Search Results (69)

Background/Objectives: Accurate dietary assessment is crucial for nutritional epidemiology, but tools like 24 h recalls (24HRs) face challenges with missing or implausible data. The Automated Self-Administered 24 h Dietary Assessment Tool (ASA24) facilitates large-scale data collection, but its lack of interviewer input may lead to implausible dietary recalls (IDRs), affecting data integrity. Multiple imputation (MI) is commonly used to handle missing data, but its effectiveness in high-variability dietary data is uncertain. This study aims to assess MI’s accuracy in estimating nutrient intake under varying levels of missing data. Methods: Data from 24HRs completed by 743 adolescents (ages 13–18) in Ontario, Canada, were used. Implausible recalls were excluded based on nutrient thresholds, creating a cleaned reference dataset. Missing data were simulated at 10%, 20%, and 40% deletion rates. MI via chained equations was applied, incorporating demographic and psychosocial variables as predictors. Imputed values were compared to actual values using Spearman’s correlation and accuracy within ±10% of true values. Results: Spearman’s rho values between the imputed and actual nutrient intakes were weak (mean ρ ≈ 0.24). Accuracy within ±10% was low for most nutrients (typically < 25%), with no clear trend by missingness level. Diet quality scores showed slightly higher accuracy, but values were still under 30%. Conclusions: MI performed poorly in estimating individual nutrient intake in this adolescent sample. While MI may preserve sample characteristics, it is unreliable for accurate nutrient estimates and should be used cautiously. Future studies should focus on improving data quality and exploring better imputation methods. Full article

Food pattern recognition plays a crucial role in modern healthcare by enabling automated dietary monitoring and personalised nutritional interventions, particularly for vulnerable populations with complex dietary needs. Current food recognition systems struggle to balance high accuracy with interpretability and computational efficiency when analysing complex meal compositions in real-world settings. We developed a novel self-explaining neural architecture that integrates specialised attention mechanisms with temporal modules within a streamlined framework. Our methodology employs hierarchical feature extraction through successive convolution operations, multi-head attention mechanisms for pattern classification, and bidirectional LSTM networks for temporal analysis. Architecture incorporates self-explaining components utilising attention-based mechanisms and interpretable concept encoders to maintain transparency. We evaluated our model on the FOOD101 dataset using 5-fold cross-validation, ablation studies, and comprehensive computational efficiency assessments. Training employed multi-objective optimisation with adaptive learning rates and specialised loss functions designed for dietary pattern recognition. Experiments demonstrate our model’s superior performance, achieving 94.1% accuracy with only 29.3 ms inference latency and 3.8 GB memory usage, representing a 63.3% parameter reduction compared to baseline transformers. The system maintains detection rates above 84% in complex multi-item recognition scenarios, whilst feature attribution analysis achieved scores of 0.89 for primary components. Cross-validation confirmed consistent performance with accuracy ranging from 92.8% to 93.5% across all folds. This research advances automated dietary analysis by providing an efficient, interpretable solution for food recognition with direct applications in nutritional monitoring and personalised healthcare, particularly benefiting vulnerable populations who require transparent and trustworthy dietary guidance. Full article

Background/Objectives: Deep learning (DL) has shown strong potential in analyzing food images, but few studies have directly predicted mass, energy, and macronutrient content from images. In addition to the importance of high-quality data, differences in country-specific food composition databases (FCDBs) can hinder model generalization. Methods: We assessed the performance of several standard DL models using four ground truth datasets derived from Nutrition5k—the largest image–nutrition dataset with ~5000 complex US cafeteria dishes. In light of developing an Italian dietary assessment tool, these datasets varied by FCDB alignment (Italian vs. US) and data curation (ingredient–mass correction and frame filtering on the test set). We evaluated combinations of four feature extractors [ResNet-50 (R50), ResNet-101 (R101), InceptionV3 (IncV3), and Vision Transformer-B-16 (ViT-B-16)] with two regression networks (2+1 and 2+2), using IncV3_2+2 as the benchmark. Descriptive statistics (percentages of agreement, unweighted Cohen’s kappa, and Bland–Altman plots) and standard regression metrics were used to compare predicted and ground truth nutritional composition. Dishes mispredicted by ≥7 algorithms were analyzed separately. Results: R50, R101, and ViT-B-16 consistently outperformed the benchmark across all datasets. Specifically, when replacing it with these top algorithms, reductions in median Mean Absolute Percentage Errors were 6.2% for mass, 6.4% for energy, 12.3% for fat, and 33.1% and 40.2% for protein and carbohydrates. Ingredient–mass correction substantially improved prediction metrics (6–42% when considering the top algorithms), while frame filtering had a more limited effect (<3%). Performance was consistently poor across most models for complex salads, chicken-based or eggs-based dishes, and Western-inspired breakfasts. Conclusions: The R101 and ViT-B-16 architectures will be prioritized in future analyses, where ingredient–mass correction and automated frame filtering methods will be considered. Full article

