Personal and Ubiquitous Computing

, Volume 17, Issue 6, pp 1147–1157

Ubiquitous monitoring and assessment of childhood obesity


    • Universidad Politécnica de Valencia, I3BH/LabHuman
    • Ciber, Fisiopatología Obesidad y Nutrición, CB06/03, Instituto de Salud Carlos III
  • Jaime Guixeres
    • Universidad Politécnica de Valencia, I3BH/GBIO
  • Mariano Alcañiz
    • Universidad Politécnica de Valencia, I3BH/LabHuman
    • Ciber, Fisiopatología Obesidad y Nutrición, CB06/03, Instituto de Salud Carlos III
  • Ausiás Cebolla
    • Labpsitec, Universitat Jaume I
  • Javier Saiz
    • Universidad Politécnica de Valencia, I3BH/GBIO
  • Julio Álvarez
    • Cardiovascular Risk Unit. Pediatrics ServiceGeneral University Hospital of Valencia
Original Article

DOI: 10.1007/s00779-012-0562-x

Cite this article as:
Zaragozá, I., Guixeres, J., Alcañiz, M. et al. Pers Ubiquit Comput (2013) 17: 1147. doi:10.1007/s00779-012-0562-x


Childhood obesity is a significant health problem in current societies that is increasing at an alarming way among population of all ages. To date, studies on the effectiveness of treatments for childhood obesity in the medium and long term suggest a moderate effect on weight loss and maintenance, which has led to suggestions that early interventions have a preventive nature on adult obesity. The long-term recovery of the weight lost is often associated with a lack of adherence to recommendations for changing life habits. Then, obesity becomes a chronic problem, difficult to approach, and the main difficulty lies in promoting and ensuring adherence to a change in lifestyle. A system known as ETIOBE has been developed to improve the treatment adherence, to promote the mechanisms of self-control in patients and to prevent relapses. An important part of the ETIOBE system is the ubiquitous monitoring platform since it enables the clinician to obtain relevant information from patients (contextual, physiological and psychological), which enables treatment customization and adaptation, depending on the patient’s evolution. The aim of this paper is to describe the monitoring platform which is intended to establish a sensor network whose focus is the obese children under clinical treatment, and the various elements that compose it: electronic PDA records to establish diet habits, HAS: home ambulatory system (data integration of biomedical devices; blood pressure to study hypertension; pulse oximeter to detect Sleep Disorders; and electronic t-shirt to detect physical activity). This paper presents the first validations of the electronic PDA records and the electronic t-shirt. These validations suggest that the monitoring platform can help to achieve the goals previously mentioned, by offering constant support and increasing motivation to change.


