Charlotte Lund Rasmussen

There is growing concern surrounding the health and wellbeing of children, as many do not meet physical activity guidelines. Augmented Reality (AR) has been successfully employed in smartphone games to promote physically active play. However, the potential social and physical risks of children moving around a play space with a smartphone have not been explored. This study aimed to investigate the impact of an AR enhanced playground on peer connection, simulator sickness and miscoordination (trips, bumps, loss of balance and falls). Seventeen pairs of 5–8-year-old children participated in an AR animal search (marker-based AR app on a smartphone) and plastic toy animal search tasks, in a laboratory playground. Searches were conducted under low-structure play conditions and during high-structure obstacle courses. Questionnaires on connection and sickness were completed after each condition and miscoordination events retrospectively coded from video. Peer connection was unaffected by animal format (AR: 4.9±2.2/7 vs non-AR: 4.7±2.2, p=0.311). Overall symptoms of simulator sickness were slightly higher during AR versus toys (AR: 0.6±1.3/10 vs non-AR: 0.4±0.9, p=0.035) however analysis of individual symptoms revealed no significant differences and 92% of questionnaire responses indicated zero symptoms. The number of miscoordination events were lower in AR conditions, even when movement intensity was taken into account. An AR enhanced playground does not appear to substantially alter risks of decreased social connection, symptoms of simulator sickness or physical miscoordination above that of a non-AR playground and therefore the benefits they could produce of increased motivation to participate in physically active play may justify their use.

Background: Accurate measurements of children’s digital technology use are essential for understanding its potential implications on health and well-being. Wearable cameras can provide such measurements, but image coding is a high burden for researchers. Machine learning–based object-recognition models have the potential to reduce this burden by identifying images containing technology. Objective: This study aims to evaluate the performance of an object recognition model, the Square Eyes model, as a screening tool for identifying technologies in wearable camera images among children for further human review, as well as to examine the potential influence of face-blurring methods on the model’s performance. Methods: This study used data collected on 48 children (aged 3‐14 y) during an approximately 1-hour laboratory session. The children performed various technology-related tasks while wearing a camera. A total of 221,226 images were coded by humans and processed through the Square Eyes model. The performance of the Square Eyes model as a screening tool was evaluated by (1) assessing agreement between the model and human coding; (2) evaluating the N-back algorithm, an algorithm embedded in the model aimed to flag images requiring human review; and (3) examining the potential influence of facial-blurring on model performance. Results: Humans detected technology in 92,745 (41.9%) images, and the Square Eyes model detected technologies with an overall accuracy of 78.0%. When considering specific technologies, agreement between the model and human coders was the highest for Television (n=19,148, 54.3%) and Laptop (n=8492, 44.5%) and lowest for smaller devices such as Smartphone (n=2600, 31.3%) and Tablet (n=3685, 25.1%). The model’s N-back algorithm effectively flagged images that required further human review, with only 7144 (3.2%) images that were not flagged for screening containing a human-coded technology. An explorative analysis indicated that using a square face-blurring with border could have reduced the model’s ability to accurately detect technologies. Conclusions: The Square Eyes model demonstrated overall satisfying accuracy in detecting technologies and successfully flagged images that required further review by humans. These findings suggest that the model could be used as an effective screening tool for reducing the burden of human coding. However, the model could be improved to more accurately detect smaller devices, and the form of facial blurring in images should be considered.

Background: Technology use by children has implications for their physical, mental, and social health and development. There are many challenges in measuring modern screen use by children. To better understand the potential impact of technology on children, it is crucial to have robust methods to measure technology use. Visual recording data may be particularly well suited to capture modern digital technology use (eg, rapid transitions, multitasking, and device switching). It would be valuable to directly compare the 2 commonly used methods (wearable camera images and room video frames) to determine what is captured best by each and how they compare when both are coded by human annotation. Objective: This study aimed to compare wearable camera images and room video frames to understand what can be captured from each source. Methods: In a laboratory study, children (aged 3‐14 years) performed various tasks while wearing a wearable camera and being video recorded by room video cameras. Comparisons between the coding of the wearable camera images and room video frames were performed on a second-by-second basis. Confusion matrices were generated to identify patterns of differences in classifications between wearable camera images and room video frame coding. Percent agreement between the wearable camera image and room video frame coding was calculated based on the confusion matrix. Results: Across 44 participants, 182,451 contemporaneous wearable camera images and room video frames were coded and merged. There were generally low levels of agreement between the codes. The highest agreement was found for “desktop” and “handheld gaming,” and the lowest agreement was found for “smartwatch.” The wearable camera and room video methods varied in sampling rate, camera movement, and field of view, leading to differences in what was captured. Conclusions: Room video might better capture technology that is within a set space and the context of the technology use, while wearable cameras could better capture active technology use, the content of technology, and different locations. Capturing technology use by children is difficult, with no gold standard and no one ideal method.