Trimethylamine N-oxide (TMAO), a gut microbiome metabolite linked to cardiovascular health, can be influenced by dietary factors like choline intake and diet quality. This study compared the effects of mixed tree nuts (MTNs) and pretzels, as part of a 12-week hypocaloric weight loss diet, on TMAO levels and identified dietary predictors. Methods: Plasma samples from 95 overweight individuals consuming either 1.5 oz. of mixed tree nuts (MTNs, n = 56) or isocaloric pretzels (n = 39) daily for 12 weeks were analyzed. Nutritional data were collected at baseline and week 12 through dietary recall using the Automated Self-Administered 24 h Dietary Assessment Tool (ASA24), and the overall diet quality was assessed via the Healthy Eating Index (HEI) score. TMAO levels were determined and analyzed using linear mixed-effect models, adjusting for covariates. Wilcoxon signed-rank tests compared baseline and week 12 TMAO and weight. Multiple linear regression identified baseline predictors of TMAO. Results: Baseline demographics, anthropometric measures, HEI scores, and dietary choline intake were similar between the MTN and pretzel groups. A significant positive association was observed between baseline dietary choline and plasma TMAO levels (p = 0.012). The 12-week hypocaloric diet led to significant weight reduction in both groups (p < 0.01), but the magnitude of weight loss did not differ significantly between the MTN (−3.47 lbs) and pretzel (−4.25 lbs) groups (p = 0.18). Plasma TMAO levels decreased significantly in both groups (p < 0.01), but the between-group difference in reduction was not significant. (MTNs: −0.34 vs. pretzels: −0.37; p = 0.43). HEI scores and dietary choline intake remained unchanged, with no significant time–intervention interaction. Participants with low baseline HEI scores (≤53.72) had a more pronounced reduction in TMAO levels in the MTN group compared to the pretzel group (MTN: −0.54 vs. pretzel: −0.23; p = 0.045) over 12 weeks, despite similar weight loss. This difference was not observed in participants with higher HEI scores. Conclusions: The 12-week hypocaloric diet reduced body weight and plasma TMAO levels similarly in both MTN and pretzel groups. Participants with lower dietary quality saw a greater reduction in TMAO levels in the MTN group, suggesting MTNs may better modulate TMAO levels, especially for those with poorer baseline diets. Full article