Children obesityE-therapyPhysical activity detectionWireless monitoring

1 Introduction

Obesity is a problem that is spreading at an alarming rate among populations of all ages, both in childhood [1] and in the adult age [2], to the extent of being regarded as a new epidemic. In fact, obesity is the most frequent nutritional disorder in developed countries, for which reason WHO (the World Health Organization) declared it a global epidemic in 1998 [3], since it affects at least 300 million people who can be diagnosed as obese—a figure which rises to one billion if overweight is included [4]. The characteristics of a person suffering from obesity include a number of significant consequences, both physically and psychologically. Firstly, and from a physical point of view, obese people have an increased risk for various types of diseases, especially of the cardiovascular type [5]. Most obese people suffer from hypertension, with all the problems this entails in terms of quality of life. In this respect, it should be noted that in addition to being a cardiovascular risk itself, obesity facilitates the development of comorbidities such as diabetes, hypertension, high cholesterol, arthritis, and breast or colon cancer [6]. The therapeutic goals of weight loss programs try to improve or eliminate the comorbidities associated with obesity and reduce the impact of future medical complications related to excess weight. Therefore, weight loss goals should not focus on achieving the ideal weight, but to produce a small weight loss (5–10 % of initial weight) which is maintained in the long term [7]. Most of the studies on the efficacy of weight loss treatments show good short-term results, such as [8, 9] in which subjects lost more than 10 % of weight during the treatment. However, weight regain is common after the end of treatment: 30–35 % of the weight lost is recovered in the year after treatment, although most patients maintained a medically significant weight loss of more than 5 % of their initial weight after 1 year [10]. The success of the treatment depends on the degree to which patients can make changes to lifestyle habits and continue to follow therapeutic advice. If patients do not receive continuous support, they are faced with numerous obstacles to recovery and with an increased risk of treatment failure. Given the multifactor nature of obesity, treatment includes, among other things, improving eating habits, increasing physical activity, psychological support and, in some cases, drug administration, in addition to the consideration of environmental and cultural factors. It is therefore essential to consider the use of psychological treatments [11]. The ultimate goal of these treatments is to increase motivation and adherence, and to produce lifestyle changes by establishing healthy habits to prevent weight gain. To this end, it is necessary to have tools that make it possible to constantly monitor patients, allowing us to know in detail patient patterns at any time (physical activity, food intake, etc. …). Adequate participation in physical activity during childhood and adolescence is considered essential for good health and normal growth and development [12]. Moreover, as several of the health outcomes related to physical activity tend to continue from childhood into adulthood [13], adequate participation in physical activity during childhood and adolescence may be of critical importance in primary prevention of chronic disease. It is crucial that researchers and practitioners have access to accurate yet user-friendly clinical instruments to assess physical activity behaviour in obese children under treatment. These professionals require non-intrusive, valid and precise methods to understand how intensity, frequency and duration of physical activity influence the health of children [14]. The concept of ambient intelligence (AmI) has arisen and was described by the ISTAG (Information Society Technologies Advisory Group). In recent years, several groups have begun working in this area [1517]. One of the main goals in ambient intelligence scenarios is obtaining as much information as possible about each element present in the scenario, with the aim of providing users with the best possible solution at any given time. Other important point is that the user feels that the system takes his state and contextual situation into account in a personalized way. The user must feel that his surroundings are redesigned in an intelligent way to adapt to his needs. Going back to therapy and clinical treatment, a structure is required to provide specialists with information to evaluate patient habits and design, with the greatest possible effectiveness, the best treatment for him. To do this, having various types of user information in real time is crucial.

1.1 Related work

Wearable systems for health monitoring have drawn a lot of attention from the research community and the industry in the last decade, as reflected in the increasing number of works in this area [18, 19]. Moreover, in the last few years several works for the promotion of physical activity and the fight against obesity have appeared. For example, a mobile phone application designed to create a support group for physical activity within an existing social network of adolescent girls [20] or NEAT-o-Games [21] where data from a wearable accelerometer are sent to a cell phone to control the animation of an avatar that represents the player in a virtual race game with other players over the cellular network. Ubifit [22] is a mobile, persuasive technology to encourage individuals to self-monitor their physical activity. Other hardware-based devices such as nike+ [23] or more recently fitbit [24] and jawbone [25] are being commercialized, very successfully in many cases. This work introduces a new ubiquitous monitoring platform for detecting cardiovascular, metabolic and physical activity patterns in children with obesity, the ETIOBE platform. This platform, apart from collecting and monitoring information from the children, allows the customization and adaptation of the treatment, depending on patient evolution. In Sect. 2.1, we describe the general structure of the ETIOBE project platform and its various components. A more detailed description of the sensor layer is described in Sect. 2.2. Finally, Sect. 3 provides user validation of the components of this layer.

