J, Sun H, Li J. Beyond confusion matrix: learning from multiple annotators with awareness of instance features. Mach Learn. 2023;112(3):1053–75. https://0-doi-org.brum.beds.ac.uk/10.1007/s10994-022-06211-x.

    Article  Google Scholar 

  31. Xu J, Zhang Y, Miao D. Three-way confusion matrix for classification: a measure driven view. Inf Sci. 2020;507:772–94. https://0-doi-org.brum.beds.ac.uk/10.1016/j.ins.2019.06.064.

    Article  Google Scholar 

  32. Amani R. Comparison between Bioelectrical Impedance Analysis and Body Mass Index methods in determination of obesity prevalence in Ahvazi Women. Eur J Clin Nutr. 2007;61(4):478–82. https://0-doi-org.brum.beds.ac.uk/10.1038/sj.ejcn.1602545.

    Article  CAS  PubMed  Google Scholar 

  33. Willett K, Jiang R, Lenart E, Spiegelman D, Willett W. Comparison of Bioelectrical Impedance and BMI in Predicting obesity-related medical conditions. Obesity. 2006;14(3):480–90. https://0-doi-org.brum.beds.ac.uk/10.1038/oby.2006.63.

    Article  PubMed  Google Scholar 

  34. Liu LN, Miaskowski C, Wang JS, Chen SC, Chen ML. Accuracy of body Mass Index to determine obesity in women with breast Cancer: an observational study of Taiwanese sample. Int J Nurs Stud. 2010;47(8):994–1000. https://0-doi-org.brum.beds.ac.uk/10.1016/j.ijnurstu.2010.01.002.

    Article  PubMed  Google Scholar 

  35. Brunani A, Perna S, Soranna D, Rondanelli M, Zambon A, Bertoli S, et al. Body composition assessment using bioelectrical impedance analysis (BIA) in a wide cohort of patients affected with mild to severe obesity. Clin Nutr. 2021;40(6):3973–81. https://0-doi-org.brum.beds.ac.uk/10.1016/J.CLNU.2021.04.033.

    Article  PubMed  Google Scholar 

  36. da Silva BR, Orsso CE, González MC, Sicchieri JMF, Mialich MS, Jordao AA, et al. Phase angle and cellular health: inflammation and oxidative damage. Rev Endocr Metab Disord. 2023;24(3):543–62. https://0-doi-org.brum.beds.ac.uk/10.1007/s11154-022-09775-0.

    Article  CAS  PubMed  Google Scholar 

  37. Anja BW, Danielzik S, Dörhöfer RP, Later W, Wiese S, Müller MJ. Phase angle from bioelectrical impedance analysis: Population reference values by age, sex, and body mass index. J Parenter Enter Nutr. 2006;30(4):309–16. https://0-doi-org.brum.beds.ac.uk/10.1177/0148607106030004309.

    Article  Google Scholar 

  38. Wells JCK. The evolution of human fatness and susceptibility to obesity: an ethological approach. Biol Rev Camb Philos Soc. 2006;81(2):183–205. https://0-doi-org.brum.beds.ac.uk/10.1017/S1464793105006974.

    Article  PubMed  Google Scholar 

  39. Wells JCK. Sexual dimorphism of body composition. Best Pract Res Clin Endocrinol Metab Baillière Tindall. 2007;21(3):415–30. https://0-doi-org.brum.beds.ac.uk/10.1016/J.BEEM.2007.04.007.

    Article  Google Scholar 

  40. Kissebah AH, Krakower GR. Regional adiposity and morbidity. Physiol Rev. 1994;74(4):761–811. https://0-doi-org.brum.beds.ac.uk/10.1152/physrev.1994.74.4.761.

    Article  CAS  PubMed  Google Scholar 

  41. Pulit SL, Karaderi T, Lindgren CM. Sexual dimorphisms in genetic loci linked to body fat distribution. Biosci Rep. 2017;37(1):BSR20160184. https://0-doi-org.brum.beds.ac.uk/10.1042/BSR20160184.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Kaplowitz PB. Link between body fat and the timing of puberty. Pediatrics. 2008;121:S208–17. https://0-doi-org.brum.beds.ac.uk/10.1542/peds.2007-1813F.

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

The authors would like to thank the Ministry of Higher Education, Science and Technology (MESCYT for its initials in Spanish) of the Dominican Republic and the 2020-2A2-102 project for their invaluable assistance in the completion of this work. The authors also express our gratitude for the volunteer’s participation and contribution to the research.

Funding

No Funding to declare.

Author information

Authors and Affiliations

  1. Autonomous University of Santo Domingo (UASD), UASD Nagua Center, Santo Domingo, Dominican Republic

    Taira Batista Luna & Ana de la Caridad Román Montoya

  2. Autonomous University of Santo Domingo (UASD), San Francisco de Macorís Campus, Santo Domingo, Dominican Republic

    Jose Luis García Bello

  3. National Center for Applied Electromagnetism (CNEA), Universidad de Oriente CP 90500, Santiago de Cuba, Cuba

    Agustín Garzón Carbonell, Alcibíades Lara Lafargue & Héctor Manuel Camué Ciria

  4. Departamento de Física, Facultad de Ciencias Naturales y Exactas, Universidad de Oriente, Santiago de Cuba, CP 90500, CP, Cuba

    Yohandys A. Zulueta

Authors
  1. Taira Batista Luna
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Jose Luis García Bello
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Agustín Garzón Carbonell
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Ana de la Caridad Román Montoya
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Alcibíades Lara Lafargue
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Héctor Manuel Camué Ciria
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Yohandys A. Zulueta
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

José Luis García Bello: Conceptualization, Investigation, Formal analysis, Writing -review & editing, Writing - original draft, Taira Batista Luna: Investigation, Formal analysis, Agustín Garzón Carbonell: Data adquisition, Data analysis, Ana de la Caridad Román Montoya: Data adquisition, Data analysis, Supervision, Alcibíades Lara Lafargue: Conceptualization, Data adquisition, Data analysis, Héctor Manuel Camué Ciria: Conceptualization, Data adquisition, Data analysis, review & editing, Writing - original draft, Yohandys A. Zulueta: Conceptualization, Supervision, Data analysis, Writing -review & editing, Writing - original draft.

Corresponding authors

Correspondence to Taira Batista Luna or Yohandys A. Zulueta.

Ethics declarations

Ethical approval and consent to participate

Ethics approval was obtained from the ethics committees and scientific councils of Provincial Blood Bank Renato Guitart, Oncological Hospital Conrado Benítez, Pediatric Hospital Juan Martinez Maceira, and Pediatric Hospital Antonio María Béguez César, and written informed consent was obtained from all parents prior to their children’s inclusion in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Luna, T.B., Bello, J.L.G., Carbonell, A.G. et al. The role of various physiological and bioelectrical parameters for estimating the weight status in infants and juveniles cohort from the Southern Cuba region: a machine learning study. BMC Pediatr 24, 313 (2024). https://0-doi-org.brum.beds.ac.uk/10.1186/s12887-024-04789-w

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://0-doi-org.brum.beds.ac.uk/10.1186/s12887-024-04789-w

Keywords

BMC Pediatrics

ISSN: 1471-2431

Contact us