AI-based digital image dietary assessment methods compared to humans and ground truth: a systematic review.

OBJECTIVE: Human error estimating food intake is a major source of bias in nutrition research. Artificial intelligence (AI) methods may reduce bias, but the overall accuracy of AI estimates is unknown. This study was a systematic review of peer-reviewed journal articles comparing fully automated AI-based (e.g. deep learning) methods of dietary assessment from digital images to human assessors and ground truth (e.g. doubly labelled water). MATERIALS AND METHODS: Literature was searched through May 2023 in four electronic databases plus reference mining. Eligible articles reported AI estimated volume, energy, or nutrients. Independent investigators screened articles and extracted data. Potential sources of bias were documented in absence of an applicable risk of bias assessment tool. RESULTS: Database and hand searches identified 14,059 unique publications; fifty-two papers (studies) published from 2010 to 2023 were retained. For food detection and classification, 79% of papers used a convolutional neural network. Common ground truth sources were calculation using nutrient tables (51%) and weighed food (27%). Included papers varied widely in food image databases and results reported, so meta-analytic synthesis could not be conducted. Relative errors were extracted or calculated from 69% of papers. Average overall relative errors (AI vs. ground truth) ranged from 0.10% to 38.3% for calories and 0.09% to 33% for volume, suggesting similar performance. Ranges of relative error were lower when images had single/simple foods. CONCLUSIONS: Relative errors for volume and calorie estimations suggest that AI methods align with - and have the potential to exceed - accuracy of human estimations. However, variability in food image databases and results reported prevented meta-analytic synthesis. The field can advance by testing AI architectures on a limited number of large-scale food image and nutrition databases that the field determines to be adequate for training and testing and by reporting accuracy of at least absolute and relative error for volume or calorie estimations.

These results suggest that AI methods are in line with ­ and have the potential to exceed ­ accuracy of human estimations of nutrient content based on digital food images.Variability in food image databases used and results reported prevented meta-analytic synthesis.The field can advance by testing AI architectures on a limited number of large-scale food image and nutrition databases that the field determines to be accurate and by reporting accuracy of at least absolute and relative error for volume or calorie estimations.Overall, the tools currently available need more development before deployment as stand-alone dietary assessment methods in nutrition research or clinical practice.

Reliability and Validity of Digital Images to Assess Child Dietary Intake in a Quick-Service Restaurant Setting.

BACKGROUND: Development of methods to accurately measure dietary intake in free-living situations-restaurants or otherwise-is critically needed to understand overall dietary patterns. OBJECTIVE: This study aimed to develop and test reliability and validity of digital images (DI) for measuring children's dietary intake in quick-service restaurants (QSRs), validating against weighed plate waste (PW) and bomb calorimetry (BC). DESIGN: In 2016, cross-sectional data were collected at two time points within a randomized controlled trial assessing children's leftovers in QSRs from parents of 4- to 12-year-old children. PARTICIPANTS/SETTING: Parents (n = 640; mean age = 35.9 y; 70.8% female) consented and agreed to provide their child's PW for digital imaging, across 11 QSRs in Massachusetts in areas with low socioeconomic status and ethnically diverse populations. OUTCOME MEASURES: Outcome measures were interrater reliability for DIs, correspondence between methods for energy consumed and left over, and correspondence between methods across varying quantities of PW. ANALYSES PERFORMED: Intraclass correlations, percent agreement, Spearman correlations, Wilcoxon signed rank tests, and Bland-Altman plots were used. RESULTS: Interrater reliability ratings for DIs had substantial intraclass correlations (ICC = 0.94) but not acceptable exact percent agreement (80.2%); DI and PW energy consumed were significantly correlated (r = 0.96, P < 0.001); DI slightly underestimated energy consumed compared with PW (Mdiff = -1.61 kcals, P < 0.001). Bland-Altman plots showed high DI-PW correspondence across various energy amounts and revealed few outliers. Energy left over by BC was highly correlated with DI (r = 0.87, P < 0.001) and PW (r = 0.90, P < 0.001); and mean differences were not significantly different from DI (Mdiff = 9.77 kcal, P = 0.06) or PW (Mdiff = -2.84 kcal, P = 0.20). CONCLUSIONS: Correspondence was high between PW and DI assessments of energy consumed, and high with BC energy left over. Results demonstrate reliability and practical validity of digital images for assessing child meal consumption in QSR settings.