Objective: The aim of this study is to determine if protein intake, diet quality, or engagement in physical activity mediate the relationship between sleep quality or duration and handgrip strength. Methods: The sample consisted of 2171 middle-aged persons examined in the 2013–2017 Healthy Aging in Neighborhoods of Diversity across the Life Span (HANDLS) prospective cohort study. Those with sleep apnea (n = 222) and missing data were excluded, resulting in an analytical sample of 1308. Handgrip strength, an objectively measured variable, was determined using a Jamar Hydraulic Hand Dynamometer and expressed relative to body mass index (based on measured height and weight). Sleep quality and duration were measured using the Pittsburgh Sleep Quality Index questionnaire. Protein intake was calculated from two 24 h recalls collected using the USDA Automated Multiple-Pass Method and expressed as g per kg of body weight. Diet quality was assessed using the Healthy Eating Index (HEI) and the energy-adjusted Dietary Inflammatory Index (e-DII). Physical activity was self-reported and expressed as meeting the Life Simple 7 criterion (≥150 min/week, 0–149 min/week, 0 min/week). Mediation analysis was conducted using the Hayes PROCESS macro, model #4, for SPSS Version 4.2. Adjustment for the self-reported covariates of age (years); sex at birth (male, female); race (African American, White); poverty status (<125% or >125% US HHS Poverty Guidelines); current cigarette smoker (yes, no); marijuana, opiate, and/or cocaine user (yes, no); medical conditions including diabetes, hypertension, and/or metabolic syndrome (yes, no); and mean energy (kcal, only protein model) was performed. Results: Protein intake, expressed as g per kg of body weight, mediated the relationship between sleep quality and sleep duration and handgrip strength (indirect effect = −0.0017 ± 0.0006, CI 95% (−0.0030, −0.0006, p < 0.05); indirect effect = 0.0057 ± 0.0019, CI 95% (0.0023, 0.0098, p < 0.05, respectively)). Diet quality, as measured using the HEI, mediated the relationship between sleep duration and handgrip strength (indirect effect = 0.0013 ± 0.0007, CI 95% (0.0001, 0.0030, p < 0.05). Conclusions: Protein intake and a healthy diet mediate the relationship between sleep and handgrip strength, suggesting that these factors may play a role in preserving muscle strength. Full article

Background: Dietary intake plays a crucial role in health research, yet existing methods for its measurement can lead to participant burden, lengthy recording, and human errors, and do not account for age-specific variations. Libro is a real-time diet-tracking mobile-based app offering flexible features. An automated food recording program within Libro was customized for young people vulnerable to eating misbehaviour. This preliminary study assessed its relative validity using a self-administered 24 h recall method as the reference method. Methods: The relative validity of Libro was tested by adopting a cross-over design that recorded food intake over a period of 3 non-consecutive weekdays and 1 weekend day with both methods. The participants were recruited online through a mental health research charity, and this study was conducted fully online. The primary outcome was the concordance of total energy intake between the two methods, with secondary outcomes focusing on the intake of protein, carbohydrates, fats, free sugars, fibre, and trans-fatty acids. Test–retest validity was assessed per method with the intraclass correlation coefficient; a Bland–Altman plot and t-test were performed to test agreement at the group level; correlation coefficient and cross-classification were performed to assess agreement at the individual level. Results: Forty-seven participants were included in the final analysis. The average intraclass correlation coefficient for energy intake measured by Libro over four days was 0.85 (95% CI: 0.76–0.91). Compared to Intake24, the average energy intake recorded using Libro was significantly lower (mean difference: −554 Kcal, 95% CI: −804.1 to −305.6 Kcal, p < 0.001), potentially driven by the reduced reporting of foods rich in free sugars. The correlation coefficient for average energy intake measured by Libro vs. Intake24 was 0.32 (95% CI: 0.03, 0.55), with only 27.7% of subjects classified in the same quartile with both methods (κ = 0.31, 95% CI: −0.03, 0.55). Concordance varied across specific dietary component measures. Conclusions: While Libro had good test–retest reliability if adopting a multiple administration method, it underreported energy and other aspects of dietary intake, along with poor classification performance compared to Intake24 in a population vulnerable to eating misbehaviour. We suggest that future studies improve user experience to increase compliance and data accuracy. Full article

Accurate dietary assessment is essential for effective health management and disease prevention. However, conventional methods that rely on manual food logging and nutritional lookup are often time consuming and error prone. This study proposes an image-based nutritional advisory system that integrates multimodal deep learning to automate food classification, volume estimation, and dietary recommendation to address these limitations. The system employs a fine-tuned CLIP model for zero-shot food recognition, achieving high accuracy across diverse food categories, including unseen items. For volume measurement, a learning-based multi-view stereo (MVS) approach eliminates the need for specialized hardware, yielding reliable estimations with a mean absolute percentage error (MAPE) of 23.5% across standard food categories. Nutritional values are then calculated by referencing verified food composition databases. Furthermore, the system leverages a large language model (Llama 3) to generate personalized dietary advice tailored to individual health goals. The experimental results show that the system attains a top 1 classification accuracy of 91% on CNFOOD-241 and 80% on Food 101 and delivers high-quality recommendation texts with a BLEU-4 score of 45.13. These findings demonstrate the system’s potential as a practical and scalable tool for automated dietary management, offering improved precision, convenience, and user experience. Full article

