Cyber-Physical System for Management and Self-Management of Cardiometabolic Health DOI: http://dx.doi.org/10.5772/intechopen.84262

the notion of the obesity paradox, i.e., increased BMI is associated with increased survival and reduced mortality among patients with cardiovascular risk [9]. A mortality study in adults showed that normal weight at the time of incident diabetes had higher mortality than adults who are overweight or obese [9, 10]. This apparent obesity paradox is best explained by insulin resistance which is the primary underlying factor in cardiovascular disease. Fat mass itself and insulin sensitivity (reciprocal of insulin resistance) may be the decisive link between mortality and weight status [10]. Moreover, a more recent study [11] confirms what most clinicians have felt for a long time that obesity or excess fat mass with associated insulin resistance is directly associated with shorter longevity and significantly increased risk of cardiovascular morbidity and mortality [11]. Furthermore, when a surrogate index of insulin resistance such as waist circumference is used to predict mortality, an elevated waistline was strongly predictive of an increased mortality rate among patients with cardiovascular disease [12], and it is an independent risk factor for CVD mortality [13].

3. Clearly, there is a need for healthy lifestyle interventions using self-management along with support team approach to prevent and treat noncommunicable diseases linked to overweight and obesity [14] to achieve cardiorespiratory fitness along with metabolic health with lowest possible insulin resistance. Effective programs and technology tools together are needed to support behavior change approaches toward healthy lifestyle. Recently, behavior change strategies have emphasized the need for feedback loops for self-directed behavior modification [15]. However, there is a paucity of personalized, time-adjusted, dynamic interventions supporting feedback control for health behavior interventions [16]. There is a needed tool to observe the slow changes of cardiovascular fitness and metabolic health metrics closely as a feedback of information for patient and primary care provider to facilitate selfdirected behavior change [3] as well as for guided therapy by the healthy lifestyle team. The hurdles to develop such behavior change models with dynamic feedback loops and corresponding supportive technology tools are (A) the lack of gold standard measures for important behavior constructs, (B) tools allowing for planning and executing dynamic changes of behavior, (C) a dynamic behavior change model using self-directed behavioral change strategies, and (D) outcome measures for optimization [3, 16].

Given the pandemic of overweight and obesity involving 1.9 billion people worldwide according to the World Health Organization, new and fresh ideas and approaches are needed. One of the goals of the current article is to introduce to researchers and clinicians a widely applicable toolset which could unleash the potential of the modern Digital Era and tackle the extraordinary burden of insulin resistance, obesity, prediabetes, metabolic syndrome, and type 2 diabetes on humanity. In this article, inspirations were taken from thoughts and works of giant and prodigious scientists of the twentieth century, the unbelievably huge potential of smartphone technologies, combined with the tremendous power of human networking through the Internet. With the current novel framework, we strive to use the minimum set of assumptions about the process and measurement.

Eugene Wigner (1902–1995) stated that there is a "miracle of the appropriateness of the language of mathematics for the formulation of the laws of physics" … which may appear to us with "unreasonable efficiency." The inspiration here is, why not use mathematical tools for the formulation of the applicable laws to the human energy metabolism such as the first and second laws of the thermodynamics

facilitate prevention and treatment of chronic noncommunicable disease, improve rehabilitation after acute cardiovascular illness, and facilitate needed behavior change for cardiometabolic risk reduction to improve cardiovascular as well as all-cause mortality. Computer-generated feedback may provide a framework for automation and self-improvement to meet daily goals of therapeutic efforts.

Based on our research in systems biology, a cyber-physical system (CPS) can be construed for noninvasively tracking, drawing trajectories, and indirectly measuring daily changes and predict the otherwise very-difficult- or impossible-to-measure slow changes of the daily state variables (SVs) of the metabolism and capture them for the first time noninvasively in freely moving humans in their natural

Components of CPS are (A) a management software tool (MST); (B) a metabolic health monitor (MHM) app; (C) software on MHM capturing biometric signals from sensors of heart rate, physical activity, and pulse oximetry from a smart watch; and (D) software on MHM capturing biometric signals related to body composition and hydration status from Ori Diagnostic Instruments' (ODI) patented apparatus for impedance spectroscopy. MHM is running ODI's proprietary self-adaptive individualized stochastic mathematical model of the human energy metabolism (SAM-HEM) [1–3] via cloud computing. Based on our published simulation studies, SAM-HEM is a suitable concept to capture daily changes of the following SVs: weight; fat mass; lean mass; protein mass; intracellular water mass; extracellular water mass; utilized macronutrient intake and substrate oxidation of carbohydrate, fat, protein; and the R-ratio (ratio of the daily lean mass change velocity divided by the daily fat mass change velocity) which could be used as a surrogate marker for insulin resistance. SAM-HEM is a self-learning algorithm with daily updates using the minimal variance Kalman filter/predictor to arrive at the best metabolic model fitting to the available measured data. The trajectories of SVs are displayed on MHM and MST with errors of calculations allowing for analysis of past events, tracking current metabolic events