2 Methods

2.1 ETIOBE project

The ETIOBE platform is an intelligent e-therapy system (e-it) for the treatment of obesity. Intelligent e-therapy systems aim to cover the need for continuous customization of patient treatment. An e-it system adapts to the patients lifestyle, offering 24/7 monitoring of the patient and represented an evolution of the cybertherapy and telepsychology tools currently in use [26]. The main objective of ETIOBE is to improve treatment adherence and promote self-control mechanisms in patients, to achieve the maintenance of goals (less body weight) and to prevent relapses by establishing healthy habits. ETIOBE is supported by different ICTs able to transfer, manage, store, interpret and react to information, as well as to customise and adapt treatment strategies according to individual patient characteristics and responses. The intelligent aspect of ETIOBE is based on the use of a sensorization module that makes it possible to obtain all the relevant information pertaining to patients (contextual, physiological and psychological), as well as on the existence of communication and computer applications capable of transferring this information, storing it, managing it, and properly interpreting it. The system reacts to this information by offering a set of personalized contents, depending on the patient’s characteristics, information and answers. These contents are orientated towards change and/or reinforcement of patient behaviour. Users can access the system independently of the platform used and location. This constitutes the ubiquitous aspect of the system. ETIOBE uses communication and computer applications capable of adapting their features to the nature of the device and the communication network that is in use. Technically, ETIOBE comprises three different applications for the users (clinical experts and patients).
  • The clinical support system (CSS) is an Internet tool allowing clinicians to design an adapted guidelines intervention, with the possibility of adapting it according to patient progress. The main aim of this application is to provide therapists with a simple, agile and functional tool for the treatment of obesity in children population, offering a joint and easy monitoring of the patient to all the actors involved in the treatment. During treatment each user has an established role (doctor, therapist, etc.) that allows him to access the various system functionalities. This platform is connected to the other platforms, which make it possible for clinicians to monitor patient progress in real time.

  • The home support system (HSS) is a website where children find the tasks selected by the clinicians (including self-reports, dietary records and physical activity records). The website also features a nutritional and healthy lifestyle knowledge component, including several “serious games” which help children absorb the information. One of the main aims of the project is to establish an attractive connection between the treatment, nutritional information and the patient. This application tries to create a social network for the children following the treatment: for this reason, children earn points for each activity performed and there is a global ranking.

  • The mobile support system (MSS) is comprised of a dietary and physical activity self-record based on PDA technology, sensorization and physiological information that is connected in real time to the CSS.

These applications have been developed using .NET platform. As previously mentioned, the sensorization module plays an important role in ETIOBE since it makes it possible for clinicians to obtain relevant patient information (contextual, physiological and psychological), enabling treatment customization and adaptation, according to patient evolution.

2.2 Sensorization module/ubiquitous monitoring network

As part of the ETIOBE project, we have developed a ubiquitous monitoring platform that establishes a sensor network where the focus is the patient. This network captures information in real time. Within this network, it is possible to distinguish 2 different types of information. On the one hand, the information that the user enters personally into the system through electronic records running on a PDA, and on the other hand, the information collected automatically by biomedical devices worn by the user. Both types of information are equally important to know the user’s condition (physical, psychological and contextual) at any time. Electronic PDA records allow patients to enter information on the amount and type of food eaten as well as on the physical exercise performed. This information is sent directly from the PDA where the self-record is made to the central server for storage and analysis. Users have three different measuring devices: a blood pressure cuff, a pulse oximeter and the most important element, a special electronic t-shirt to obtain physiological and contextual information about the free-living conditions of obese children. The two former devices are commercial and the latter has been developed specifically for this project. This smart t-shirt is placed on the patients only during specific weeks designated by the physician completing the data on physical activity habits collected from electronic records. To download and integrate the information from such devices, we have developed a simple and intuitive application (HAS: home ambulatory system). This application also provides users with simple guides on how to use these devices (connecting, charging battery, etc). Figure 1 shows the intuitive appearance of the application.
Fig. 1