The impact of a community social marketing campaign on children's meal orders and consumption: main outcomes from a group randomised controlled trial.

OBJECTIVE: Restaurants may be important settings for interventions to reduce children's energy intake. The objective of this study was to test the impact of a parent-focused social marketing campaign to promote healthy children's meals on calories ordered and consumed by children at quick-service restaurants (QSR). DESIGN: Using a repeated cross-sectional study design, two urban communities were randomised to intervention (IN) v. control (C) condition. A community-wide social marketing campaign was implemented in the IN community to empower Black and Latinx mothers who frequent QSR (priority population) to select healthier options for their child. SETTING: Data were collected in 2016 at QSR located within the communities pre- and post-IN and analysed in 2017. PARTICIPANTS: Parents (n 1686; n 819 and n 867 for I and C conditions, respectively) were recruited after placing their QSR order; a survey, receipt and their child's leftovers were collected. RESULTS: Calories ordered did not differ significantly between the IN and C conditions (changeadj = -146·4 kJ (-35·0 kcal); 95 % CI -428·0 kJ (-102·3 kcal), 134·6 kJ (32·2 kcal)). In a sub-analysis of only the priority audience, children in the IN community ordered significantly fewer calories compared to C children in unadjusted models (changeunadj = -510·4 kJ (-122·0 kcal); 95 % CI -1013·4 kJ (-242·2 kcal), -7·5 kJ (-1·8 kcal)), but the trend did not persist after adjusting for covariates (changeadj = -437·2 kJ (-104·5 kcal); 95 % CI -925·5 kJ (-221·2 kcal), 50·6 kJ (12·1 kcal)). Calories consumed followed similar trends. CONCLUSION: The campaign did not significantly reduce children's QSR calories ordered or consumed. However, a quantitatively important mean reduction in calories was suggested among the priority audience, indicating potential for community-wide promotion of healthful children's meals.

A Mixed-Methods Evaluation of a School Wellness Initiative: An Examination of Longer Lunch Periods and More Physical Activity Opportunities.

BACKGROUND: There are currently no national standards for lunch period lengths or physical activity in schools. Research is needed to better understand the impact of school initiatives that improve policies related to lunch and movement opportunities on student outcomes. Additionally, best practices are necessary to support schools that are considering initiatives that address these factors. OBJECTIVE: This study examined the impact of implementing longer lunch periods, recess, and other movement opportunities on student outcomes and best practices for implementation. DESIGN: A mixed-methods study including surveys and semistructured interviews and focus groups conducted during the 2019-2020 school year. PARTICIPANTS/SETTING: Surveys (n = 5107) from students in grades 3 and 4 attending 19 pilot and 11 matched control elementary schools and interviews/focus groups among principals, cafeteria managers, teachers, and parents in a representative subsample (n = 6) of pilot schools in Anchorage Alaska. MAIN OUTCOME MEASURES: Students' self-reported hunger levels and mood and perceptions and supportive strategies from school principals, cafeteria staff, teachers, and parents were examined. STATISTICAL ANALYSES PERFORMED: Mixed-model analysis of variance accounting for student demographics with students as a random effect (students nested within schools) were used to examine differences in hunger and mood. For interviews/focus groups, responses were analyzed qualitatively using principles of content analysis. RESULTS: Longer lunch periods were associated with significantly reduced hunger at the end of lunch period and significantly increased self-reported happiness in the cafeteria. Based on interviews/focus groups with school staff and parents, the initiative was generally perceived positively with reported benefits including reductions in disciplinary issues and improvements in student focus, social and emotional learning, and overall student happiness and well-being. Several supportive strategies were identified. CONCLUSIONS: Initiatives that increase lunch period lengths and physical activity opportunities have the potential to reduce students' hunger levels and improve focus and behaviors in the classroom. Schools should consider similar initiatives that incorporate the suggested strategies to potentially improve outcomes among students.

A positive deviance-based qualitative study of stress, coping, and feeding practices among low-income, Hispanic mothers whose children do versus do not meet guidelines for fruit and vegetable intake.