real time, predicting metabolic changes in the future, and supporting self-

The challenges ahead are the following: The prevalence of obesity and type 2 diabetes (T2D) is ranked the highest in the USA and Mexico in the American continent [4]. The proportion of the population with abnormal glucose tolerance is 52.4% for the USA (14.4% T2D, 38% prediabetes) [5] and 33.5% for Mexico (14.1%

noncommunicable chronic diseases including the most prominent one, cardiovascular disease (CVD). CVD is substantially higher in individuals with unhealthy lifestyle characteristics, including obesity, prediabetes, diabetes, insulin resistance, metabolic syndrome, physical inactivity, poor diet, and cigarette smoking [7]. In forging the battle against these problems, I see the following paramount problems:

1. There is a certain degree of fatigue toward "dieting," "weight loss," and

2. The problem with targeting weight loss only is that it does not distinguish between the loss of adipose and lean tissues. Further, it intuitively contradicts

hearing the word "obese." According to a recent survey [8], many sufferers of obesity wanted to become "healthy" so they could be "fit" and "strong" and

Central to our mission in primary care is to fight the burden of

realize CPS to improve cardiometabolic health.

expressed the wish for general health.

T2D, 19.1% prediabetes) [6].

114

management as well as guided therapies. We envision also that the same smart watch can provide sufficient information to track cardiorespiratory fitness by estimated maximum oxygen uptake. Our innovation is to merge the assessment of metabolic fitness/flexibility measurements with the assessment of cardiorespiratory fitness and

environment outside of a metabolic laboratory setting.

Type 2 Diabetes - From Pathophysiology to Modern Management

when considering, for example, the fat balance, i.e., fat in minus fat out? Though indirect calorimetry already makes reference to these laws, the indirect calorimetry technology use is intricately connected to respiratory gas exchange measurements which are difficult to do with the daily routine of life. However, mathematical models can be created with input variables with easier realization in daily life such as weight and fat weight measurement with bioimpedance fat scale. Further, appropriately built mathematical models can provide indirect measurements of difficult-to-measure variables of the human energy metabolism like fat versus carbohydrate oxidation rate or changes of insulin resistance [1–3] and provide a solution to gain a special quantified insight into the fat and the entire energy metabolism. Currently, the computational model of the human energy metabolism (CM-HEM) [17] and its improved version [18] is considered the most complete. CM-HEM uses the three compartmental partitioning of the entire energy flow, centered around the major macronutrient energy stores: glycogen G, fat F, and protein P. Hall was able to test his model and found satisfactory agreement between the model predictions and the measured group averaged data from the Minnesota Study [17], as well as 50 other studies [18]. Although CM-HEM behaves appropriately for different groups of subjects, it is presently unclear whether individual subject responses can be predicted [19], and CM-HEM may be limited in its ability to provide precise information on an individual basis [19]. Further, CM-HEM uses food intake as an input rather than an output variable, and it would be particularly interesting to determine utilized food intake from body composition changes [19]. CM-HEM is neither linear nor recursive nor individualized to a particular subject, and therefore, it is not suited to performing recursive parameter identification of the human energy metabolism, nor is it able to perform inverse calculation of utilized energy intake. Further, insulin resistance is not considered in CM-HEM when in fact insulin resistance plays a crucial role influencing fat and carbohydrate oxidation rates and the entire dynamic of body composition change [1–3]. Obviously, individualized models are needed which can be tied to easily measurable input variables such as weight and fat weight and provide insight into the fat and nonfat energy balance and change of insulin resistance.

a posteriori results and (2) providing a posteriori estimation for the process variables such as lean mass L, glycogen mass G, fat mass F, protein P, intracellular water mass ICW, and extracellular water mass ECW. This will realize a dynamic statespace modeling connecting the measured variables with the process variables, never losing the measured reality and keeping full statistical knowledge about confidence intervals and other statistical properties of results. The beauty of the Kalman estimators is that they operate also as a predictor when no updated measurements are

Cyber-Physical System for Management and Self-Management of Cardiometabolic Health