Background: Technology is integrated into many children’s daily lives, with parents’ and health professionals’ perspectives shaping children’s technology use. Measuring and understanding these perspectives are essential for developing strategies for supporting adults in decision-making that help children thrive in a digital world. Objective: This systematic review aimed to investigate the psychometric properties of questionnaires used to assess parents’ and health professionals’ perspectives on young children’s use of technology related to health, well-being, and development. The secondary aim was to synthesize findings on these perspectives. Methods: Peer-reviewed papers published between January 2010 and September 2024 were identified through searches in 7 electronic databases. Studies were included if they examined parental or health care professionals’ perspectives on technology use among children aged birth to 5 years. Two reviewers (CLR and IPHA) independently conducted the data extraction and study quality assessment. Reported psychometric properties of the questionnaires were synthesized. Deductive thematic analysis was used to explore the content focus of the questionnaire used in the included studies and synthesize the reported perspectives. Results: In total, 85 studies were included, all involving parents. No study investigated health professionals’ perspectives. The methodological quality of the studies was generally low, with 62 studies scoring below the threshold for acceptable quality. In total, 52 studies reported psychometric properties of the questionnaires used, of which, only 15 studies reported more than 1 measure of validity or reliability. A total of 75 studies reported participants’ perspectives on children’s technology use. Findings revealed that parents generally supported the role of digital devices in enhancing learning but expressed concerns about potential negative impacts on children’s physical health, emotions, and behaviors. Conclusions: Parents’ perspectives on children’s technology use were frequently assessed through questionnaires, though the validity of these questionnaires was often poor, with limited psychometric testing. Parental perspectives were mixed with educational benefits being recognized, while countered with concerns about the impact on children’s physical health and development. High-quality questionnaires are needed to generate stronger evidence informing strategies to support families in technology use decision-making with and for children.

BACKGROUND Accurate measurements of children’s digital technology use is essential for understanding its potential implications on health and wellbeing. Wearable cameras can provide such measurements, but the image coding is a high researcher burden. Machine learning based object recognition models have the potential to reduce this burden by identifying images containing technologies. OBJECTIVE To evaluate the performance of an object recognition model, the Square Eyes model, as a screening tool for identifying technologies in wearable camera images among children for further human review. Additionally, to examine the potential influence of face-blurring methods on the model’s performance. METHODS This study used data collected on 48 children (3-14 years of age) during a ~1 hour laboratory session. The children performed various technology related tasks while wearing a camera. A total of 211,226 images were coded by humans and processed through the Square Eyes model. The performance of the Square Eyes model as a screening tool was evaluated by 1) assessing agreement between the model and human coding; 2) evaluating the N-back algorithm, an algorithm embedded in the model aimed to flag images requiring human review, and 3) examining the potential influence of facial-blurring on model performance. RESULTS Humans detected technology in 92,745 (41.9%) images The Square Eyes model detected technologies with an overall accuracy of 78.0%. When considering specific technologies, agreement between the model and human coders was the highest for Television (54.3%) and Laptop (44.5%) and lowest for smaller devices such as Smartphone (31.3%) and Tablet (25.1%). The model’s N-back algorithm effectively flagged images that required further human review, with only 7,144 images (3.2% of all images) not flagged for screening containing a human-coded technology. An explorative analysis indicated that using a square face-blurring with border could have reduced the model’s ability to accurately detect technologies. CONCLUSIONS The Square Eyes model demonstrated overall satisfying accuracy in detecting technologies and successfully flagged images that required further review by humans. These findings suggest that the model could be used as an effective screening tool for reducing the burden of human coding. However, the model could be improved for accurately detecting smaller devices and the form of facial blurring in images should be considered.