Background: Culinary medicine (CM) interventions in teaching kitchens have emerged as novel approaches for influencing dietary behaviors, but their efficacy, content, and delivery vary. Objective: The effects of a virtual vegan CM intervention on behavioral determinants, cooking competencies, diet quality, and skin carotenoid status were assessed. Methods: This analysis from a 9-week randomized crossover study evaluated behavioral survey assessments, Whole Plant Food Density (WPFD) as a diet quality indicator utilizing Automated Self-Administered 24 h Dietary Recall data, and skin carotenoid status (SCS) via pressure-mediated reflection spectroscopy at multiple timepoints. Adults at ≥5% atherosclerotic cardiovascular disease (ASCVD) risk followed a vegan diet pattern that was high or low in extra virgin olive oil (EVOO) for 4 weeks each with weekly virtual cooking classes, separated by a 1-week washout period. Qualitative feedback was collected for thematic analysis. Results: In 40 participants (75% female; body mass index, 32 ± 7 kg/m²; age, 64 ± 9 years mean ± SD), perceived control over trajectory of heart disease, knowledge of lifestyle behaviors for heart health, and confidence in cooking skills and preparing a variety of plant-based foods improved post intervention (all p ≤ 0.001). WPFD increased by 69–118% from baseline. Greater SCS changes occurred after high-EVOO (+51.4 ± 13.9 mean ± SEM, p < 0.001) compared to low-EVOO (+6.0 ± 16.4, p = 0.718) diets. Conclusions: A virtual vegan CM intervention improved dietary behaviors and quality, which was associated with reductions in CVD risk factors. SCS is influenced by EVOO intake, warranting consideration when used to estimate fruit and vegetable intake. The potential impacts of CM on behaviors and health outcomes warrant continued research efforts in medical and public health settings. Full article

Normal weight obesity (NWO) is a condition characterized by a normal body mass index (BMI; 18.5–24.9 kg·m⁻²) yet excess body fat. Those with this condition have an increased risk of cardiometabolic diseases associated with obesity. The prevalence of NWO is not well investigated in adolescents, particularly in the United States. This study examined the prevalence of NWO and dietary behaviors among adolescents aged 14–19 years old (n = 139) who live in a rural area in the United States. Data were collected from December 2019 through February 2020. Body composition was assessed via bioelectrical impedance analysis and diet was assessed using an Automated Self-Administered 24 h food recall questionnaire. Participants were categorized by BMI and body fat percentage as NWO, normal weight lean (NWL), or obese (OB). The sample prevalence of NWO was 13.6%, with girls having a higher prevalence (22.2%) than boys (1.8%). Those with NWO had significantly lower dry lean mass than OB (p = 0.02), but there were no differences between NWL and OB (p = 0.08). There was significantly higher caloric intake (p = 0.02) among NWL compared to OB, and NWL consumed more fiber than both NWO (p = 0.02) and OB (p = 0.03). Overall, this study gives us a better understanding of the prevalence of NWO in the adolescent population and the dietary habits associated with each group. Those with NWO may be at increased risk for negative long-term health outcomes commonly associated with obesity. Additionally, the higher caloric intake among NWL was unexpected and should be investigated further. Full article

Background/Objectives: Persons with multiple sclerosis (MS) are interested in diet as a second-line approach for disease management. This study examined potential variables that correlate with inaccuracy of self-reported energy intake (EI) in adults with MS. Methods: Twenty-eight participants completed two assessment appointments within a 14-day period that included a standard doubly labeled water (DLW) protocol for estimating total energy expenditure (TEE). The participants reported their EI using the Automated Self-Administered 24 h (ASA24) Dietary Assessment Tool. The primary variables of interest for explaining the discrepancy between TEE and ASA24 EI (i.e., inaccuracy) included cognition (processing speed, visuospatial memory, and verbal memory), hydration status (total body water), and device-measured physical activity. Pearson’s correlations assessed the association between absolute and percent inaccuracy in reporting of EI with outcomes of interest, followed by linear regression analyses for identifying independent correlates. Results: California Verbal Learning Test—Second Edition (CVLT-II) z-scores and light physical activity (LPA) were significantly associated with mean absolute difference in EI (r = –0.53 and r = 0.46, respectively). CVLT-II z-scores and LPA were the only variables significantly associated with mean percent difference in EI (r = –0.48 and r = 0.42, respectively). The regression analyses indicated that both CVLT-II and LPA significantly explained variance in mean absolute difference in EI, and only CVLT-II explained variance for percent difference in EI. Conclusions: The results from this study indicate that verbal learning and memory and LPA are associated with inaccuracy of self-reported EI in adults with MS. This may guide timely research identifying appropriate protocols for assessment of diet in MS. Full article