DOI: http://dx.doi.org/10.5772/intechopen.84262

Central to the development of the noninvasive metrics for the human energy

accumulation and reduced capacity of fat oxidation and inflexibility in regulating fat oxidation combined with the increased propensity of glucose oxidation and glucose-induced suppression of fat oxidation [22]. Experimental weight perturbation showed concordant changes of the glucose vs. fat oxidation fraction in skeletal muscle [23]. The correlation between BMI/weight/body composition and insulin resistance measured, for example, with Homeostatic Model Assessment of Insulin Resistance (HOMA-IR), is well documented in the medical literature [24, 25]. It is increasingly recognized also that there is a dynamic correlation between changes of weight, fat weight, and insulin sensitivity/resistance changes. Building on the above observations and reviewing energy perturbation studies from the international literature, we observed also a high level of correlation between weight, fat weight, and HOMA-IR [26]. We found also that our newly defined R-ratio and Rw-ratio showed highly significant correlation with

HOMA-IR, and we proposed these measures as metrics for insulin resistance [2, 3, 26]. We recognized that monitoring R-ratio and Rw-ratio may give an important tool for monitoring changes of insulin resistance; we developed CPS for this purpose. We have provided the derivation of our formulas used in CPS in the Appendix. The lists of

The essential input parameters, weight Wk and fat weight Fk, are captured by the software on MHM from a "bathroom scale" performing measurements of body composition (weight, fat weight) and hydration status (intracellular and extracellular water mass) developed by Ori Diagnostic Instruments (ODI), which is a patented apparatus for impedance spectroscopy [27, 28]. Without calorie counting and just using the required input ΔWk, ΔFk, and EBk, the fat and nonfat energy balance can be estimated along with the weight-related alpha α^wk, the energy

lean mass-related alpha α^k, the energy density parameter for lean mass change ϱ^Lk, the lean mass-related R-ratio Rk, the nonprotein respiratory quotient Rnpk, and the

If steady-state equilibrium in the metabolism can be assumed and the total energy expenditure is known, then it is possible to calculate food fraction φk, total metabolized energy intake MEIk, fat intake FIk, and fat oxidation FOk in addition to

macronutrient calorie counting on designated calibration days (maybe every 2 weeks). This could improve accuracy and would allow deeper insight into the dynamics of SVs

If equilibrium state is uncertain, then we would recommend additional

With additional measurement of physical activity (PAE) energy expenditure via smart watch sensors and using the measured or calculated value of the basal metabolic rate BMRk (by either using Harris-Benedict formula or by actual measurement of BMRk with indirect calorimetry), the total energy expenditure can be obtained as

<sup>W</sup> k, the weight-related Rw-ratio Rwk, the

measured and derived variables in CPS are listed in the glossary.

density parameter for weight change ϱ^

fat-burning fraction χ<sup>k</sup> as in Eqs. (1)–(19).

<sup>W</sup> <sup>k</sup>, ϱ^Lk, Rk, Rwk, Rnpk, and χk.

in Eq. (20).

α^wk, α^k, ϱ^

117

of the metabolism.

metabolism is to have a novel metric for insulin resistance from energy flow point of view through the body. Insulin resistance is related to ectopic fat

provided.

A second insightful guidance to our approach comes from John von Neumann (1903–1957) for the fight against insulin resistance. During his time, he foresaw already that "science, as well as technology, will in the near and in the farther future increasingly turn to problems of structure, organization, information, and control." This raises the question, why not use Neumann's self-organization and system theory ideas to control and prevent insulin resistance and obesity? In this regard using certain universal principles for energy calculation, such as the principle of "least action/ stationary action," for example, in Lagrangian and Hamiltonian mechanics, can be instrumental in setting up suitable control equations or functionals, such as the Hamilton-Jacobi-Bellman equation in control theory for dynamic optimized control [20]. This Hamilton-Jacobi-Bellman equation can allow for dynamic optimization of the energy system to achieve the desired state in the shortest possible time with minimized efforts.

The third inspirational insight comes from Rudolf E. Kálmán (1930–2016) and his invention of the "Kalman filter." This is briefly a statistical tool with tremendously widely used successful applications to control a vast array of consumer, health, commercial, and defense products. According to Grewal [21], the Kalman Filter is possibly the greatest discovery in the twentieth century and made the moon landing among others possible. This raises the question, why not use the Kalman filter to estimate and predict fat mass change? A potential application of Kalman's minimum variance estimator and predictor could be twofold: (1) Updating the a priori estimation equations for the measurement variables (weight, fat weight) with

## Cyber-Physical System for Management and Self-Management of Cardiometabolic Health DOI: http://dx.doi.org/10.5772/intechopen.84262

a posteriori results and (2) providing a posteriori estimation for the process variables such as lean mass L, glycogen mass G, fat mass F, protein P, intracellular water mass ICW, and extracellular water mass ECW. This will realize a dynamic statespace modeling connecting the measured variables with the process variables, never losing the measured reality and keeping full statistical knowledge about confidence intervals and other statistical properties of results. The beauty of the Kalman estimators is that they operate also as a predictor when no updated measurements are provided.