Aim : This study aimed to assess the accuracy of deep learning (DL) models trained and tested on thigh-worn accelerometry from children to estimate their postures and movements and compare their performance metrics against more traditional machine learning (ML) models and a decision tree algorithm. Method : This study utilized data from a laboratory study involving 48 children (3–14 years old), wearing a SENS accelerometer on the thigh during a 1-hr structured laboratory session. Human coders annotated nine different postures and movement categories from video data (30 Hz). The raw accelerometer data was processed using two traditional ML models: Balanced Random Forest and Extreme Gradient Boost; two DL models: a Long Short-Term Memory-Convolutional Neural Network with (a) standard deviation and (b) minmax normalization; and a decision tree algorithm (ActiMotus). The efficacy of each model was compared using accuracy, precision, recall, and F1-scores. Results : Across all movement categories, the Extreme Gradient Boost ML model (F1 = 0.769) was found to slightly outperform the DL models (F1 = 0.743 and 0.728), with all ML and DL models performing statistically better than the widely used decision tree model (F1 = 0.645). These differences in performance were consistent across all postures and movement categories with the exception of standing and walking. Conclusions : The results of this study suggest that DL models are not necessarily superior to ML models for the classification of children’s posture and movements, but both are more accurate than a currently available decision tree algorithm. Future work should continue to refine deep and ML models for classifying children’s physical behaviors.

Aim : This study compared posture and movement classification accuracy and agreement of two thigh-worn accelerometers (SENS and Axivity) among children aged 3–14 years using the same classification algorithm. Method : Forty-eight children attended a structured 1-hr laboratory session, of which 46 provided valid data. The session was video recorded and accelerometer data were collected from accelerometers worn on the front of the thigh. Human-coded video provided the reference standard to calculate accuracy of ActiMotus algorithm posture and movement classifications. Agreement between devices was compared using multiple metrics. Accelerometer variables (acceleration variation and sensor orientation metrics) used for classification were also compared between devices. Results : The devices showed comparable classification accuracy (F1-score = 66%–68%; balanced accuracy = 82%–83%) with a trend for SENS to have slightly higher accuracy, but the only significant difference was found for running where SENS had lower metrics (F1-score = 55%; balanced accuracy = 77%) than Axivity (F1-score = 56%; balanced accuracy = 81%). Agreement in posture and movement duration estimates was highest for lying, sitting, and standing (85%–88%), but lower for dynamic movements, particularly running (65%) and stair climbing (53%). When comparing the accelerometer variables, the greatest relative difference was observed for the SD of x -axis for running (mean absolute percent error = 18.4%, SD = 8.9). Conclusions : We observed differences between SENS and Axivity for posture and movement classification accuracy and agreement which may be linked to differences in hardware features and thereby derived accelerometer variables. The results suggest that hardware choice might influence estimates of postures and movements among children even when using the same placement and algorithm.

Aim : To evaluate the criterion validity of the SENS Motion Activity algorithm to classify postures and movements among children aged between 3 and 14 years in a laboratory setting by comparing with human-coded video. Method : Data were collected on 48 Australian children who attended a structured ∼1 hr data collection session at a laboratory with their caregivers. The session was video recorded, and thigh acceleration was measured using a SENS accelerometer. Data from the accelerometer were processed and classified into four postures and movements using the SENS algorithm. Human-coded video provided the reference standard to calculate the performance metrics sensitivity, specificity, precision, F1-score, and balanced accuracy. Results : Overall, the SENS Motion Activity algorithm classified postures and movements with performance metrics F1-score of 73% and balanced accuracy of 87%. Lying/sitting had the highest F1-score and balanced accuracy (99%). Standing, walking, and running had somewhat lower performance metrics (F1-score and balanced accuracy of 69%–77%, 57%–92%, and 68%–82%, respectively). Difficulties with human coding likely contributed to lower performance. A higher balanced accuracy was found for boys compared with girls for running. The algorithm had better performance for children older than 11 years compared with younger children overall and for walking and running. Conclusion : Our results suggest that the SENS Motion Activity algorithm can provide accurate measurements for detecting lying/sitting, standing, walking, and running among children. Further research could usefully explore algorithm development for classifying movement in sitting and standing postures and brief sporadic movements, especially among children younger than 11 years.