Background: Healthy dietary patterns can support the maintenance of cognition and brain health in older age and are negatively associated with cardiometabolic risk. Cardiometabolic risk factors are similarly important for cognition and may play an important role in linking diet to cognition. Aim: This study aimed to explore the relationship between dietary patterns and cognition and to determine whether cardiometabolic health markers moderate these relationships in older adulthood. Design: A cross-sectional analysis of observational data from the baseline of the ACTIVate study. Participants: The cohort included 426 cognitively normal adults aged 60–70 years. Methods: The Australian Eating Survey (AES) Food Frequency Questionnaire was used to collect data on usual dietary intake, along with additional questions assessing intake of dietary oils. Principal component analysis (PCA) was applied to reduce the dimensionality of dietary data. Cardiometabolic risk was quantified using the metabolic syndrome severity score (MetSSS). Tests from the Cambridge Neuropsychological Test Automated Battery (CANTAB) were used to derive composite scores on four cognitive domains: processing speed, executive function, short-term memory, and long-term memory. Results: Three dietary patterns were identified using PCA: a plant-dominant diet, a Western-style diet, and a meat-dominant diet. After controlling for age, sex, total years of education, energy intake, and moderate-to-vigorous physical activity (MVPA), there was a small, negative association between the meat-dominant diets and long-term memory. Subsequent moderation analysis indicated that MetSSS significantly moderated this relationship. Conclusions: Findings highlight the link between diet, cardiometabolic health, and cognitive function in older, cognitively healthy adults. However, longitudinal studies are needed to confirm observations and evaluate the dynamics of diet, cardiometabolic health, and cognitive function over time. Full article

Tools to briefly assess diet among US Spanish-speaking adults are needed to identify individuals at risk for cardiometabolic disease (CMD) related to diet. Two registered dietitian nutritionists (RDNs) recruited bilingual medical students to translate the validated Diet Risk Score (DRS) into Spanish (DRS-S). Participants were recruited from a federally qualified health center. Students administered the DRS-S and one 24-h recall (Automated Self-Administered 24-Hour (ASA24^®) Dietary Assessment Tool) on one day; a second recall was administered within 1 week. Recalls were scored using the Healthy Eating Index (HEI)-2015, a measure of adherence to the Dietary Guidelines for Americans. Spearman correlations, weighted kappa, and ANOVA were conducted using SAS 9.4 to assess the relative validity of the DRS-S. Thirty-one Spanish-speaking adults (female: n = 17, 53%; mean age: 58 (42–69)) completed assessments. The mean DRS-S was 9 (SD = 4.2) (max: 27; higher score = higher risk) and the mean HEI-2015 score was 65.7 (SD = 9.7) (max: 100; higher score = lower risk), with significant agreement between measures (r: −0.45 (p = 0.01)), weighted kappa: −0.3 (p = 0.03). The DRS-S can be used in resource-constrained settings to assess diet for intervention and referral to RDNs. The DRS-S should be tested in clinical care to assess the impact of dietary changes to reduce CMD risk. Full article