Home ambulatory system

2.2.1 Electronic PDA records

The aim of self-records is to evaluate behaviours in a natural setting (e.g. home or school). The information obtained allows the clinic to identify behaviour cues, as well as the thoughts and emotions associated with the behaviour. A more accurate evaluation can therefore be made, and treatment effects and patient evolution can be assessed. As regards obesity, the most important targets to be self-monitored are: information about diet and physical activity. As an assessment tool, a fundamental benefit of diary methods is that they allow the examination of events and experiences in their natural, spontaneous contexts [27]; furthermore, they allow patients to see the positive changes they are making and to track their progress. Because the self-record application runs on PDAs using the Windows Mobile operating system, it was developed using .net technology based on Compact Framework 3.5 for mobile devices [28]. This application allows patients to enter personal information into the system. In the first version, patients can fill in two types of records, for diet and activity. The application starts automatically when the PDA is turned on; this makes it easier for children to use. First, the main screen appears showing the logo of the ETIOBE system and a button with the tag “self-record” The application always remains open; therefore, if the PDA is set to “standby/energy saver” mode, the main screen with the ETIOBE logo automatically appears when the active mode is resumed. When users want to fill in a record, they must press the “self-record” button. The login screen then appears, and after completing the self-record, the user presses the exit button and returns to the main screen of the application.

The dietary record (see Fig. 2) allows users to specify the type of food eaten, the amount (number of pieces, the portion size, etc.) and the type and amount of drink (1, 2 or 3 glasses). Users must also indicate the social situation at the time of the ingestion (alone, with friends or with parents) the place where the meal was eaten (home, school, bar, bakery or kiosk) and the emotion experienced prior to eating.
Fig. 2

Screenshots of dietary electronic record

The physical activity record (see Fig. 3) allows users to select the kind of activity performed (football, gymnastics, running, riding, dancing or others), how long they have been performing the activity (in minutes) and the intensity of the exercise (light, moderate, strong, very strong). At any moment, users can press the “save self-records” button on the main screen and all self-records introduced are sent to the server for storage. A more detailed description of these electronic PDA records is provided in [29].
Fig. 3

Screenshots of activity electronic record

2.2.2 Blood pressure monitors: study of hypertension

Blood pressure is measured using a clinically validated digital blood pressure monitor (Omron HEM-705IT) [30]. Measurements are taken during the day (normally 4 times) according to the planning designed by physicians in the CSS. The device internally stores the measurements and is synchronized with HAS to download all the information to the central server in the hospital.

2.2.3 Pulse oximetry: study of sleep disorders

A portable pulse oximeter monitor (Nonin Medical’s WristOx 3100) [31] is placed at night on the child’s wrist and a sensor is placed on his finger with a flex sensor so as not to disturb him. The pulse oximeter measures oxygen saturation (±0.01 %) and pulse rate (±3 %) during all the night with a sample rate of 1 Hz. In the morning, the device is connected to the home computer and the HAS platform download all the information from the device. At the CSS, an algorithm is in charge of analyse the data and detect number of instances of apnoea, number of instances of hypopnoea, apnoea index, hypopnoea index, apnoea–hypopnea index (AHI) and oxygen saturation (mean and minimum) measured by the pulse oximeter. With this information, ETIOBE may help to study the relation between OSAHS and Obesity. This platform is under revision by means of a validation study for testing usability and feasibility of platform.

2.2.4 Physical activity detection with smart fabrics (t-shirt)

Smart fabrics enable the monitoring of patients over extended periods and in a natural context, in biomedicine as well as in several health-focused disciplines, such as bio-monitoring, rehabilitation, telemedicine, teleassistance, ergonomics and sport medicine [32]. The innovation in this field is originated by the development of a new generation of textile sensors, combining electronics and informatics novelties, leading to the integration of multiple, smart functions into textiles-based sensing interfaces, aiming to the reduction of any impediment [33]. Due to the fact that e-textiles do not have wiring harnesses between discrete components, they have a distinctive advantage over conventional electronics for home health care. Both the wires and the components are part of the fabric and thus much less visible, and more importantly, they are cannot become entangled snagged by the surroundings. Consequently, e-textiles can be worn in everyday situations where currently available electronic devices would hinder or perhaps embarrass the user. Within ETIOBE, we have developed a smart t-shirt that uses smart fabrics from the medical company NUUBO [34]. The t-shirt can record vital constants and physical activity patterns for E-therapy treatments. Figure 4 shows the shirt developed. The main properties for the t-shirt are:
  • Conductive electrodes integrated inside the fabric using metal yarns.