Central to the development of the noninvasive metrics for the human energy metabolism is to have a novel metric for insulin resistance from energy flow point of view through the body. Insulin resistance is related to ectopic fat accumulation and reduced capacity of fat oxidation and inflexibility in regulating fat oxidation combined with the increased propensity of glucose oxidation and glucose-induced suppression of fat oxidation [22]. Experimental weight perturbation showed concordant changes of the glucose vs. fat oxidation fraction in skeletal muscle [23]. The correlation between BMI/weight/body composition and insulin resistance measured, for example, with Homeostatic Model Assessment of Insulin Resistance (HOMA-IR), is well documented in the medical literature [24, 25]. It is increasingly recognized also that there is a dynamic correlation between changes of weight, fat weight, and insulin sensitivity/resistance changes. Building on the above observations and reviewing energy perturbation studies from the international literature, we observed also a high level of correlation between weight, fat weight, and HOMA-IR [26]. We found also that our newly defined R-ratio and Rw-ratio showed highly significant correlation with HOMA-IR, and we proposed these measures as metrics for insulin resistance [2, 3, 26]. We recognized that monitoring R-ratio and Rw-ratio may give an important tool for monitoring changes of insulin resistance; we developed CPS for this purpose. We have provided the derivation of our formulas used in CPS in the Appendix. The lists of measured and derived variables in CPS are listed in the glossary.

The essential input parameters, weight Wk and fat weight Fk, are captured by the software on MHM from a "bathroom scale" performing measurements of body composition (weight, fat weight) and hydration status (intracellular and extracellular water mass) developed by Ori Diagnostic Instruments (ODI), which is a patented apparatus for impedance spectroscopy [27, 28]. Without calorie counting and just using the required input ΔWk, ΔFk, and EBk, the fat and nonfat energy balance can be estimated along with the weight-related alpha α^wk, the energy density parameter for weight change ϱ^ <sup>W</sup> k, the weight-related Rw-ratio Rwk, the lean mass-related alpha α^k, the energy density parameter for lean mass change ϱ^Lk, the lean mass-related R-ratio Rk, the nonprotein respiratory quotient Rnpk, and the fat-burning fraction χ<sup>k</sup> as in Eqs. (1)–(19).

With additional measurement of physical activity (PAE) energy expenditure via smart watch sensors and using the measured or calculated value of the basal metabolic rate BMRk (by either using Harris-Benedict formula or by actual measurement of BMRk with indirect calorimetry), the total energy expenditure can be obtained as in Eq. (20).

If steady-state equilibrium in the metabolism can be assumed and the total energy expenditure is known, then it is possible to calculate food fraction φk, total metabolized energy intake MEIk, fat intake FIk, and fat oxidation FOk in addition to α^wk, α^k, ϱ^ <sup>W</sup> <sup>k</sup>, ϱ^Lk, Rk, Rwk, Rnpk, and χk.

If equilibrium state is uncertain, then we would recommend additional macronutrient calorie counting on designated calibration days (maybe every 2 weeks). This could improve accuracy and would allow deeper insight into the dynamics of SVs of the metabolism.

when considering, for example, the fat balance, i.e., fat in minus fat out? Though indirect calorimetry already makes reference to these laws, the indirect calorimetry technology use is intricately connected to respiratory gas exchange measurements which are difficult to do with the daily routine of life. However, mathematical models can be created with input variables with easier realization in daily life such as weight and fat weight measurement with bioimpedance fat scale. Further, appropriately built mathematical models can provide indirect measurements of difficult-to-measure variables of the human energy metabolism like fat versus carbohydrate oxidation rate or changes of insulin resistance [1–3] and provide a solu-

Type 2 Diabetes - From Pathophysiology to Modern Management

tion to gain a special quantified insight into the fat and the entire energy

nonfat energy balance and change of insulin resistance.

shortest possible time with minimized efforts.