Less than 1% of children in the United States concurrently meet guidelines for fruit/vegetable intake, physical activity, screen time, and sugar-sweetened beverages. Prior evidence suggests that parents of this 1% potentially cope with stress differently. This qualitative study used a positive deviance-based approach to locate mothers whose children avoided negative feeding outcomes despite being 'high-risk' for obesity. Semi-structured interviews were conducted in Spanish for two groups: low-income, Hispanic mothers whose children were normal weight and met recommendations for fruits/vegetables and physical activity (n = 5); and a comparison group whose children had obesity and did not meet guidelines (n = 8). Topics included weight-related parenting practices, attitudes toward health, and stress management. Interviews were transcribed, translated, and coded using NVivo for theoretically driven thematic analysis. Results suggested that mothers viewed stress differently. Mothers of healthy weight children believed stress could be prevented, such as by paying children more attention or directing one's attention away from stressors; comparison group mothers tended to report stress about managing their child's eating and about financial worries. Future research is needed to understand the underlying sources of these differences (e.g. personality traits, coping practices) and test whether stress prevention interventions can promote healthy parental feeding practices.

Orders of Healthier Adult Menu Items in a Full-Service Restaurant Chain with a Healthier Children's Menu.

This study evaluated orders of adult menu items designated as healthier at the Silver Diner, a regional full-service restaurant chain serving over 4 million customers annually. This restaurant implemented a healthier children's menu in April 2012. Orders of adult menu items were abstracted from before (September 2011-March 2012; PRE; n = 1,801,647) and after (September 2012-March 2013; POST; n = 1,793,582) the healthier children's menu was introduced. Entrées, appetizers, and sides listed as healthier options on the menu were coded as healthier. PRE to POST changes in the percentage of orders of healthier items, soda, and dessert were evaluated using McNemar tests of paired proportions. Orders of healthier entrées, appetizers, and sides on the adult menu increased PRE to POST (8.9% to 10.4%, 25.5% to 27.5%, and 7.3% to 9.3%, respectively), and soda and dessert orders decreased (23.2% to 21.7% and 29.0% to 28.3%, respectively). All shifts were statistically significant (p < 0.0001). Our findings demonstrate improvements in orders of healthier adult menu options during the same time frame as a healthy children's menu change. Future research can help elucidate mechanisms to inform future health promotion efforts in restaurants in ways that have the potential to impact both adults and children.

Multiple Health Behavior Change for Weight Loss: A Scoping Review.

Objectives: In this scoping review, we examine the current state of literature on weight-related Multiple Health Behavior Change (MHBC). Specifically, we investigate: (1) MHBC versus single health behavior change (SHBC) interventions and (2) simultaneous versus sequential MHBC approaches. Secondarily, we explore (3) attributes that predict success in MHBC, and (4) the utilization of theoretical frameworks. Methods: We retrieved studies from PubMed, Web of Science, and Google Scholar within the 2000 to 2018 range. Results: MHBC interventions proved superior for long-term weight loss when compared to SHBC approaches. However, the literature is limited. Studies investigating simultaneous and sequential MHBC approaches are also limited and have mixed results. Predictive characteristics of MHBC include behavior adherence, risk level, stage of change, self-efficacy, social support, environmental barriers, and treatment approaches. Whereas evidence evaluating theory in MHBC programs remains scant, there is promising research on constructs from the Transtheoretical Model and Social Cognitive Theory. Conclusions: MHBC approaches may better support weight loss efforts. However, further research is needed to understand the effects of behavior change order and timing, predictive features of participants and interventions, and theoretical framework utilization in these weight-loss programs.

Weighed Plate Waste Can Accurately Measure Children's Energy Consumption from Food in Quick-Service Restaurants.

BACKGROUND: Quick-service restaurants (QSRs) serve one-third of US children on any given day, yet no methods can directly measure energy (kcal) consumed in QSRs. Weighed plate waste is one feasible option, but the accuracy is unknown. OBJECTIVE: The objective of this study was to determine the accuracy of weighed plate waste for measuring children's energy consumption in QSRs. METHODS: Children's plate waste (entrées and sides) was collected for a larger study assessing a community-wide health messaging campaign to inform parents' orders for children in QSRs; a subsample (n = 194) was used for validation. Gross energy left over estimated by weighed plate waste combined with restaurant-stated nutrition information was compared to gross energy determined by bomb calorimetry, the gold-standard energy assessment technique. Analyses were conducted at the meal level (all food items, combined) and stratified by the number of items per meal (1, 2, or 3). Pearson correlations and paired t tests analyzed agreement; Bland-Altman statistics examined differences between energy estimations for the total and stratified subsample. RESULTS: Overall, significant agreement was observed between weighed plate waste and bomb calorimetry (r = 0.99, P < 0.001). On average, weighed plate waste underestimated energy content by <2 kcal compared with bomb calorimetry (mean percent difference ± SD of 0.3% ± 10.7%); 94% of estimations fell within the limits of agreement (-23.5 to 26.8 kcal), and 63% and 24% of estimations differed by <10 or <20 net kcal, respectively. Although stratification by item number showed slight variation, mean differences for all strata were <5 kcal (t test P > 0.80), suggesting the accuracy of weighed plate waste for measuring meals of various sizes. CONCLUSIONS: Weighed plate waste is an accurate and valid field technique for measuring children's energy consumption from food in QSRs. Future improvements to capturing beverages, self-serve condiments, and sharing behaviors may improve the overall feasibility and accuracy.

