Reproducibility of Dietary Intake Measurement From Diet Diaries, Photographic Food Records, and a Novel Sensor Method.

Objective: No data currently exist on the reproducibility of photographic food records compared to diet diaries, two commonly used methods to measure dietary intake. Our aim was to examine the reproducibility of diet diaries, photographic food records, and a novel electronic sensor, consisting of counts of chews and swallows using wearable sensors and video analysis, for estimating energy intake. Method: This was a retrospective analysis of data from a previous study, in which 30 participants (15 female), aged 29 ± 12 y and having a BMI of 27.9 ± 5.5, consumed three identical meals on different days. Four different methods were used to estimate total mass and energy intake on each day: (1) weighed food record; (2) photographic food record; (3) diet diary; and (4) novel mathematical model based on counts of chews and swallows (CCS models) obtained via the use of electronic sensors and video monitoring system. The study staff conducted weighed food records for all meals, took pre- and post-meal photographs, and ensured that diet diaries were completed by participants at the end of each meal. All methods were compared against the weighed food record, which was used as the reference method. Results: Reproducibility was significantly different between the diet diary and photographic food record for total energy intake (p = 0.004). The photographic record had greater reproducibility vs. the diet diary for all parameters measured. For total energy intake, the novel sensor method exhibited good reproducibility (repeatability coefficient (RC) of 59.9 (45.9, 70.4), which was better than that for the diet diary [RC = 79.6 (55.5, 103.3)] but not as repeatable as the photographic method [RC = 43.4 (32.1, 53.9)]. Conclusion: Photographic food records offer superior precision to the diet diary and, therefore, would be valuable for longitudinal studies with repeated measures of dietary intake. A novel electronic sensor also shows promise for the collection of longitudinal dietary intake data.

Dietary Intake and Physical Activity Assessment: Current Tools, Techniques, and Technologies for Use in Adult Populations.

Accurate assessment of dietary intake and physical activity is a vital component for quality research in public health, nutrition, and exercise science. However, accurate and consistent methodology for the assessment of these components remains a major challenge. Classic methods use self-report to capture dietary intake and physical activity in healthy adult populations. However, these tools, such as questionnaires or food and activity records and recalls, have been shown to underestimate energy intake and expenditure as compared with direct measures like doubly labeled water. This paper summarizes recent technological advancements, such as remote sensing devices, digital photography, and multisensor devices, which have the potential to improve the assessment of dietary intake and physical activity in free-living adults. This review will provide researchers with emerging evidence in support of these technologies, as well as a quick reference for selecting the "right-sized" assessment method based on study design, target population, outcome variables of interest, and economic and time considerations. THEME INFORMATION: This article is part of a theme issue entitled Innovative Tools for Assessing Diet and Physical Activity for Health Promotion, which is sponsored by the North American branch of the International Life Sciences Institute.

The Pediatric SmartShoe: Wearable Sensor System for Ambulatory Monitoring of Physical Activity and Gait.

Cerebral palsy (CP) is a group of nonprogressive neuro-developmental conditions occurring in early childhood that causes movement disorders and physical disability. Measuring activity levels and gait patterns is an important aspect of CP rehabilitation programs. Traditionally, such programs utilize commercially available laboratory systems, which cannot to be utilized in community living. In this study, a novel, shoe-based, wearable sensor system (pediatric SmartShoe) was tested on 11 healthy children and 10 children with CP to validate its use for monitoring of physical activity and gait. Novel data processing techniques were developed to remove the effect of orthotics on the sensor signals. Machine learning models were developed to automatically classify the activities of daily living. The temporal gait parameters estimated from the SmartShoe data were compared against reference measurements on a GAITRite mat. A leave-one-out cross-validation method indicated a 95.3% average accuracy of activity classification (for sitting, standing, and walking) for children with CP and 96.2% for healthy children. Average relative errors in gait parameter estimation (gait cycle, stance, swing, and step time, % single support time on both lower extremities, along with cadence) ranged from 0.2% to 6.4% (standard deviation range = 1.4%-9.9%). These results suggest that the pediatric SmartShoe can accurately measure physical activity and gait of children with CP and can potentially be used for ambulatory monitoring.

Development of the RT-GAIT, a Real-Time feedback device to improve Gait of individuals with stroke.

