International Journal of Hypertension

International Journal of Hypertension / 2019 / Article

Research Article | Open Access

Volume 2019 |Article ID 7320365 | 13 pages | https://doi.org/10.1155/2019/7320365

Blood Pressure Classification Using the Method of the Modular Neural Networks

Academic Editor: Tomohiro Katsuya
Received08 Nov 2018
Revised27 Dec 2018
Accepted13 Jan 2019
Published23 Jan 2019

Abstract

In this paper, we present a new model based on modular neural networks (MNN) to classify a patient’s blood pressure level (systolic and diastolic pressure and pulse). Tests are performed with the Levenberg-Marquardt (trainlm) and scaled conjugate gradient backpropagation (traincsg) training methods. The modular neural network architecture is formed by three modules. In the first module we consider the diastolic pressure data; in the second module we use details of the systolic pressure; in the third module, pulse data is used and the response integration is performed with the average method. The goal is to design the best MNN architecture for achieving an accurate classification. The results of the model show that MNN presents an excellent classification for blood pressure. The contribution of this work is related to helping the cardiologist in providing a good diagnosis and patient treatment and allows the analysis of the behavior of blood pressure in relation to the corresponding diagnosis, in order to prevent heart disease.

1. Introduction

The learning ability of neural networks and their pattern classification characteristics are the reasons why these models can be of great importance for medical applications. Nowadays there are many approaches in intelligent computing, such as evolutionary computing, fuzzy systems [17] and neural networks [818], which are used in the areas of medicine [1722].

Hypertension that threatens to be present in most of the people of the world is a dangerous disease and leads to fatal consequences such as death and is a risk factor for people who suffer from it: obesity, diabetes mellitus, etc.

Hypertension is a global problem as it affects more than a billion people and causes more than ten million (avoidable) deaths each year. The only way to know if a person suffers from this disease is to constantly check the blood pressure and effectively diagnose and prevent this disease [23, 24].

Currently, there are several computer techniques that have been applied in medicine, such as neural networks and fuzzy systems to diagnose hypertension; by using these methods we can provide information about the factors and risks the patient may have.

The main contribution of this work is the proposed Arterial Hypertension Classification and Diagnosis model based on modular neural networks for disease prevention. In this way, the cardiologist with the help of the model may prescribe the necessary treatment to the patient since hypertension is a disease that can evolve without showing any symptoms; this is the reason it is also known as “the silent killer”.

In this work, an MNN model is used to classify the patient’s hypertension level, tests are also performed with the Levenberg-Marquardt (trainlm) and scaled conjugate gradient backpropagation (traincsg) training methods, and this neural network consists of three modules. In the first module we consider the diastolic pressure data; in the second module we use details of the systolic pressure; in the third module, pulse data is used. Therefore, we obtain the patient’s blood pressure through ambulatory blood pressure monitoring (ABPM); so far we have 300 records.

1.1. Overview of Related Works

Artificial neural networks with the back propagation learning algorithm to obtain the hypertension diagnosis were presented by Sumathi B. et al. [25]. The method was designed with eight risk factors: smoking, stress, family history, high cholesterol, etc., where the result of neural network classification shows if the patient suffers from arterial hypertension.

Huang S. et al. [26] presented a study to investigate the factors of Hypertension (HTN) and was implemented as a prediction model for 35-year-old people in a rural area of China, with a modular neural network, considering risk factors, such as socioeconomic status and education level.

Vilkov V.G. et al. [13, 27] presented a comparative study with models of daily blood pressure monitoring was performed in 34 apparently healthy subjects and 72 patients with arterial hypertension (AH). They compared the efficiency of diagnosis of latent AH using models based on artificial neural networks of different architectures.

Barman M. et al. [28] presented an intelligent system based on a fuzzy rule system, to diagnose heart diseases and the number of heart attacks; such fuzzy system has seven inputs and uses the Cleveland database.

Patil P. et al. [29] designed a sensor which measures the pulse and temperature of the patient and is controlled by a fuzzy system which shows the patient’s pulse via remote and sends a warning to relatives, doctors, or ambulances, in case it presents an emergency for patients.

Morsi I. et al. [30] presented a model to diagnose blood pressure. A group of 105 patients is used to design this model and another group with the same number of patients is used to test and thus check the efficiency of fuzzy systems in the field of medicine.

Hussein S. [31] analyzed the risk factors of hypertension and a model was designed for the prediction of rural residents over 35 years of age, considering several factors such as education level, sedentary work, and history of hypertension in the family.

Touyz M. R. et al. [32] presented an ANFIS system; methodology is designed to diagnose and compare an existing fuzzy expert system, regarding performance metrics accuracy and sensitivity.