  • ECG signal acquisition with medical quality (sample rate: 250 Hz, complete morphology of the QRS complex and P wave, RR interval measuring on real time, new filters for Baseline and Motion and Muscle Noise reduction).

  • Physical activity and metabolic real time algorithm integrated.

  • High memory capacity.

  • Low power consumption (in Holter Mode it can be on up to 30 h).

  • Reduced Size (module size 4 by 1 cm and weight only 50 g).

  • Data protocol communication (different modes (continuous mode, intelligent mode (with alarms), programmed mode (summarize data) and temporised mode).

  • Wireless transmission (Bluetooth or Zigbee).

  • T-Shirts are made for different genders and sizes to offer the best comfort to the patient. Patient can wash the t-shirt normally without special considerations.
Fig. 4

T-shirt developed between I3BH and NUUBO medical. On the left, t-shirt worn during the laboratory experiment. On the right, the electronic module (white device in centre) connected to fabric electrodes (grey zone on fabric)

We also have developed an algorithm to detect patient activity states and posture. In order to develop this module it was made an exhaustive study of the different options that might make it possible to determine body position, physical activity and energy consumption, taking into account that patient’s daily activity should not be interrupted [35]. By blocking the sensor position on the patient body (first on the hip, then on the chest), our system distinguishes among periods of activity (light, moderate, vigorous and rest), recognizes the postural orientation of the wearer, and detects events such as walking and falls with a reasonable degree of accuracy, as well as provides an estimation of metabolic energy expenditure. The algorithm developed is based on the classification framework presented by Mathie et al. [36], which involves a hierarchical binary structure where broad classifications are made in the top levels of the decision tree, and more detailed sub-classifications are made at lower levels. This structure makes it possible to modularize the required classification decisions and to test the associated algorithms independently. The algorithm also uses the heart beat for the classification of states. After analysing certain studies and articles applied to physical activity estimation [37], we made a frequency analysis of acceleration due to movement to determine the step frequency of the new states (walk, race) [38]. Depending on the disorder to be treated, the metabolic equations will be different. As will be shown in the results section, a new estimation model for obese children, combining biomechanics and physiological parameters, has been established. We have also developed a desktop application that records the values obtained from the t-shirt and allows making notes data files to record changes in activity or position during the validation studies.

3 Results

3.1 Validation of electronic PDA records

At this time, a validation of the full ETIOBE system (Clinical Support System, Home Support System and Mobile Support System) is being performed. The objective of this validation is to compare the effectiveness of the system compared to a traditional system. This validation began in late 2011 and will continue throughout 2012. As part of the validation protocol, children receive a PDA to record what they eat and their physical activity performed during treatment. To test the correct operation of all system before starting the validation, we performed a small pilot study with 5 children. Each child was given a PDA and asked to record his intake and physical activity during a week. During the week of the pilot study, children had no serious problems recording information and the server received 109 dietary records and 30 physical activity records. As expected, given the structure of the electronic records (a step-by-step wizard), all records received were complete, which means data were included in all fields. Thus, it is shown that one of the main strengths of the PDA records is that they prevent incomplete records, always transmitting the full information to the server.

Furthermore, the information is stored in real time, so the therapist can see it and analyse it at any time in the clinical system.