Regaining the ability to walk is a major rehabilitation goal after a stroke. Recent research suggests that, in people with stroke, task-oriented and intensive rehabilitation strategies can drive cortical reorganization and increase activity levels. This paper describes development and pilot testing of a novel wearable device for Real-Time Gait and Activity Improving Telerehabilitation (RT-GAIT), designed for use with such rehabilitation strategies. The RT-GAIT provides auditory or tactile feedback to the individual wearing the platform. The feedback is based on the amount of time spent in stance phase on each foot, as measured by the pressure sensors embedded into the insoles. The system was initially bench-validated using sensor signals collected in a previous study. Next, a clinical case study was conducted with one post-stroke individual. The results of the case study suggest that the RT-GAIT device can potentially improve the gait parameters. Mean difference in stance times between the healthy limb and paretic limb was improved by 48% and the standard deviation for the same was improved by 87.5%, between baseline measurements and the measurements taken after the treatment with the RT-GAIT.

Highly accurate recognition of human postures and activities through classification with rejection.

Monitoring of postures and activities is used in many clinical and research applications, some of which may require highly reliable posture and activity recognition with desired accuracy well above 99% mark. This paper suggests a method for performing highly accurate recognition of postures and activities from data collected by a wearable shoe monitor (SmartShoe) through classification with rejection. Signals from pressure and acceleration sensors embedded in SmartShoe are used either as raw sensor data or after feature extraction. The Support vector machine (SVM) and multilayer perceptron (MLP) are used to implement classification with rejection. Unreliable observations are rejected by measuring the distance from the decision boundary and eliminating those observations that reside below rejection threshold. The results show a significant improvement (from 97.3% ± 2.3% to 99.8% ± 0.1%) in the classification accuracy after the rejection, using MLP with raw sensor data and rejecting 31.6% of observations. The results also demonstrate that MLP outperformed the SVM, and the classification accuracy based on raw sensor data was higher than the accuracy based on extracted features. The proposed approach will be especially beneficial in applications where high accuracy of recognition is desired while not all observations need to be assigned a class label.

Validation of two novel monitoring devices to measure physical activity in healthy women.

Measurement of physical activity is increasingly important in health research. We sought to determine the accuracy and sensitivity to non-exercise activity of three activity monitors worn simultaneously by healthy adult women participating in a structured activity protocol. Ten normal-weight women wore the Intelligent Device for Energy Expenditure and Activity (IDEEA), the SmartShoe, and the SenseWear Armband, during activities that included standing, sitting still, sitting and fidgeting, lying down, and walking at varying speeds. Percentage of time postures correctly identified was determined for the IDEEA and the SmartShoe, and activity counts collected from all three devices were compared. Posture was detected with high accuracy by both the IDEEA and the SmartShoe (97.4% and 94.2% accuracy, respectively). The SmartShoe showed superior sensitivity to movement while seated ("fidgeting") compared with the IDEEA (p=0.004 and 0.049 difference between postures, respectively); all three devices distinguished between fast and slow walking. Data support the ability of the IDEEA and the SmartShoe to recognize basic postures in healthy normal-weight women, as well as to detect fidgeting within the seated position.

Using decision trees to measure activities in people with stroke.

Improving community mobility is a common goal for persons with stroke. Measuring daily physical activity is helpful to determine the effectiveness of rehabilitation interventions. In our previous studies, a novel wearable shoe-based sensor system (SmartShoe) was shown to be capable of accurately classify three major postures and activities (sitting, standing, and walking) from individuals with stroke by using Artificial Neural Network (ANN). In this study, we utilized decision tree algorithms to develop individual and group activity classification models for stroke patients. The data was acquired from 12 participants with stroke. For 3-class classification, the average accuracy was 99.1% with individual models and 91.5% with group models. Further, we extended the activities into 8 classes: sitting, standing, walking, cycling, stairs-up, stairs-down, wheel-chair-push, and wheel-chair-propel. The classification accuracy for individual models was 97.9%, and for group model was 80.2%, demonstrating feasibility of multi-class activity recognition by SmartShoe in stroke patients.

Development of SmartStep: an insole-based physical activity monitor.