116

metabolism. Currently, the computational model of the human energy metabolism (CM-HEM) [17] and its improved version [18] is considered the most complete. CM-HEM uses the three compartmental partitioning of the entire energy flow, centered around the major macronutrient energy stores: glycogen G, fat F, and protein P. Hall was able to test his model and found satisfactory agreement between the model predictions and the measured group averaged data from the Minnesota Study [17], as well as 50 other studies [18]. Although CM-HEM behaves appropriately for different groups of subjects, it is presently unclear whether individual subject responses can be predicted [19], and CM-HEM may be limited in its ability to provide precise information on an individual basis [19]. Further, CM-HEM uses food intake as an input rather than an output variable, and it would be particularly interesting to determine utilized food intake from body composition changes [19]. CM-HEM is neither linear nor recursive nor individualized to a particular subject, and therefore, it is not suited to performing recursive parameter identification of the human energy metabolism, nor is it able to perform inverse calculation of utilized energy intake. Further, insulin resistance is not considered in CM-HEM when in fact insulin resistance plays a crucial role influencing fat and carbohydrate oxidation rates and the entire dynamic of body composition change [1–3]. Obviously, individualized models are needed which can be tied to easily measurable input variables such as weight and fat weight and provide insight into the fat and

A second insightful guidance to our approach comes from John von Neumann (1903–1957) for the fight against insulin resistance. During his time, he foresaw already that "science, as well as technology, will in the near and in the farther future increasingly turn to problems of structure, organization, information, and control." This raises the question, why not use Neumann's self-organization and system theory ideas to control and prevent insulin resistance and obesity? In this regard using certain universal principles for energy calculation, such as the principle of "least action/ stationary action," for example, in Lagrangian and Hamiltonian mechanics, can be instrumental in setting up suitable control equations or functionals, such as the Hamilton-Jacobi-Bellman equation in control theory for

dynamic optimized control [20]. This Hamilton-Jacobi-Bellman equation can allow for dynamic optimization of the energy system to achieve the desired state in the

The third inspirational insight comes from Rudolf E. Kálmán (1930–2016) and his invention of the "Kalman filter." This is briefly a statistical tool with tremendously widely used successful applications to control a vast array of consumer, health, commercial, and defense products. According to Grewal [21], the Kalman Filter is possibly the greatest discovery in the twentieth century and made the moon landing among others possible. This raises the question, why not use the Kalman filter to estimate and predict fat mass change? A potential application of Kalman's minimum variance estimator and predictor could be twofold: (1) Updating the a priori estimation equations for the measurement variables (weight, fat weight) with

One of the goals of this chapter is to demonstrate the feasibility of the concept of CPS for its main function which is to predict changes of insulin resistance and fat oxidation from serial measurements of weight and fat mass. Unfortunately, there is a paucity of published data with longitudinal observations and serial measurements of the measurable components of the energy metabolism including measuring markers of insulin resistance. A complete data set to study the insulin resistance and weight-fat weight relationship would require the following data: serial measurements of macronutrient energy intake (EI), total energy expenditure (TEE) and serial fat mass (F), and lean body mass (L) or weight (W) measurements. Very few trial data are published only with serial measurements of markers of insulin sensitivity or resistance. Nevertheless, we were able to identify a study suitable for our aim, which is to demonstrate the feasibility of our concept of CPS to track and predict SVs and markers of insulin resistance. Here we use published data from the study entitled "Effects of brief perturbation in energy balance on indices of glucose homeostasis in healthy lean men (EBPE) [29]."

For calculation of correlations between HOMA-IR and weight, fat weight, Rratio, Rw-ratio, fat-burning fraction χk, and nonprotein respiratory quotient Rnpk, I

Cyber-Physical System for Management and Self-Management of Cardiometabolic Health

For demonstration purposes, I plugged the mandatory input variables ΔWk, ΔFk, and EBk as well as the known ingested macronutrient calories CIk, FIk, and PIk into SAM-HEM algorithm with Kalman filter [1–3]. As a measure of goodness of fit of the metabolic model SAM-HEM, I calculated the predicted mean value and standard deviation of the modeling error, i.e., model-predicted value minus the

The input weight and fat weight data are shown in Figure 1a for Protocol A and in Figure 1b for Protocol B. The measured data points for fat mass are connected with MATLAB's Piecewise Cubic Hermite Interpolating Polynomial. The results of

(a) Daily changes of Rw-ratio, χk, and Homa-IR in protocol A. (b) Daily changes of Rw-ratio, χk, and Homa-

(a) Daily changes of Rnp and Homa-IR in protocol A. (b) Daily changes of Rnp and Homa-IR in protocol B.

Rw-ratio Rwk and fat-burning fractionation χ<sup>k</sup> for Protocols A and B are in Figure 2a and b, respectively. The measured data points for HOMA-IR are connected with MATLAB's Piecewise Cubic Hermite Interpolating Polynomial.

used MATLAB's corrcoef function.

DOI: http://dx.doi.org/10.5772/intechopen.84262

3. Results

Figure 2.

Figure 3.

119

IR in protocol B.

known trajectory of weight, fat weight, and lean mass.