Table 1 shows the distribution of the records received organized by day and type (dietary and physical activity).
Table 1

Distribution of dietary and activity records


Day 1

Day 2

Day 3

Day 4

Day 5

Day 6

Day 7

No. dietary regs








No. activity regs








On average, each child filled in 4.29 dietary records and 0.85 activity records per day. As can be seen, the number of dietary records is higher than the number of physical activity records, since children were asked to record separately each meal they took during the day. While analysing the records received, we realized that when a child forgot recording a meal, he would include the information for both meals (the current one and the one previously forgotten) in his following record, although the system would count it only as one record. On the other hand, the results show that the children record a higher number of physical activities on the first few days; this may be due to the newness of the registration system. Finally, it is worth highlighting that some children said that sometimes, when trying to record, they had problems with the network and had to repeat the process. This is because the system must have a data connection (internet access) to send the information to the server, and if there are coverage problems and there is no internet access the system cannot continue and the recording process should be repeated. We hope that, given the advances in mobile technology, there will be less coverage problems and these cases will become fewer and eventually disappear

3.2 Laboratory validation of the t-shirt

In order to validate algorithms, filters implemented on electronic module of the t-shirt and feasibility of smart fabrics electrodes, a laboratory protocol was implemented. The phases of the protocol were:
  • 2 min on sedentary phase (stand up)

  • 2 min walking on a treadmill (4 km/h)

  • 2 min jogging on a treadmill (6 km/h)

  • 2 min walking on a treadmill (4 km/h)

  • 2 min on sedentary phase (stand up)

An adult (age = 35, weight = 80 kg) completed, during five consecutive days at different times of the day, the explained protocol. Signals were acquired by desktop application and exported to Matlab for analysis. ECG signals were separated in phases (sedentary, walking and jogging), and fiducial points detected by RR algorithm from the t-shirt were represented joined to ECG signals (a sample of these signals is shown in Fig. 5). Fiducial points were revised manually by an expert physician and results were compared with automatic detection from t-shirt. The sensitivity or true positive rate (TPR) (Eq. 1) was calculated as:
$$ \text{TPR}=\text{TP}/\text{P}=\text{TP}/(\text{TP}+\text{FN}) $$
where TP, true positive; FN, false negative.
Fig. 5

Sample of ECG and QRS detection for interval sample in different experiment phases. Sensitivity calculated by a manual revision of the entire signal on each phase. Upper graph (sedentary phase), in the middle (walking phase) and lower graph (running phase)

Figure 5 presents real results of the sensitivity calculated for each phase using the entire signals during 5 days.

Regarding the accelerometer signal, it was obtained, joined to the t-shirt, using a commercial and clinical validated accelerometer, (Actigraph GT1X) [3941]. The correlation factor during all the phases was 0.902 (p < 0.05). The response from the accelerometer signal has the correct intensity for the different phases accomplished on the experiment (sedentary, light and moderate physical activity). Red line marks the threshold defined by Mathie et al. [36] for the differentiation between active and sedentary level.

3.3 Clinical study of models for estimating physical activity in obese children

A clinical protocol to test the best model for clinical estimation of metabolic activity in obese children has been developed. The protocol is presented in Table 2.
  • Two groups (control and obese group) of 20 children (male/female subjects 10 ± 4 years).

  • Nurse measures body composition, anthropometrics and initial heart rate and blood pressure.

  • In the first phase, the patient is measured with an indirect calorimeter while lying on a bed for 20 min to obtain the resting metabolic rate RMR.

  • Afterwards, patient must complete various activity stages for 45 min, from a sedentary level to a vigorous level (Table 2).

  • During the protocol Respiration Rate, RR interval and Accelerometer against O2 consumption are measured [4244] with t-shirt and Indirect calorimeter (COSMED, Fitmate PRO). Indirect Calorimeter is a gold-standard measuring of Energy Expenditure. All signals are synchronized for the processing stage.

  • Between stages, blood pressure is obtained (three measurements each time).

  • Questionnaires for Perceived Exertion (Borg scale), physical activity habits (3-Day Physical Activity Recall; 3DPAR), diet habits (a food frequency questionnaire) and ergonomics are administered at the end by the psychologist.