In our previous research we developed a SmartShoe--a shoe based physical activity monitor that can reliably differentiate between major postures and activities, accurately estimate energy expenditure of individuals, measure temporal gait parameters, and estimate body weights. In this paper we present the development of the next stage of the SmartShoe evolution--SmartStep, a physical activity monitor that is fully integrated into an insole, maximizing convenience and social acceptance of the monitor. Encapsulating the sensors, Bluetooth Low Energy wireless interface and the energy source within an assembly repeatedly loaded with high forces created during ambulation presented new design challenges. In this preliminary study we tested the ability of the SmartStep to measure the pressure differences between static weight-bearing and non-weight-bearing activities (such as no load vs. sitting vs. standing) as well as capture pressure variations during walking. We also measured long-term stability of the sensors and insole assembly under cyclic loading in a mechanical testing system.

A comparison of energy expenditure estimation of several physical activity monitors.

INTRODUCTION: Accurately and precisely estimating free-living energy expenditure (EE) is important for monitoring energy balance and quantifying physical activity. Recently, single and multisensor devices have been developed that can classify physical activities, potentially resulting in improved estimates of EE. PURPOSE: This study aimed to determine the validity of EE estimation of a footwear-based physical activity monitor and to compare this validity against a variety of research and consumer physical activity monitors. METHODS: Nineteen healthy young adults (10 men, 9 women) completed a 4-h stay in a room calorimeter. Participants wore a footwear-based physical activity monitor as well as Actical, ActiGraph, IDEEA, DirectLife, and Fitbit devices. Each individual performed a series of postures/activities. We developed models to estimate EE from the footwear-based device, and we used the manufacturer's software to estimate EE for all other devices. RESULTS: Estimated EE using the shoe-based device was not significantly different than measured EE (mean ± SE; 476 ± 20 vs 478 ± 18 kcal, respectively) and had a root-mean-square error of 29.6 kcal (6.2%). The IDEEA and the DirectLlife estimates of EE were not significantly different than the measured EE, but the ActiGraph and the Fitbit devices significantly underestimated EE. Root-mean-square errors were 93.5 (19%), 62.1 kcal (14%), 88.2 kcal (18%), 136.6 kcal (27%), 130.1 kcal (26%), and 143.2 kcal (28%) for Actical, DirectLife, IDEEA, ActiGraph, and Fitbit, respectively. CONCLUSIONS: The shoe-based physical activity monitor provides a valid estimate of EE, whereas the other physical activity monitors tested have a wide range of validity when estimating EE. Our results also demonstrate that estimating EE based on classification of physical activities can be more accurate and precise than estimating EE based on total physical activity.

Evaluation of Chewing and Swallowing Sensors for Monitoring Ingestive Behavior.

Monitoring Ingestive Behavior (MIB) of individuals is of special importance to identify and treat eating patterns associated with obesity and eating disorders. Current methods for MIB require subjects reporting every meal consumed, which is burdensome and tend to increase the reporting bias over time. This study presents an evaluation of the burden imposed by two wearable sensors for MIB during unrestricted food intake: a strain sensor to detect chewing events and a throat microphone to detect swallowing sounds. A total of 30 healthy subjects with various levels of adiposity participated in experiments involving the consumption of four meals in four different visits. A questionnaire was handled to subjects at the end of the last visit to evaluate the sensors burden in terms of the comfort levels experienced. Results showed that sensors presented high comfort levels as subjects indicated that the way they ate their meal was not considerably affected by the presence of the sensors. A statistical analysis showed that chewing sensor presented significantly higher comfort levels than the swallowing sensor. The outcomes of this study confirmed the suitability of the chewing and swallowing sensors for MIB and highlighted important aspects of comfort that should be addressed to obtain acceptable and less burdensome wearable sensors for MIB.

A Sensor System for Automatic Detection of Food Intake Through Non-Invasive Monitoring of Chewing.

Objective and automatic sensor systems to monitor ingestive behavior of individuals arise as a potential solution to replace inaccurate method of self-report. This paper presents a simple sensor system and related signal processing and pattern recognition methodologies to detect periods of food intake based on non-invasive monitoring of chewing. A piezoelectric strain gauge sensor was used to capture movement of the lower jaw from 20 volunteers during periods of quiet sitting, talking and food consumption. These signals were segmented into non-overlapping epochs of fixed length and processed to extract a set of 250 time and frequency domain features for each epoch. A forward feature selection procedure was implemented to choose the most relevant features, identifying from 4 to 11 features most critical for food intake detection. Support vector machine classifiers were trained to create food intake detection models. Twenty-fold cross-validation demonstrated per-epoch classification accuracy of 80.98% and a fine time resolution of 30 s. The simplicity of the chewing strain sensor may result in a less intrusive and simpler way to detect food intake. The proposed methodology could lead to the development of a wearable sensor system to assess eating behaviors of individuals.