1.2. Artificial Neural Networks

Neural networks are integrated by many interrelated components. A neural network can have a structure of multiple inputs and outputs; these systems operate similarly to the human brain. A neural network learns from input values, and this helps us learn about the human being [3335].

1.3. Hypertension

In Mexico, a large number of professional studies have been made with the idea of determining the prevalence of hypertension, defined as the frequency of the disease at a given time in a particular place. The most important studies use different methodological criteria, which make them difficult to compare. In these studies, different blood pressure levels are used to define this disease. They even propose an adequate standardization to measure it, which usually leads to overdiagnosis. The number of measurements made at the time of the survey, be it on the same day or on different days, impacts significantly the prevalence of the disease. Another study that was decided to be included on the findings reported is the study of heart diseases of San Antonio (ECSA), which includes Mexican-American population and has a branch in Mexico City: the study of diabetes in Mexico City (EDCM). In these studies, the prevalence and incidence of hypertension and other cardiovascular factors are also reported.

1.4. Arterial Hypertension

Hypertension may be essential (unknown etiology, but with hereditary background) or secondary (with demonstrable cause) and can also be isolated or as metabolic syndrome; this disease is incapacitating and deadly due to the damage caused to important organs: blood vessels, heart, kidney, and eyes. Normal levels of blood pressure are those below 139/89 mmHg; secondary hypertension can be suspected in young people younger than 35 years of age or when there is no hypertension history in the family or in the absence of a family. Treatment of hypertension helps reducing damage to organs or even reverses it if possible; this treatment may be a drug with antidepressant use or nonpharmacological treatment, which includes changes in hygienic-dietetic habits (reduction weight, stop smoking, and drinking alcoholic beverages).

1.4.1. Development of Systolic and Diastolic Hypertension

The risk of cardiovascular complications begins, apparently with blood pressure values of 115mmHg for systolic and 75 mmHg for diastolic. In the clinical area, several subtypes of hypertension determined by isolated elevations of systolic and diastolic, or the combination of both are used. These subgroups have their own natural history and present a different cardiovascular risk.

Isolated systolic hypertension (ISH) is common after 50 years of age, affecting nearly 50% of people between 50 and 59 years of age, reaching 90% in those over 80 years old. This subtype of hypertension reflects the increase in the stiffness of the aorta and great vessels without an increase in arteriolar resistance.

When there is an increase in arteriolar resistance combined with a lack of increased arterial stiffness, the isolated diastolic pressure subtype (IDP) occurs; this subtype predominates in people younger than 40, comprising almost 60% of the population [36].

1.4.2. Pulse Pressure

Hypertension includes calculating pulse pressure (PP), which is done by subtracting the diastolic pressure and systolic quantities [37, 38]. In the elderly, increased systolic blood pressure reflects an increase in the degree of stiffness of arteries; as a result, pulse pressure increases. This is related to an increased incidence of cardiovascular events. Blood pressure PP is closely related to the changes produced by age in people over 50, increasing diastolic coronary mortality rates, and after 60 they stop. To most people, pulse and systolic pressure values become the most important risk triggers.

Previously, the pulse was measured when the examiner or physician would sit comfortably on the right side to support the patient’s elbow and, with his right thumb, explore the antecubital fosse, where the brachial artery is. The patient’s arm reflex should be activated. When the thumb or finger of the examiner is correctly in place, he or she can raise or lower the patient’s forearm by varying the pressure applied to the artery, feeling the maximum pulse. The right thumb of the examiner can feel the patient’s carotid artery similarly, which he or she may feel by gently grasping the patient’s fingertips with theirs. The digital pulse can be counted exactly by simultaneously palpating the radial artery while the examiner supports the patient’s wrist. The femoral pulse of a child of tender age should be sought only while the leg is relaxed voluntarily.

The arteries pulsations provide information about heart rhythms and speed, arterial pulse differential (right and left limbs, or top and bottom), thrills (shudders), and waveform. [39, 40].

Bradycardia. The bradycardia term simply means slower rate. Bradycardia athletes: the heart of an athlete is much more powerful than a normal person, which allows their heart force a greater volume of blood with each beat, a large proportion of blood driven into the arterial tree with each beat probably produces sufficient circulatory reflexes to begin causing bradycardia.

Tachycardia. Tachycardia means rapid heart rate. The three causes of tachycardia are increased body temperature and heart stimulation by the sympathetic and toxic states of the heart. Increased heart rate of about 10 beats per minute for each degree Celsius increases body temperature, up to 41°C; at this temperature, the heart rate may actually decrease by increasing muscle wasting as a result of fever. Tachycardia causes hyperthermia and increases the frequency of the heart rhythm [4145].