Background: Smartwatch activity trackers are devices that measure physical activity levels with features that aim to encourage physically active behaviors. These devices have shown promise for increasing physical activity levels and reducing sedentary behaviors among school-aged children, adolescents, and adults. Recently, commercially available products have been adapted so that they are suitable for use by preschool-aged children. However, it is unclear whether the intended use of these devices is feasible and effective in young children. Objective: The purpose of this study was to explore parents’ perspectives on the use of smartwatch activity trackers by young children. Methods: Semistructured interviews were conducted with 22 parents (17/22, 77% female) of children aged 3-5 years. Interviews explored perspectives on the feasibility of their children wearing the devices, implications of use by young children, and how families could make use of these devices to support their children’s physically active behaviors. Interviews were audio-recorded, transcribed verbatim, and data analyzed using thematic analysis. Results: Parents perceived that the use of these devices by young children is feasible, with developmental stage or abilities and personality or temperament being important individual determinants of feasibility. However, parents expressed concerns related to the devices providing extrinsic motivation to move, being disruptive or distracting, being a burden on parents, and for the safety and privacy of their child’s information. Most parents believed that young children are inherently active and do not need devices to support physical activity. Furthermore, most parents expressed an interest in knowing how physically active their children were and thought that there may be a role for these devices for children who are less physically active. Conclusions: Parents reported developmental stage or abilities and temperament as relevant considerations related to the feasibility of smartwatch activity tracker use by young children. Parents also indicated that there is a potential role for these devices in young, less active children.

Background: Only 10% of Australian children meet the recommended daily physical activity guidelines. Augmented reality (AR) is increasingly being used in primary education and clinical rehabilitation, with high enjoyment and motivation to participate frequently reported. AR has increased physical activity participation in adult populations, but whether AR can increase physical activity in young children has yet to be investigated. Objective: This study aimed to determine if an indoor AR-enhanced playground is enjoyed by young children and prompts physical activity in both low- and high-structured scavenger hunts. Methods: Seventeen pairs of 5‐ to 8-year-olds participated in 2 animal search tasks (ie, AR and non-AR) in 2 activity structure levels (ie, low-structured and high-structured) in a 2×2 repeated measures design in an indoor laboratory playground. Children searched for either AR animals (a custom AR app on a smartphone) or toy animals and followed a set obstacle course route (high-structured) or moved wherever they wished (low-structured). Questionnaires assessed child enjoyment, perceived physical activity, and caregiver perception of enjoyment. Thigh-worn accelerometers (SENS; SENS Innovation ApS) assessed postures and movements, and a video camera recorded engagement time. Results: Children rated AR conditions (low-structured: mean 4.4, SD 1.0, and high-structured: mean 4.5, SD 0.9) as more enjoyable than high-structured non-AR (mean 4.1, SD 1.0; P=.03). When asked which condition was the most enjoyable, 15 chose the low-structured AR, followed by the high-structured AR (n=11) and low-structured non-AR (n=8). Caregiver perception of children’s enjoyment ratings generally aligned. Ratings of perceived physical activity level were the same in all conditions (mean 4.3, SD 0.7; P>.05). Accelerometry showed that a greater percentage of time was spent in low-intensity postures and movements during AR conditions (AR: mean 50%, SD 13% vs non-AR: mean 35%, SD 14%; P<.001), namely in sitting and standing, and in high-intensity movements during non-AR (AR: mean 21%, SD 12% vs non-AR: mean 32%, SD 18%; P<.001). During low-structured conditions, engagement time was significantly longer with the AR animals compared to the toy animals (AR: mean 263.1, SD 65.7 seconds vs non-AR: mean 197.3, SD 76.5 seconds; P=.002). Conclusions: While the intensity of physical activity was lower during AR, the greater enjoyment and longer engagement time may lead to greater overall accumulation of active play by motivating young children to go to and stay longer at playgrounds. The high-structured AR conditions resulted in higher-intensity physical activity compared to low-structured AR conditions; however, enjoyment ratings from children and caregivers were generally higher in the low-structured AR. Therefore, AR may be suitable to implement in both low- and high-structured play environments. Future research should investigate whether these findings hold true at outdoor playgrounds and examine the impact of novelty over time.