Using a Marketing Evaluation Tool to Optimize a Social Marketing Campaign: Insights From a Copy Test of a You're the Mom Campaign Ad.

A copy test is a business tool for assessing advertisements. This report provides an example of how copy test may be used within nutrition education practice and research. A public health nutrition advertisement for You're the Mom was copy tested with a market research firm. Mothers (n = 300) were aged 22-49years, had a household income <$50,000 and ≥1 child aged 4-8years and bought fast food ≥2-3times/mo. Compared with advertisements for for-profit goods, the advertisement scored high on impact (77th percentile) and moderate on persuasiveness (46th percentile) and communicated 2 key messages at higher rates than norms (51% and 46%) and a third at a lower rate (37%). Copy test results can best inform campaign development when the distinct purpose of the ad (versus the overall campaign) is clear; and when evaluation tools are designed to align directly with that purpose.

Direct and indirect effects of parent stress on child obesity risk and added sugar intake in a sample of Southern California adolescents.

OBJECTIVE: Research indicates that children are at higher risk for obesity if their parents have been exposed to a larger number of stressors, yet little is known about effects of parents' subjective, perceived experience of stress on children's eating behaviours and adiposity and whether weight-related parenting practices (i.e. parent rules and positive family meal practices) mediate this relationship. The present study evaluated the direct and mediated relationship between parent perceived stress and child waist circumference and parent stress and child consumption of added sugars one year later. DESIGN: Longitudinal panel data. SETTING: Eleven communities in Southern California, USA. SUBJECTS: Data were collected over two waves from parent-child dyads (n 599). Most parents were female (81 %) and Hispanic (51 %); children were 11 years old on average (sd 1·53; range 7-15 years) and 31 % received free school lunch. RESULTS: Perceived parent stress was not significantly associated with child waist circumference or consumption of added sugars one year later, and mediating pathways through parenting practices were not significant. However, parent rules were significantly associated with lower child consumption of added sugars (ß=-0·14, P<0·001). CONCLUSIONS: Results suggest that parent rules about the types of foods children can eat, clearly explained to children, may decrease child consumption of added sugars but not necessarily lead to changes in obesity risk. Parent- and family-based interventions that support development of healthy rules about child eating have the potential to improve child dietary nutrient intake.

Child and parent perspectives on healthier side dishes and beverages in restaurant kids' meals: results from a national survey in the United States.

BACKGROUND: Children frequently consume foods from restaurants; considering the quick-service sector alone, 1/3 of children eat food from these restaurants on a given day, and among these consumers, 1/3 of their daily calories come from fast food. Restaurant foods and beverages are second only to grocery store foods and beverages in their contribution to total energy intake of U.S. 4- to 11-year-olds. Shifting their restaurant consumption in healthier directions could have a positive impact on child health. In 2014 this study examined self-reported child receptivity and parent awareness of child receptivity to ordering a fruit or vegetable side dish instead of French fries; and milk, water, or flavored water instead of soda/pop with a kids' meal when eating out. Child receptivity to side dishes was compared between 2010 and 2014. METHODS: An online survey was administered by Nielsen via their Harris Poll Online to a national panel of 711 parents and their 8- to 12-year-old child, as part of a larger study. Frequencies, logistic regressions, t-tests, chi-square tests, and percent agreement were used to evaluate child likelihood of ordering certain side dishes; receptivity to healthier side dish and beverage alternatives; changes in receptivity to healthier sides across years; and parent awareness. RESULTS: A majority of children said they were likely to order a meal with a vegetable (60%), fruit (78%), or French fry (93%) side dish. They were receptive to receiving a fruit or vegetable (FV) side dish instead of French fries (68%); or milk, water, or flavored water instead of soda (81%) with their restaurant kids' meal. Liking/taste was the most common reason for children's feelings. Child receptivity to a FV side dish instead of French fries was high in both years and significantly higher in 2014 (t = -2.12, p = 0.034). The majority of parent and child reports of child receptivity were concordant (85%). CONCLUSIONS: These national survey results indicate that children are receptive to FV side dishes and healthier beverage options with their restaurant meals. Their receptivity has remained high in the recent past, and parents are aware of child receptivity. An opportunity exists for restaurants to leverage child receptivity to healthier sides and beverages by providing and promoting healthy options.