1.5. Ambulatory Blood Pressure Monitoring

Nowadays modern laboratory methods often require outpatient-monitoring equipment to measure a variety of the indicators for blood pressure (BP), for 24 hours continuously. The biological rhythms are physiological functions and pathological alterations. The BP with an average heart rate of 72 beats per minute and 103.680 pulse waves is produced with corresponding changes in BP [46]. While at first the method was used in research studies, it is now increasingly used in clinical practice, as it provides additional data of measures from office and home. Moreover, only the ABMP can shed some light on symptomatic episodes occurring within 24 hours, either by raising or lowering the BP [47]. This means that it can be used not only for diagnosis of arterial hypertension (HA), but also to evaluate the frequency and severity of acute episodes of hyper- or hypotension. The ABMP is very useful to investigate the effects of new drugs for a period of 24 hours.

2. The Proposed Method

This section presents the proposed method for blood pressure classification, which consists of designing of modular neural networks (MNN) and the integration of responses of MNN with an average method. The main goals are to implement and find the best MNN architecture; the MNN consists of three modules; the first module is for the systolic pressure, the next module is the diastolic pressure, and in the last module we have pulse, this way we classify the arterial hypertension of a person.

Figure 1 illustrates the MNN structure, which considers the diastolic, systolic pressure, and the pulse for the MNN inputs, and in this case has 3 modules. The tests are performed by changing the number of layers that are between 1 and 3 and the number of neurons from 1 to 50, and in this way we obtained the responses of the MNN and integrate them with the average integration method and we obtained the classification of blood pressure.

In Figure 2, we present the data used for the classification of the arterial hypertension, where 300 patient samples are used for training all the modules in the modular neural network and we considered other 40 patients for tests with 45 records for each patient in the complete architecture.

Table 1 presents the parameters of the MNN that are manually changed to obtain the best architecture.


Number of LayersNumber of NeuronsEpochsLearning RateError GoalTraining Methods

1 to 31 to 505000.0010.01(i) Levenberg-Marquardt (trainlm)
(ii) Scaled Conjugate Gradient Back Propagation (traincsg).

Table 2 presents the classification of arterial hypertension according to the European guidelines.


CategorySystolicDiastolic

Optimal<120And<80

Normal120-129And/or80-84

High Normal130-139And/or85-89

Grade 1 Hypertension140-159And/or90-99

Grade 2 Hypertension160-179And/or100-109

Grade 3 Hypertension≥180And/or≥110

Isolated Systolic hypertension≥140And/or<90

3. Discussion and Results

The proposed method to classify the blood pressure of a patient was validated with tests performed on 16 patients and positive results were obtained for the MNN.

The results of the best MNN architecture are shown in Figure 3. For each of the modules of the MNN, the goal error was of 0.002 and 500 epochs were used; the number of neurons used was 14 in the first layer and 15 in the second layer.

Table 3 presents the results of the MNN with the “trainlm” method for the classification arterial hypertension.


No. PersonsNumber of NeuronsTimeSystolicDiastolicPulseClassification

Person114,1500:06:481187068Optimal

Person 214,1500:06:481097776Optimal

Person 314,1500:06:481127679Optimal

Person 414,1500:06:481207373Optimal

Person 514,1500:06:481468677Grade 1 Hypertension

Person 614,1500:06:481076392Optimal

Person 714,1500:06:481288397Normal

Person 814,1500:06:481126696Optimal

Person 914,1500:06:481307673Normal

Person 1014,1500:06:481237957Normal

Person 1114,1500:06:481386565High Normal

Person 1214,1500:06:481388474High Normal

Person 1314,1500:06:481237679Normal

Person 1414,1500:06:481146378Optimal

Person 1514,1500:06:481247972Normal

Person 1614,1500:06:481348489High Normal

Person 1714,1500:06:481257780Normal

Person 1814,1500:06:481066579Optimal

Person 1914,1500:06:481106879Optimal

Person 2014,1500:06:481237680Normal

Person 2114,1500:06:481157276Optimal

Person 2214,1500:06:481127378Optimal

Person 2314,1500:06:481227677Normal

Person 2414,1500:06:481176890Optimal

Person 2514,1500:06:481217492Optimal

Person 2614,1500:06:481308289Normal

Person 2714,1500:06:481216386Optimal

Person 2814,1500:06:481127390Optimal

Person 2914,1500:06:481238280Normal

Person 3014,1500:06:48956173Optimal

Person 3114,1500:06:481086572Optimal

Person 3214,1500:06:481107073Optimal

Person 3314,1500:06:481166771Normal

Person 3414,1500:06:481308680Normal

Person 3514,1500:06:481177380Optimal

Person 3614,1500:06:481175481Optimal

Person 3714,1500:06:481137474Optimal

Person 3814,1500:06:481328679Normal

Person 3914,1500:06:481288078Normal

Person 4014,1500:06:481318570Normal

Table 4 presents the average of the test of the MNN for each of the patients.