A robust classification scheme for detection of food intake through non-invasive monitoring of chewing.

Automatic methods for food intake detection are needed to objectively monitor ingestive behavior of individuals in a free living environment. In this study, a pattern recognition system was developed for detection of food intake through the classification of jaw motion. A total of 7 subjects participated in laboratory experiments that involved several activities of daily living: talking, walking, reading, resting and food intake while being instrumented with a wearable jaw motion sensor. Inclusion of such activities provided a high variability to the sensor signal and thus challenged the classification task. A forward feature selection process decided on the most appropriate set of features to represent the chewing signal. Linear and RBF Support Vector Machine (SVM) classifiers were evaluated to find the most suitable classifier that can generalize the high variability of the input signal. Results showed that an average accuracy of 90.52% can be obtained using Linear SVM with a time resolution of 15 sec.

Recognition of household and athletic activities using SmartShoe.

The ability to provide real time feedback concerning a person's activity level and energy expenditure can be beneficial for improving activity levels of individuals. Examples include biofeedback systems used for body weight and physical activity management and biofeedback systems for rehabilitation of stroke patients. A critical aspect of any such system is being able to accurately classify data in real-time so that active and timely feedback can be provided. In the paper we demonstrate feasibility of real-time recognition of multiple household and athletic activities on a cell phone using the data collected by a wearable sensor system consisting of SmartShoe sensor and a wrist accelerometer. The experimental data were collected for multiple household and athletic activities performed by a healthy individual. The data was used to train two neural networks, one to be used primarily for sedentary individuals and one for more active individuals. Classification of household activities including ascending stairs, descending stairs, doing the dishes, vacuuming, and folding laundry, achieved 89.62% average accuracy. Classification of athletic activities such as jumping jacks, swing dancing, and ice skating, was performed with 93.13% accuracy. As proof of real-time processing on a mobile platform the trained neural network for healthy individuals was timed and required less than 4 ms to perform each feature vector construction and classification.

Highly accurate classification of postures and activities by a shoe-based monitor through classification with rejection.

Monitoring human beings' major daily activities is important for many biomedical studies. Some monitoring applications may require highly reliable identification of certain postures and activities with desired accuracies well above 99% mark. This paper suggests a method for performing highly accurate classification of postures and activities from data collected by a wearable shoe monitor (SmartShoe) through classification with rejection. The classifier used in this study is support vector machines that uses posterior probability based on the distance of an observation to the separating hyperplane to reject unreliable observations. The results show that a significant improvement (from 95.2% ± 3.5% to 99% ± 1%) of the classification accuracy has been reached after the rejection, as compared to the accuracy reported previously. Such an approach will be especially beneficial in application where high accuracy of recognition is desired while not all observations need to be assigned a class label.

Classification of posture and activities by using decision trees.

Obesity prevention and treatment as well as healthy life style recommendation requires the estimation of everyday physical activity. Monitoring posture allocations and activities with sensor systems is an effective method to achieve the goal. However, at present, most devices available rely on multiple sensors distributed on the body, which might be too obtrusive for everyday use. In this study, data was collected from a wearable shoe sensor system (SmartShoe) and a decision tree algorithm was applied for classification with high computational accuracy. The dataset was collected from 9 individual subjects performing 6 different activities--sitting, standing, walking, cycling, and stairs ascent/descent. Statistical features were calculated and the classification with decision tree classifier was performed, after which, advanced boosting algorithm was applied. The computational accuracy is as high as 98.85% without boosting, and 98.90% after boosting. Additionally, the simple tree structure provides a direct approach to simplify the feature set.

Automatic detection of temporal gait parameters in poststroke individuals.