In industry 4.0, where the automation and digitalization of entities and processes are fundamental, artificial intelligence (AI) is increasingly becoming a pivotal tool offering innovative solutions in various domains. In this context, nutrition, a critical aspect of public health, is no exception to the fields influenced by the integration of AI technology. This study aims to comprehensively investigate the current landscape of AI in nutrition, providing a deep understanding of the potential of AI, machine learning (ML), and deep learning (DL) in nutrition sciences and highlighting eventual challenges and futuristic directions. A hybrid approach from the systematic literature review (SLR) guidelines and the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines was adopted to systematically analyze the scientific literature from a search of major databases on artificial intelligence in nutrition sciences. A rigorous study selection was conducted using the most appropriate eligibility criteria, followed by a methodological quality assessment ensuring the robustness of the included studies. This review identifies several AI applications in nutrition, spanning smart and personalized nutrition, dietary assessment, food recognition and tracking, predictive modeling for disease prevention, and disease diagnosis and monitoring. The selected studies demonstrated the versatility of machine learning and deep learning techniques in handling complex relationships within nutritional datasets. This study provides a comprehensive overview of the current state of AI applications in nutrition sciences and identifies challenges and opportunities. With the rapid advancement in AI, its integration into nutrition holds significant promise to enhance individual nutritional outcomes and optimize dietary recommendations. Researchers, policymakers, and healthcare professionals can utilize this research to design future projects and support evidence-based decision-making in AI for nutrition and dietary guidance. Full article

Background and Objectives: Biological and mechanistic data can support observational evidence to aid causal inference. The substantial body of available epidemiological evidence on the role of diet, nutrition, physical activity, and body weight and cancer has been systematically reviewed within the World Cancer Research Fund Global Cancer Update Program (WCRF CUP Global) over the past few decades. Mechanistic data can provide substantial additional support to established or suspected associations between diet and cancer but has not previously been systematically reviewed within the CUP Global. Here, we describe the development of a framework for the evaluation of biological and mechanistic data to support CUP Global in their evaluations. Methods: The protocol to evaluate mechanistic data utilizes a two-stage, iterative approach: (1) use of expert knowledge in combination with text mining automated tools (https://www.temmpo.org.uk/ accessed on 14 February 2024 to identify a set of the main potential mechanisms (typically 2–3 mechanisms) and their associated intermediate phenotypes (IPs) that link the factor of interest (exposure: E) to the cancer outcome of interest (outcome: O) and (2) for selected mechanisms, perform systematic literature reviews of human studies to evaluate the associations between E and IPs and between IPs and O. An expert committee then assesses the level of evidence for the role of each potential mechanism in the E–O association. If appropriate, additional literature reviews of experimental studies will be performed to address specific questions. Results: A protocol has been developed that can be used to systematically review data on mechanisms in a timely manner. As a first test case, the proposed protocol will be tested to evaluate mechanisms linking dietary patterns and colorectal cancer development. Discussion: This project will produce a framework for the systematic evaluation of mechanistic research to support causal associations between diet, nutrition, physical activity, body weight and cancer risk within WCRF CUP Global. Full article

Photo-based dietary assessment is becoming more feasible as artificial intelligence methods improve. However, advancement of these methods for dietary assessment in research settings has been hindered by the lack of an appropriate dataset against which to benchmark algorithm performance. We conducted the Surveying Nutrient Assessment with Photographs of Meals (SNAPMe) study (ClinicalTrials ID: NCT05008653) to pair meal photographs with traditional food records. Participants were recruited nationally, and 110 enrollment meetings were completed via web-based video conferencing. Participants uploaded and annotated their meal photos using a mobile phone app called Bitesnap and completed food records using the Automated Self-Administered 24-h Dietary Assessment Tool (ASA24^®) version 2020. Participants included photos before and after eating non-packaged and multi-serving packaged meals, as well as photos of the front and ingredient labels for single-serving packaged foods. The SNAPMe Database (DB) contains 3311 unique food photos linked with 275 ASA24 food records from 95 participants who photographed all foods consumed and recorded food records in parallel for up to 3 study days each. The use of the SNAPMe DB to evaluate ingredient prediction demonstrated that the publicly available algorithms FB Inverse Cooking and Im2Recipe performed poorly, especially for single-ingredient foods and beverages. Correlations between nutrient estimates common to the Bitesnap and ASA24 dietary assessment tools indicated a range in predictive capacity across nutrients (cholesterol, adjusted R² = 0.85, p < 0.0001; food folate, adjusted R² = 0.21, p < 0.05). SNAPMe DB is a publicly available benchmark for photo-based dietary assessment in nutrition research. Its demonstrated utility suggested areas of needed improvement, especially the prediction of single-ingredient foods and beverages. Full article