PersonTimeSystolicDiastolicPulse

Person 100:15:341157367

Person 200:15:341057277

Person 300:15:341147080

Person 400:15:341197172

Person 500:15:341458475

Person 600:15:341046190

Person 700:15:341258796

Person 800:15:341096473

Person 900:15:341297356

Person 1000:15:341227765

Person 1100:15:341366370

Person 1200:15:341368171

Person 1300:15:341207478

Person 1400:15:341106277

Person 1500:15:341196870

Person 1600:15:341318280

Person 1700:15:341257780

Person 1800:15:341066579

Person 1900:15:341106879

Person 2000:15:341237680

Person 2100:15:341157276

Person 2200:15:341127378

Person 2300:15:341227677

Person 2400:15:341176890

Person 2500:15:341217492

Person 2600:15:341308289

Person 2700:15:341216386

Person 2800:15:341127390

Person 2900:15:341238280

Person 3000:15:34956173

Person 3100:15:341086572

Person 3200:15:341107073

Person 3300:15:341166771

Person 3400:15:341308680

Person 3500:15:341177380

Person 3600:15:341175481

Person 3700:15:341137474

Person 3800:15:341328679

Person 3900:15:341288078

Person 4000:15:341318570

The results of the best MNN architecture are shown in Figure 4. For each of the modules of the MNN, the goal error was of 0.002 and 500 epochs were use;, the number of neurons used was 26 in the first layer and 29 in the second layer.

Table 5 presents the results of the MNN with the “trainscg” method for the classification of arterial hypertension.


No. PersonsNumber of NeuronsTimeSystolicDiastolicPulseClassification

Person126,2900:07:121167267Optimal

Person 226,2900:07:121067277Optimal

Person 326,2900:07:121147080Optimal

Person 426,2900:07:121197172Optimal

Person 526,2900:07:121458475Grade1 Hypertension

Person 626,2900:07:121076190Normal

Person 726,2900:07:121308797High Normal

Person 826,2900:07:121206497Optimal

Person 926,2900:07:121317473High Normal

Person 1026,2900:07:121227857Normal

Person 1126,2900:07:121366365High Normal

Person 1226,2900:07:121357570High Normal

Person 1326,2900:07:121207578Normal

Person1426,2900:07:121177277Optimal

Person 1526,2900:07:121216970Normal

Person 1626,2900:07:121329078High Normal

Person 1726,2900:07:121267680Normal

Person 1826,2900:07:121066582Optimal

Person 1926,2900:07:121106884Optimal

Person 2026,2900:07:121237690Normal

Person 2126,2900:07:121127178Optimal

Person 2226,2900:07:121117070Optimal

Person 2326,2900:07:121227371Normal

Person 2426,2900:07:121166780Optimal

Person 2526,2900:07:121207480Optimal

Person 2626,2900:07:121298079Normal

Person 2726,2900:07:121206174Optimal

Person 2826,2900:07:121127381Optimal

Person 2926,2900:07:121218267Normal

Person 3026,2900:07:12956177Optimal

Person 3126,2900:07:121066580Optimal

Person 3226,2900:07:121167572Optimal

Person 3326,2900:07:121167175Normal

Person 3426,2900:07:121308690Normal

Person 3526,2900:07:121177497Optimal

Person 3626,2900:07:121175888Optimal

Person 3726,2900:07:121137073Optimal

Person 3826,2900:07:121317180Normal

Person 3926,2900:07:121288182Normal

Person 4026,2900:07:121348581Normal

In Table 6 the average of the test of the MNN for each of the persons is presented.


PersonTimeSystolicDiastolicPulse

Person 100:17:191157367

Person 200:17:191057277

Person 300:17:191147080

Person 400:17:191197172

Person 500:17:191458475

Person 600:17:191046190

Person 700:17:191258796

Person 800:17:191096473

Person 900:17:191297356

Person 1000:17:191227765

Person 1100:17:191366370

Person 1200:17:191368171

Person 1300:17:191207478

Person 1400:17:191106277

Person 1500:17:191196870

Person 1600:17:191318280

Person 1700:17:191267680

Person 1800:17:191066582

Person 1900:17:191106884

Person 2000:17:191237690

Person 2100:17:191127178

Person 2200:17:191117070

Person 2300:17:191227371

Person 2400:17:191166780

Person 2500:17:191207480

Person 2600:17:191298079

Person 2700:17:191206174

Person 2800:17:191127381

Person 2900:17:191218267

Person 3000:17:19956177

Person 3100:17:191066580

Person 3200:17:191167572

Person 3300:17:191167175

Person 3400:17:191308690

Person 3500:17:191177497

Person 3600:17:191175888

Person 3700:17:191137073

Person 3800:17:191317180

Person 3900:17:191288182

Person 4000:17:191348581

Figure 5 presents the modeling data of diastolic pressure for the MNN; the pink line represents the real data and the green line represents data modeled with the MNN. The results of this model for the diastolic pressure that were obtained were good with respect to the records used to use the tests, since the trend according to the cardiologist was good.

