Balancing the Encoder and Decoder Complexity in Image Compression for Classification.

This paper presents a study on the computational complexity of coding for machines, with a focus on image coding for classification.We first conduct a comprehensive set of experiments to analyze the size of the encoder (which encodes images to bitstreams), the size of the decoder (which decodes bitstreams and predicts class labels), and their impact on the rate-accuracy trade-off in compression for classification.Through empirical investigation, we demonstrate a complementary relationship between the encoder size and the decoder size, \ie, it is better to employ a large encoder with a small decoder and vice versa.Motivated by this relationship, we introduce a feature compression-based method for efficient image compression for classification.By compressing features at various layers of a neural network-based image classification model, our method achieves adjustable rate, accuracy, and encoder (or decoder) size using a single model.Experimental results on ImageNet classification show that our method achieves competitive results with existing methods while being much more flexible.The code will be made publicly available.

A machine-learning-enabled smart neckband for monitoring dietary intake.

The increasing need for precise dietary monitoring across various health scenarios has led to innovations in wearable sensing technologies. However, continuously tracking food and fluid intake during daily activities can be complex. In this study, we present a machine-learning-powered smart neckband that features wireless connectivity and a comfortable, foldable design. Initially considered beneficial for managing conditions such as diabetes and obesity by facilitating dietary control, the device's utility extends beyond these applications. It has proved to be valuable for sports enthusiasts, individuals focused on diet control, and general health monitoring. Its wireless connectivity, ergonomic design, and advanced classification capabilities offer a promising solution for overcoming the limitations of traditional dietary tracking methods, highlighting its potential in personalized healthcare and wellness strategies.

Accuracy of energy and nutrient intake estimation versus observed intake using four technology-assisted dietary assessment methods: a randomized crossover feeding study.

BACKGROUND: Technology-assisted 24-hour dietary recalls (24HR) have been widely adopted in population nutrition surveillance. Evaluations of 24HR inform improvements but direct comparisons of 24HR methods for accuracy in reference to a measure of true intake are rarely undertaken in a single study population. OBJECTIVE: To compare the accuracy of energy and nutrient intake estimation of four technology-assisted dietary assessment methods relative to true intake across breakfast, lunch, and dinner. METHODS: In a controlled feeding study with a crossover design, 152 participants (55% women; mean age 32y (SD 11); mean BMI 26 kg/m2 (SD 5)) were randomized to one of three separate feeding days to consume breakfast, lunch, and dinner, with unobtrusive weighing of foods and beverages consumed. Participants undertook a 24HR the following day (Automated Self-Administered Dietary Assessment Tool-Australia© (ASA24); Intake24-Australia©; mobile Food Record™ - Trained Analyst (mFR-TA); or Image-Assisted Interviewer-Administered 24-hour recall (IA-24HR)). When assigned to IA-24HR, participants referred to images captured of their meals using the mobile Food Record™ app. True and estimated energy and nutrient intakes were compared, and differences among methods were assessed using linear mixed models. RESULTS: The mean difference between true and estimated energy intake as a percentage of true intake was 5.4% (95% CI 0.6, 10.2) using ASA24, 1.7% (95% CI -2.9, 6.3) using Intake24, 1.3% (95% CI -1.1, 3.8) using mFR-TA, and 15.0% (95% CI 11.6, 18.3) using IA-24HR. The variances of estimated and true energy intakes were statistically significantly different for all methods (P<0.01), apart from Intake 24 (P=0.1). Differential accuracy in nutrient estimation was present among the methods. CONCLUSIONS: Under controlled conditions, Intake24, ASA24, and mFR-TA estimated average energy and nutrient intakes with reasonable validity, but intake distributions were estimated accurately by Intake24 only (energy and protein). This study may inform considerations regarding instruments of choice in future population surveillance. TRIAL REGISTRATION: Australian New Zealand Clinical Trials Registry Number ACTRN12621000209897; https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?id=381165&isReview=true.

QARV: Quantization-Aware ResNet VAE for Lossy Image Compression.