Associations among dispositional mindfulness, self-compassion, and executive function proficiency in early adolescents.

The study objective was to examine the effects of two conceptually related constructs, self-compassion and dispositional mindfulness, on executive function (EF) proficiency among early adolescents. Executive function refers to a set of psychological processes governing emotional regulation, organization, and planning. While the benefits of positive psychology appear evident for mental health and wellness, little is known about the etiological relationship between dispositional mindfulness and self-compassion in their associations with EF. Two hundred and ten early adolescents attending middle school (age M=12.5 years; SD=0.5; 21% Hispanic, 18% Mixed/bi-racial, 47% White, and 9% Other/Missing; 37.1% on free lunch program) self-reported levels of dispositional mindfulness (Mindful Attention Awareness Scale; MAAS), self-compassion (Self-Compassion Scale; SCS; self-judgment and self-kindness domains), and EF proficiency (Behavior Rating Inventory of Executive Function; BRIEF-SR). A sequential linear regression stepwise approach was taken entering the independent variables as separate models in the following order: self-kindness, self-judgement, and dispositional mindfulness. All models controlled for participant age and sex. SCS self-kindness was not associated with EF proficiency, but SCS self-judgment (reverse-coded) contributed to the variance in EF (ß=0.40, p<.001). When adding MAAS to the model, MAAS scores were significantly associated with EF (ß=0.64, p<.001) and accounted for the initial variance explained by SCS. When considering the influence of positive psychology constructs on EF proficiency in adolescents, the measure of dispositional mindfulness appears to outweigh that of specific self-compassion domains, when independent of contemplative training.

Brief report: Examining children's disruptive behavior in the wake of trauma - A two-piece growth curve model before and after a school shooting.

School shootings may have serious negative impacts on children years after the event. Previous research suggests that children exposed to traumatic events experience heightened fear, anxiety, and feelings of vulnerability, but little research has examined potential aggressive and disruptive behavioral reactions. Utilizing a longitudinal dataset in which a local school shooting occurred during the course of data collection, this study sought to investigate whether the trajectory of disruptive behaviors was affected by the shooting. A two-piece growth curve model was used to examine the trajectory of disruptive behaviors during the pre-shooting years (i.e., piece one) and post-shooting years (i.e., piece two). Results indicated that the two-piece growth curve model fit the data better than the one-piece model and that the school shooting precipitated a faster decline in aggressive behaviors. This study demonstrated a novel approach to examining effects of an unexpected traumatic event on behavioral trajectories using an existing longitudinal data set.

The Acute Relationships Between Affect, Physical Feeling States, and Physical Activity in Daily Life: A Review of Current Evidence.

Until recently, most studies investigating the acute relationships between affective and physical feeling states and physical activity were conducted in controlled laboratory settings, whose results might not translate well to everyday life. This review was among the first attempts to synthesize current evidence on the acute (e.g., within a few hours) relationships between affective and physical feeling states and physical activity from studies conducted in free-living, naturalistic settings in non-clinical populations. A systematic literature search yielded 14 eligible studies for review. Six studies tested the relationship between affective states and subsequent physical activity; findings from these studies suggest that positive affective states were positively associated with physical activity over the next few hours while negative affective states had no significant association. Twelve studies tested affective states after physical activity and yielded consistent evidence for physical activity predicting higher positive affect over the next few hours. Further, there was some evidence that physical activity was followed by a higher level of energetic feelings in the next few hours. The evidence for physical activity reducing negative affect in the next few hours was inconsistent and inconclusive. Future research in this area should consider recruiting more representative study participants, utilizing higher methodological standards for assessment (i.e., electronic devices combined with accelerometry), reporting patterns of missing data, and investigating pertinent moderators and mediators (e.g., social and physical context, intensity, psychological variables). Knowledge gained from this topic could offer valuable insights for promoting daily physical activity adoption and maintenance in non-clinical populations.