Figure 6 presents the modeling data of systolic pressure with the MNN proposed; the pink line represents the real data and the green line represents data modeled with the MNN. The results of this model for the systolic pressure that were obtained were good with respect to the records used to use the tests, since the trend according to the cardiologist was good.

Figure 7 present the modeling data of pulse pressure with the MNN; the pink line represents the real data and the green line represents data modeled with the modular neural network. The obtained results of this model for the pulse pressure were good with respect to the records used in the tests, since the trend according to the cardiologist is good.

4. Statistical Comparative Study

In this section a hypothesis test is made based on the errors obtained with the architecture of the modular network using the Levenberg-Marquardt learning method (trainlm) to obtain the trend of the systolic pressure. In addition, the results are compared with linear regression models based on the obtained errors.

The model used to perform the statistical comparison was the well-known linear regression. This model describes the relationship between a dependent variable and (also known as the output or answer) as a function of one or more independent variables (called predictors). The general equation corresponding to a linear regression model is as follows:where represents a parameter that establishes the linear relationship between variables. represents the random error terms. represents the real data.y is the variable for classification.

The formulas to estimate the beta parameter values are given byIn this case the values for βo and β1 are the following (for each of the modules):

Module 1βo =0.101663348532528β1= 1.028385618868264

Module 2βo = 0.04938271β1= 1.024901397932274

Module 3βo =0.565544448530641β1= 1.104635529315485

A set of 30 experiments are carried out to compare the results; for this, we use the parametric Z test of two samples, which is used with the following formula: where is the observed difference. is the expected difference. is the standard error of the difference.

The null hypothesis establishes that the mean of the errors of the systolic neural network are greater than or equal to the average of the errors obtained by the regression, being the alternative hypothesis that the mean of the errors of the systolic neural network are lower than the average of the errors obtained by the regression; the parameters of the hypothesis test are shown in Table 7.


Parameters

Confidence Interval95%

Alfa0.05

Hoμ1μ2

Haμ1 < μ2

Critical ValueZ= -1.645

In Table 8 we show the descriptive statistics for this test.


VariableObservationsMeanStd. Derivation

MNN(sys)309.8201.997

Regression3016.8304.508

Table 9 shows the results obtained by applying formula (1) for Module 1.


Difference-7.010

z (Observed Value)-7.788

z (Critical Value)-1.645

p-value<3.33066907387547x10−15

Alfa0.05

Since the result of the p value is lower than the level of significance alpha = 0.05, we reject the null hypothesis and accept the alternative hypothesis, so we can conclude that there is sufficient evidence with a 5% level of significance to support the claim that the means of the errors of the modular neural network for the obtaining of the systolic pressure tendency are smaller than those obtained by the regression method.

In Table 10 the descriptive statistics for Module 2 (diastolic) test is shown.


VariableObservationsMeanStd. Derivation

MNN(sis)3023.1773.096

Regression3034.7785.438

Table 11 shows the results obtained by applying formula (1) for this test.


Difference-9.462

z (Observed Value)-8.2383

z (Critical Value)-1.645

p-value<1.11022302462516x10−16

Alfa0.05

Since the result of the p value is lower than the level of significance alpha = 0.05, we reject the null hypothesis and accept the alternative hypothesis, so we can conclude that there is sufficient evidence with a 5% level of significance to support the claim that the means of the errors of the modular neural network for the obtaining of the diastolic pressure tendency are smaller than those obtained by the regression method.

We show in Table 12 the descriptive statistics for Module 3 (pulse) test.


VariableObservationsMeanStd. Derivation

MNN(sis)3014.7741.821

Regression3028.3674.733

Table 13 shows the results obtained by applying formula (1) for the third module (pulse tendency).


Difference-13.593

z (Observed Value)-14.682

z (Critical Value)-1.645

p-value<2.59524148975464x10−17

Alfa0.05

Since the result of the p value is less than the level of significance alpha = 0.05, we reject the null hypothesis and accept the alternative hypothesis, so we can conclude that there is sufficient evidence with a 5% level of significance to support the claim that the means of the errors of the modular neural network for obtaining the pulse tendency are smaller than those obtained by the regression method.

When comparing the model of the modular neural network with the linear regression models by means of the z-statistic tests, we can realize that when using intelligent computing techniques, in this case the modular neural networks, we have a more efficient technique to classify the systolic and diastolic pressure and the pulse and this could help the cardiologist detect and prevent diseases in the blood.