This paper addresses the problem of lossy image compression, a fundamental problem in image processing and information theory that is involved in many real-world applications. We start by reviewing the framework of variational autoencoders (VAEs), a powerful class of generative probabilistic models that has a deep connection to lossy compression. Based on VAEs, we develop a new scheme for lossy image compression, which we name quantization-aware ResNet VAE (QARV). Our method incorporates a hierarchical VAE architecture integrated with test-time quantization and quantization-aware training, without which efficient entropy coding would not be possible. In addition, we design the neural network architecture of QARV specifically for fast decoding and propose an adaptive normalization operation for variable-rate compression. Extensive experiments are conducted, and results show that QARV achieves variable-rate compression, high-speed decoding, and better rate-distortion performance than existing baseline methods.

Unified Architecture Adaptation for Compressed Domain Semantic Inference.

Advances in both lossy image compression and semantic content understanding have been greatly fueled by deep learning techniques, yet these two tasks have been developed separately for the past decades. In this work, we address the problem of directly executing semantic inference from quantized latent features in the deep compressed domain without pixel reconstruction. Although different methods have been proposed for this problem setting, they either are restrictive to a specific architecture, or are sub-optimal in terms of compressed domain task accuracy. In contrast, we propose a lightweight, plug-and-play solution which is generally compliant with popular learned image coders and deep vision models, making it attractive to vast applications. Our method adapts prevalent pixel domain neural models that are deployed for various vision tasks to directly accept quantized latent features (other than pixels). We further suggest training the compressed domain model by transferring knowledge from its corresponding pixel domain counterpart. Experiments show that our method is compliant with popular learned image coders and vision task models. Under fair comparison, our approach outperforms a baseline method by a) more than 3% top-1 accuracy for compressed domain classification, and b) more than 7% mIoU for compressed domain semantic segmentation, at various data rates.

Integration of USDA Food Classification System and Food Composition Database for Image-Based Dietary Assessment among Individuals Using Insulin.

New imaging technologies to identify food can reduce the reporting burden of participants but heavily rely on the quality of the food image databases to which they are linked to accurately identify food images. The objective of this study was to develop methods to create a food image database based on the most commonly consumed U.S. foods and those contributing the most to energy. The objective included using a systematic classification structure for foods based on the standardized United States Department of Agriculture (USDA) What We Eat in America (WWEIA) food classification system that can ultimately be used to link food images to a nutrition composition database, the USDA Food and Nutrient Database for Dietary Studies (FNDDS). The food image database was built using images mined from the web that were fitted with bounding boxes, identified, annotated, and then organized according to classifications aligning with USDA WWEIA. The images were classified by food category and subcategory and then assigned a corresponding USDA food code within the USDA's FNDDS in order to systematically organize the food images and facilitate a linkage to nutrient composition. The resulting food image database can be used in food identification and dietary assessment.

Long-Tailed Food Classification.

Food classification serves as the basic step of image-based dietary assessment to predict the types of foods in each input image. However, foods in real-world scenarios are typically long-tail distributed, where a small number of food types are consumed more frequently than others, which causes a severe class imbalance issue and hinders the overall performance. In addition, none of the existing long-tailed classification methods focus on food data, which can be more challenging due to the inter-class similarity and intra-class diversity between food images. In this work, two new benchmark datasets for long-tailed food classification are introduced, including Food101-LT and VFN-LT, where the number of samples in VFN-LT exhibits real-world long-tailed food distribution. Then, a novel two-phase framework is proposed to address the problem of class imbalance by (1) undersampling the head classes to remove redundant samples along with maintaining the learned information through knowledge distillation and (2) oversampling the tail classes by performing visually aware data augmentation. By comparing our method with existing state-of-the-art long-tailed classification methods, we show the effectiveness of the proposed framework, which obtains the best performance on both Food101-LT and VFN-LT datasets. The results demonstrate the potential to apply the proposed method to related real-life applications.

Comparing different environmental enrichments for improving the welfare and walking ability of male turkeys.