Approximately one-third of people who recover from a stroke require some form of assistance to walk. Repetitive task-oriented rehabilitation interventions have been shown to improve motor control and function in people with stroke. Our long-term goal is to design and test an intensive task-oriented intervention that will utilize the two primary components of constrained-induced movement therapy: massed, task-oriented training and behavioral methods to increase use of the affected limb in the real world. The technological component of the intervention is based on a wearable footwear-based sensor system that monitors relative activity levels, functional utilization, and gait parameters of affected and unaffected lower extremities. The purpose of this study is to describe a methodology to automatically identify temporal gait parameters of poststroke individuals to be used in assessment of functional utilization of the affected lower extremity as a part of behavior enhancing feedback. An algorithm accounting for intersubject variability is capable of achieving estimation error in the range of 2.6-18.6% producing comparable results for healthy and poststroke subjects. The proposed methodology is based on inexpensive and user-friendly technology that will enable research and clinical applications for rehabilitation of people who have experienced a stroke.

Swallowing detection by sonic and subsonic frequencies: a comparison.

The detection of swallowing events by acoustic means represents an important tool to assess and diagnose swallowing disorders as well as to objectively monitor ingestive behavior of individuals. Acoustic sensors used to register swallowing sounds may also capture sound artifacts arising from intrinsic speech and external noise affecting the detection. In this paper we tested if subsonic frequencies are less prone to artifacts from speech, chewing and other intrinsic sounds than sonic frequencies. A simple method using a throat and an ambient microphone was employed to compare the swallowing detection accuracy by acoustic signals acquired in the sonic (20-2500 Hz) and subsonic (≤ 5 Hz) ranges. Averaged recall values were higher than 85% for both ranges. However, averaged precision values of 50% for subsonic frequencies and of 42% for sonic frequencies were caused by a high number of false positives. These results indicated no significant difference between averaged precision values which may suggest that subsonic frequencies were not less prone to intrinsic sound artifacts than frequencies in the sonic range. Further examination with the addition of a signal classification layer is proposed as a future step to confirm this statement.

Prediction of bodyweight and energy expenditure using point pressure and foot acceleration measurements.

Bodyweight (BW) is an essential outcome measure for weight management and is also a major predictor in the estimation of daily energy expenditure (EE). Many individuals, particularly those who are overweight, tend to underreport their BW, posing a challenge for monitors that track physical activity and estimate EE. The ability to automatically estimate BW can potentially increase the practicality and accuracy of these monitoring systems. This paper investigates the feasibility of automatically estimating BW and using this BW to estimate energy expenditure with a footwear-based, multisensor activity monitor. The SmartShoe device uses small pressure sensors embedded in key weight support locations of the insole and a heel-mounted 3D accelerometer. Bodyweight estimates for 9 subjects are computed from pressure sensor measurements when an automatic classification algorithm recognizes a standing posture. We compared the accuracy of EE prediction using estimated BW compared to that of using the measured BW. The results show that point pressure measurement is capable of providing rough estimates of body weight (root-mean squared error of 10.52 kg) which in turn provide a sufficient replacement of manually-entered bodyweight for the purpose of EE prediction (root-mean squared error of 0.7456 METs vs. 0.6972 METs). Advances in the pressure sensor technology should enable better accuracy of body weight estimation and further improvement in accuracy of EE prediction using automatic BW estimates.

Design of a instrumentation module for monitoring ingestive behavior in laboratory studies.

The development of accurate and objective tools for monitoring of ingestive behavior (MIB) is one of the most important needs facing studies of obesity and eating disorders. This paper presents the design of an instrumentation module for non-invasive monitoring of food ingestion in laboratory studies. The system can capture signals from a variety of sensors that characterize ingestion process (such as acoustical and other swallowing sensors, strain sensor for chewing detection and self-report buttons). In addition to the sensors, the data collection system integrates time-synchronous video footage that can be used for annotation of subject's activity. Both data and video are simultaneously and synchronously acquired and stored by a LabVIEW-based interface specifically developed for this application. This instrumentation module improves a previously developed system by eliminating the post-processing stage of data synchronization and by reducing the risks of operator's error.

Characterizing walking activity in people with stroke.

Stroke is the leading cause of disability in the U.S. Many people with stroke have limited walking ability and are inactive. In this paper we describe a novel shoe based sensor, SmartShoe, and a signal processing technique to identify walking activity. The technique was validated with 6 people with walking impairment due to stroke. The results suggest that the SmartShoe is able to accurately identify walking activity. This device could be used to monitor walking activity as well as provide behavioral enhancing feedback to increase activity levels and walking ability in people with stroke for extended periods of time in the real world.