5. Conclusion

In this paper we have obtained good results with the proposed model. The MNN classification model for arterial hypertension was implemented with two training methods for the modular neural network, namely, the Scale Conjugate Gradient Backpropagation (trainscg) and Levenberg-Marquardt (trainlm), and we achieved good results with the second method (trainlm). Good results were also obtained in the diastolic, systolic, and pulse models, since the trend was good with respect to the records used to perform the tests. In addition, we have made a comparison between the neural network model and the regression equations, showing that the MNN model statistically outperforms the regression model. In this paper we conclude that this classification method is effective and could help the cardiologist to detect and prevent a patient’s blood pressure.

Data Availability

The data that was used in this research to support the findings of this study are available from the corresponding author upon request by email pmelin@tectijuana.mx.

Conflicts of Interest

The authors declare that there are no conflicts of interest, financial or nonfinancial, with respect to this research study.

Authors’ Contributions

The three authors of the paper were responsible for (1) concept and design of the system, (2) acquisition of data, (3) analysis and interpretation of data, and (4) preparation of the manuscript. The authors have agreed to authorship the paper and the order of authorship for this manuscript of this research.

Acknowledgments

We would like to express our gratitude to the CONACYT for Research Grant no. 246774 and Tijuana Institute of Technology for the facilities and resources granted for the development of this research.