Table 2

Protocol designed for studying ambulatory activity detection in obese children




Duration (seg)

3 Blood pressure measurements

 Sedentary phase (20 min)

Sitting watching TV/sitting playing videogame

Patient is watching a film sitting in a chair/patient is playingbrain training games sitting in a chair


 Effort test (until exhaust)

Balke Modified Protocol/Constant speed at 5.3 km/h while the inclination is from 2 to 15 % each minute. Since that point the speed is increased 03 km/h each minute until 6.2 km/h limit. The test is finished when the patients arrives to his 85 % HRmax or is on 75 % Hrmax during 3 min

Until exhaust

 Recovery phase (15 min)

Sitting answering questionnaires

Patient remains sitting while answering questionnaires with the help of the psychologist. The measures are stopped when the patient arrivesto 50 % HRmax

Until 50 % HRmax

First, test data were downloaded from the pulse oximeter and metabolic chart to the control computer database. Analytics, demographic and anthropometrics data were recorded in the database before the test; blood pressure and the t-shirt measurements were stored in real time in the database during the experiment. Data were processed with Matlab routines to filter the signals, eliminate outliers and check the validity of the signals for each patient. Data were summarized over a 10-s epoch and 1 min epoch, respectively. Data were analysed for the entire protocol and divided into resting phase, sedentary phase and effort test phase. The acceleration signals from the three axes were processed to obtain the Signal Magnitude Area (SMA) using Eq. 2, a validated indicator reference for physical activity intensity. Measure units for SMA are “Gcounts. This standard unit gives the “counts” of the accelerometer in Gs (9.8 m/s2) integrated at the selected epoch time.
$$ \text{SMA}_n = 1/N\int(|\text{AccXbm}|+|\text{AccYbm}|+|\text{AccZbm}|) $$
where N, number of samples for the chosen epoch; AccXbm, AccYbm and AccZbm; acceleration due to movement on the X, Y and Z axis (lowpass filter implemented [fc = 0.5 HZ] to eliminate acceleration due to gravity).
HR values obtained from the t-shirt were correlated with HR values collected from indirect calorimeter (Fig. 6). Figure 7 shows the correct correlation of signals for the entire protocol in one patient, especially on the signals summarized by minutes. The correlation factor among all the HR signals summarized in minutes during experiment in all patients was 0.933, an index that ensures the validity of the t-shirt for measuring HR during the experiment.
Fig. 6

Physical activity level obtained from accelerometer signal during the experiment
Fig. 7

Sample of heart rate of a participant during the entire protocol. The green text shows the sedentary phases, the red text the beginning of the effort phase, and the blue text the beginning of recovery phase. The blue signal is the continuous HR measured by the t-shirt and the red signal is the HR measured by the Indirect Calorimeter

Stepwise linear regression, curve fit and nonlinear regression were used to predict VO2/kg/min, a gold-standard measure of the energy expenditure of the human body. During the result analysis, we obtained several models and prediction equations from the obese group. To determine the validity of these prediction equations extracted from the obese group, a cross validation technique was used. A clinically obese subsample (n = 9) was randomly selected, and prediction equations derived from the obese group were used to predict metabolic variables in the obese subsample. Table 3 and Fig. 8 display the results of the new model implemented in the t-shirt (Eq. 3). Results show a significantly low error in both phases. This suggests that prediction models combining HR and accelerometer as t-shirt platform data provide the most accurate estimate of metabolic cost as compared to accelerometer or HR single-measure models. Puyau et al. [45] validated two accelerometer single models in children with 76–79 % and 81 % of variability in Energy Expenditure due to Activity and concluded that single accelerometer models can be used to discriminate between levels of intensity. Reilly et al. [46] assessed the validity of two equations based on the Actigraph with a mean error of +0.3 MJ/d and concluded that they appear to be inadequate for the estimation of free-living TEE in young children. Norman et al. [47] compared accelerometer models to detect physical activity in healthy children and children with cerebral palsy and they found no significant differences between the mean caloric estimated by the accelerometer model and the indirect calorimeter. These data justify the use of a model combining heart rate and Accelerometers used with our t-shirt.
$$ \begin{aligned} \text{VO}2(\text{ml}/\text{min}*\text{kg})&= \text{ACC}(\text{gcounts}) + \text{HR}(\text{bmp})\\ &\quad +\text{restingHR}(\text{bmp})+\text{waist}(\text{cm}) \end{aligned} $$
Table 3