References

  1. A. A. Abdullah, Z. Zakaria, and N. F. Mohammad, “Design and development of fuzzy expert system for diagnosis of hypertension,” in Proceedings of the 2nd International Conference on Intelligent Systems, Modelling and Simulation (ISMS '11), vol. 56, pp. 113–117, IEEE, Kuala Lumpur, Malaysia, January 2011. View at: Publisher Site | Google Scholar
  2. A. A. Abdullah, Z. Zakaria, and N. F. Mohammad, “Design and development of fuzzy expert system for diagnosis of hypertension,” in Proceedings of the 2nd International Conference on Intelligent Systems, Modelling and Simulation (ISMS '11), vol. 10, pp. 131–141, IEEE, Kuala Lumpur, Malaysia, January 2011. View at: Publisher Site | Google Scholar
  3. R. Kaur and A. Kaur, “Hypertension Diagnosis Using Fuzzy Expert System,” International Journal of Engineering Research and Applications (IJERA), pp. 2248–9622, 2014. View at: Google Scholar
  4. R. Poli, S. Cagnoni, G. Coppini, and G. Valli, “A Neural Network Expert System for Diagnosing and Treating Hypertension,” The Computer Journal, vol. 24, no. 3, pp. 64–71, 1991. View at: Publisher Site | Google Scholar
  5. R. Fuller and S. Giove, “A Neuro-Fuzzy Approach to FMOLP Problems,” in Proceedings of CIFT94, pp. 97–101, Trento, Italy, 1994. View at: Google Scholar
  6. R. Nohria and P. S. Mann, “Diagnosis of Hypertension using Adaptive Neuro-Fuzzy Inference System,” IJCST, vol. 6, pp. 2229–4333, 2015. View at: Google Scholar
  7. A. Zeinab and T. Hamid, Design of a Fuzzy Expert System and A Multi-layer Neural Network System for Diagnosis of Hypertension, vol. 2, Department of Computer Engineering, Mashhad Branch, Islamic Azad University, Mashhad, Iran, 2015.
  8. R. Yafawi, M. E. Knauft, K. Stokem, J. M. Palminteri, and M. WirthJ, “Pulmonary arterial hypertension,” Encyclopedia of Cardiovascular Research and Medecine, pp. 181–194, 2018. View at: Google Scholar
  9. M. S. Kallistratos, L. E. Poulimenos, and A. J. Manolis, “Atrial fibrillation and arterial hypertension,” Pharmacological Research, vol. 128, pp. 322–326, 2018. View at: Publisher Site | Google Scholar
  10. C. Cuspidi, M. Tadic, G. Grassi, and G. Mancia, “Treatment of hypertension: The ESH/ESC guidelines recommendations,” Pharmacological Research, vol. 128, pp. 315–321, 2018. View at: Publisher Site | Google Scholar
  11. A. Corrado, M. Correale, N. Mansueto et al., “Nailfold capillaroscopic changes in patients with idiopathic pulmonary arterial hypertension and systemic sclerosis-related pulmonary arterial hypertension,” Microvascular Research, vol. 114, pp. 46–51, 2017. View at: Publisher Site | Google Scholar
  12. P. Srivastava, A. Srivastava, A. Burande, and A. Khandelwal, “A note on hypertension classification scheme and soft computing decision making system,” ISRN Biomathematics, vol. 11, pp. 13–22, 2013. View at: Google Scholar
  13. S. Huang, Y. Xu, L. Yue et al., “Evaluating the risk of hypertension using an artificial neural network method in rural residents over the age of 35 years in a Chinese area,” Hypertension Research, vol. 33, no. 7, pp. 722–726, 2010. View at: Publisher Site | Google Scholar
  14. P. Srivastava, A. Srivastava, A. Burande, and A. Khandelwal, “A note on hypertension classification scheme and soft computing decision making system,” ISRN Biomathematics, vol. 12, Article ID 342970, pp. 20–23, 2013. View at: Publisher Site | Google Scholar
  15. M. Ture, I. Kurt, A. Turhan Kurum, and K. Ozdamar, “Comparing classification techniques for predicting essential hypertension,” Expert Systems with Applications, vol. 29, no. 3, pp. 583–588, 2005. View at: Publisher Site | Google Scholar
  16. P. Melin, I. Miramontes, and G. Prado-Arechiga, “A hybrid model based on modular neural networks and fuzzy systems for classification of blood pressure and hypertension risk diagnosis,” Expert Systems with Applications, vol. 107, pp. 146–164, 2018. View at: Publisher Site | Google Scholar
  17. P. Melin and G. Prado-Arechiga, New Hybrid Intelligent Systems for Diagnosis and Risk Evaluation of Arterial Hypertension, Springer, 2018.
  18. J. C. Guzman, P. Melin, and G. Prado-Arechiga, “Design of an optimized fuzzy classifier for the diagnosis of blood pressure with a new computational method for expert rule optimization,” Algorithms, vol. 10, no. 3, p. 79, 2017. View at: Google Scholar
  19. S. Das, P. K. Ghosh, and S. Kar, “Hypertension diagnosis: A comparative study using fuzzy expert system and neuro fuzzy system,” in Proceedings of the 2013 IEEE International Conference on Fuzzy Systems, FUZZ-IEEE 2013, vol. 148, pp. 113–121, Durgapur, India, July 2013. View at: Google Scholar
  20. X. Y. Djam and Y. H. Kimbi, “Fuzzy expert system for the management of hypertension,” Pacific Journal of Science and Technology, vol. 11, pp. 390–402, 2011. View at: Google Scholar
  21. R. E. Klabunde, Cardiovascular Physiologic Concepts, Wolters Kluwer Health, 2nd, 2011.
  22. I. K. Ludmila and F. Steimann, Fuzzy Medical Diagnosis, School of Mathematics, University of Wales, Banggor, UK, 2008.
  23. H. Zhang and F. C. Lin, “Medical Diagnosis by the Virtual Physician,” IEEE Xplore Computer Based Medical System, pp. 296–302, 1999. View at: Google Scholar
  24. Y. Clec'h, C. Vicaut, E. Marbeuf-G et al., “Can fuzzy logic make things more clear?” Critical Care, vol. 3, no. 1, pp. 116–120, 2009. View at: Google Scholar
  25. M. R. Grübler, M. Gaksch, K. Kienreich et al., “Effects of Vitamin D3 on asymmetric- and symmetric dimethylarginine in arterial hypertension,” The Journal of Steroid Biochemistry and Molecular Biology, vol. 175, pp. 157–163, 2018. View at: Publisher Site | Google Scholar