Results of the validation for the oxygen consumption model



Mean error

Error (%)

SD error











The first line corresponds to the resting phase and the second line to the effort phase
Fig. 8

Validation response for the proposed model for oxygen consumption. The red line is the real VO2, the blue line the model estimation obtained from the control group and the green line is the model obtained from the obese group

After the effort test, the children answered some questions in order to test acceptance of the sensor. Results (Figs. 7, 8) (p < 0.05):
  • Wearing the t-shirt all day. All children were asked about the possibility of wearing this new ubiquitous sensor all day while completing the activities proposed in the ETIOBE treatment. All children (n = 20) save one answered positively to this question. Moreover, 45 % of children stated that they would wear the t-shirt with low or none inconvenience.

  • Acceptance of the t-shirt as a daily element. Another important issue was testing whether children accept the t-shirt as a regular tool in their daily life. For that purpose, all children were asked whetehr it would bother them if other people saw them wear this sensor. No child answered “yes, I can’t wear this sensor” and only 3 children (15 %) answered “It would bother me a lot” or “It would bother me a quite a bit. The remaining children (85 %) answered “It would bother be a little bit or nothing.

4 Discussion

The results section has presented the validation results of the two non-commercial acquiring systems designed by I3BH: firstly, the data of a clinical study with electronic records have been presented, and secondly, a validation study for the t-shirt has been shown. This validation proves the feasibility of this tool for obtaining the metabolic, physical activity and ergonomic response of the obese child during an effort test. The results of the t-shirt validations show a good response for detecting the physical activity levels, heart rate and metabolic response during effort of the obese children. The one-system design of the t-shirt provides advantages over simultaneous accelerometry and HR monitoring with separate monitors [48], including lower researcher and participant burden, and intrinsic time synchronization. Corder et al. [49] found, in another work with models, a good correlation with a combined model for children (r2 = 0.86) that was better than single-measure models (HR and ACC only models). These data are consistent with data from the validation study of the t-shirt where the model with a low bias was the model combining HR and ACC signals. In addition, use of a standard unit of measurement for accelerometry (Gcounts) can remove the problems associated with the use of different monitor types for measuring acceleration [50] and will enable the comparison of results of future studies for different devices and experiments that work with this standard unit. It remains to be shown whether the combined predictions of the t-shirt are significantly more accurate than single-objective methods for the prediction of Physical Activity Energy Expenditure in obese children. The positive response of patients to using the ambulatory system during the day and their acceptance as a daily element ensures children’s cooperation with this kind of devices and their ubiquitous capacity. All these data confirm that the developed t-shirt is a good tool for the ubiquitous monitoring of physical activity patterns in obese children during clinical treatment.

5 Conclusions

In this article, a brief description of the ETIOBE project was presented. Then, the ubiquitous monitoring layer elements were described: electronic records (for diet and physical activity recall), blood pressure monitor (for the study of hypertension), pulse oximeter (for studying sleep disorders) and a special t-shirt (personal monitor to obtain physical activity patterns). The various validations performed so far show that the monitoring layer developed can help to achieve the main goals of the ETIOBE project: to improve treatment adherence, promote self-control mechanisms in patients, and prevent relapses. At this time, and all through 2012, a validation of the complete ETIOBE system (Clinical Support System, Home Support System and Mobile Support System) is being performed. This validation will evaluate the effectiveness of the system. The results of this validation will be analysed in a future paper.


This study was funded by Ministerio de Educación y Ciencia Spain, Project Game Teen (TIN2010-20187) and partially by projects Consolider-C (SEJ2006-14301/ PSIC), “CIBER of Physiopathology of Obesity and Nutrition, an initiative of ISCIII” and Excellence Research Program PROMETEO (Generalitat Valenciana. Conselleria de Educación, 2008-157).

Copyright information

© Springer-Verlag London Limited 2012