  26. B. Sumathi and A. Santhakumaran, “Pre-diagnosis of hypertension using artificial neural network,” Global Journal of Computer Science and Technology, Coimbator, Tamil Nadu, vol. 3, no. 6, pp. 21–32, 2011. View at: Google Scholar
  27. V. G. Vilkov, R. G. Oganov, and S. A. Shal'nova, “Comparative accuracy of neural network models for diagnosing latent arterial hypertension on the basis of data on daily blood pressure monitoring,” Human Physiology, vol. 32, no. 6, pp. 657–661, 2006. View at: Publisher Site | Google Scholar
  28. A. Kaur and A. Bhardwaj, “Genetic neuro fuzzy system for hypertension diagnosis,” International Journal of Computer Science and Information Technologies, vol. 5, no. 4, pp. 4986–4989, 2014. View at: Google Scholar
  29. M. Barman and J. Choudhury, “A fuzzy rule base system for the diagnosis of heart disease,” International Journal of Computer Applications, pp. 46–53, 2012. View at: Google Scholar
  30. P. Patil and S. Mohsin, “Fuzzy Logic based Health Care System using Wireless Body Area Network,” International Journal of Computer Applications, vol. 80, no. 12, pp. 46–51, 2013. View at: Publisher Site | Google Scholar
  31. I. Morsi and Y. Z. Abd El Gawad, “Fuzzy logic in heart rate and blood pressure measuring system,” in Proceedings of the 8th IEEE Sensors Applications Symposium, SAS 2013, pp. 113–117, IEEE, Galveston, Tex, USA, February 2013. View at: Google Scholar
  32. S. Hosseini, C. Jutten, and S. Charbonnier, “Neural network modeling of ambulatory systolic blood pressure for hypertension diagnosis,” in Artificial Neural Nets Problem Solving Methods, vol. 2687 of Lecture Notes in Computer Science, pp. 599–606, Springer, Heidelberg, Germany, 2003. View at: Publisher Site | Google Scholar
  33. R. M. Touyz, N. N. Lang, J. Herrmann, A. H. Van Den Meiracker, and A. H. J. Danser, “Recent Advances in Hypertension and Cardiovascular Toxicities With Vascular Endothel ial Growth Factor Inhibition,” Recent Advances in Hypertension, pp. 1–3, 2017. View at: Google Scholar
  34. L. Feng, A. H. Khan, I. Jehan, J. Allen, and T. H. Jafar, “Albuminuria and kidney function as prognostic marker of left ventricular mass among South Asians with hypertension,” Journal of the American Society of Hypertension, vol. 11, no. 12, pp. 811–822.e2, 2017. View at: Publisher Site | Google Scholar
  35. J. L. Nierenberg, C. Li, J. He et al., “Blood Pressure Genetic Risk Score Predicts Blood Pressure Responses to Dietary Sodium and Potassium: The GenSalt Study (Genetic Epidemiology Network of Salt Sensitivity),” Hypertension (Dallas, Tex. : 1979), vol. 70, no. 6, pp. 1106–1112, 2017. View at: Publisher Site | Google Scholar
  36. J. S. R. Jang, C. T. Sun, and E. Mizutani, Neuro-Fuzzy and Soft Computing, Prentice Hall, 1996.
  37. C. Torlasco, A. Faini, E. Makil et al., “Cardiovascular risk and hypertension control in Italy. Data from the 2015 World Hypertension Day,” International Journal of Cardiology, vol. 243, pp. 529–532, 2017. View at: Publisher Site | Google Scholar
  38. G. Beevers, G. Y. H. Lip, and E. O'brien, “Blood pressure measurement: Part I—Sphygmomanometry: Factors common to all techniques,” BMJ, vol. 322, no. 7292, p. 981, 2001. View at: Publisher Site | Google Scholar
  39. D. Bernstein, “Evaluation of the cardiovascular system: history and physical,” in Evaluation, R. M. Kliegman, B. F. Stanton, and J. W. St. Geme, Eds., Nelson Textbook of Pediatrics, pp. 385–450, 2015. View at: Google Scholar
  40. Harrison, “Tachyarrhythmias,” in Principles of Internal Medicine, vol. 2, pp. 2226–2232, McGraw-Hill, 19a edition, 2006, Section 14. View at: Google Scholar
  41. “American Heart Association,” 2016, https://www.heart.org/. View at: Google Scholar
  42. D. L. Simel, L. Goldman, and A. I. Schafer, “Approach to the Patient: History and Physical Examination,” in Goldman's Cecil Medicine, 2012. View at: Google Scholar
  43. “World Health Organization,” International Society of Hypertension Group, 2003. View at: Google Scholar
  44. J. A. Whitworth, “World Health Organization (WHO)/International Society of Hypertension (ISH) statement on management of hypertension,” Journal of Hypertension, vol. 21, no. 11, pp. 1983–1992, 2003. View at: Publisher Site | Google Scholar
  45. G. Parati and G. Mancia, “Ambulatory Blood Pressure Monitoring,” in Manual of Hypertension, G. Mancia, J. Chalmers, S. Julius et al., Eds., pp. 153–171, Churchill-Livingstone, London, UK, 2002. View at: Google Scholar
  46. M. Sokolow, “Ambulatory blood pressure a personal historical account,” American Journal of Hypertension, vol. 6, no. 6, pp. 1615–1655, 1993, J Hypertension; 20: 1917-1923. View at: Publisher Site | Google Scholar
  47. G. Parati and G. Mancia, “Ambulatory blood pressure monitoring in clinical practice,” Journal of Hypertension, vol. 20, no. 10, pp. 1925–1927, 2002. View at: Publisher Site | Google Scholar

Copyright © 2019 Martha Pulido et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.


More related articles

1488 Views | 368 Downloads | 3 Citations
 PDF  Download Citation  Citation
 Download other formatsMore
 Order printed copiesOrder

Related articles

We are committed to sharing findings related to COVID-19 as quickly and safely as possible. Any author submitting a COVID-19 paper should notify us at help@hindawi.com to ensure their research is fast-tracked and made available on a preprint server as soon as possible. We will be providing unlimited waivers of publication charges for accepted articles related to COVID-19. Sign up here as a reviewer to help fast-track new submissions.