This study investigated age-related changes in turkey welfare measures (wounds, feather quality (FQ), feather cleanliness, and footpad condition (FCON)) and walking ability (gait) as influenced by different types of environmental enrichment (EE). Tom turkeys (n = 420) were randomly assigned to: straw bale (S), platform (P), platform + straw bale (PS), pecking block (B), tunnel (T) or control (C; no enrichment) group. Welfare measures and gait were assessed at 8, 12, 16 and 19 wk and analyzed using PROC LOGISTIC with Firth bias-correction. Better wing FQ with age was observed in turkeys in S and T groups. Turkeys in the S group had better wing FQ at 16 (P = 0.028) and 19 wk (P = 0.011) vs. 8 wk. Wing FQ (P = 0.008) was better at 19 vs. 8 wk for T turkeys. FCON worsened over time for turkeys in all treatment groups except for the S group. FCON was worse at 19 vs.8 wk for P (P = 0.024), PS (P = 0.039), B (P = 0.011), T (P = 0.004) and C (P = 0.014) turkeys and was worse at 19 vs. 12 wk for B (P = 0.038), T (P = 0.015) and C (P = 0.045) turkeys. FCON was worse at 19 vs. 16 wk for T (P = 0.007) and C (P = 0.048) turkeys. FCON was also worse at 16 vs. 8 wk for B (P = 0.046) turkeys. Gait worsened with increasing age in all treatment groups. Gait was worse at 19 wk for S (P < 0.001), P (P < 0.001), PS (P < 0.001) and B turkeys (P < 0.001) vs. earlier ages, while gait in T (P < 0.001) and C turkeys (P < 0.001) worsened starting at 16 wk.

Doxorubicin inhibits cholesterol efflux through the miR-33/ABCA1 pathway.

Doxorubicin (DOX) is extensively used for the treatment of kinds of cancers, and cardiovascular toxicity is one of the side effects. However, it is unclear whether DOX causes impairment of cardiac function by promoting atherosclerosis. Thus, we investigated the role of DOX in regulating the lipid deposition of macrophages and its molecular mechanism. RAW 264.7 cell line was stimulated with DOX in the presence or absence of low-density lipoprotein (LDL). We found that DOX increased miR-33 and reduced ATP binding cassette transporter A1 (ABCA1) protein. Moreover, cholesterol efflux was suppressed by DOX, which was more efficient under a high-cholesterol condition. After transfecting mimics or inhibitors of miR-33 into cells, ABCA1 protein was respectively decreased and increased, and intracellular lipid accumulation was correspondingly regulated. Overall, DOX suppresses the expression of ABCA1 protein by upregulating miR-33, promoting an intracellular lipid deposition in macrophages, which is a sign of early atherosclerosis. This provides new insights for clinical observation and evaluation of the side effects of DOX.

Differences in Dietary Intake Exist among U.S. Adults by Diabetic Status Using NHANES 2009-2016.

The objective was to determine the most frequently consumed food items, food subcategories, and food categories, and those that contributed most to total energy intake for the group of U.S. adults reporting taking insulin, those with type 2 diabetes (T2D) not taking insulin, and those without diabetes. Laboratory tests and questionnaires of the National Health and Nutrition Examination Survey 2009-2016 classified 774 participants reporting taking insulin, 2758 participants reporting T2D not taking insulin, and 17,796 participants without diabetes. Raw and weighted frequency and energy contributions of each food item, food subcategory, and food category were calculated and ranked. Comparisons among groups by broad food category used the Rao-Scott modified chi-square test. Soft drinks ranked as the 8th and 6th most consumed food subcategory of participants with T2D not taking insulin and those without diabetes, and contributed 5th and 2nd most to energy, respectively. The group reporting taking insulin is likely to consume more protein foods and less soft drink compared to the other two groups. Lists of the most frequently reported foods and foods contributing most to energy may be helpful for nutrition education, prescribing diets, and digital-based dietary assessment for the group reporting taking insulin.

An Active Image-Based Mobile Food Record Is Feasible for Capturing Eating Occasions among Infants Ages 3-12 Months Old in Hawai'i.

The ability to comprehensively assess the diet of infants is essential for monitoring adequate growth; however, it is challenging to assess dietary intake with a high level of accuracy. Infants rely on surrogate reporting by caregivers. This study aimed to determine if surrogate reporters (e.g., caregivers) could use an image-based mobile food record adapted (baby mFR) to record infants' eating occasions, and via caregiver feedback, could assess the usability and feasibility of the baby mFR in recording infants' diets. This was a cross-sectional study in which surrogate reporters (e.g., caregivers) recorded all food and beverage intake (including human milk) of the infant over a 4-day period. Trained research staff evaluated all images submitted during data collection for different indicators of quality. All surrogate reporters were asked to complete a usability questionnaire at the end of the 4-day data collection period. Basic descriptive analyses were performed on the infants 3-12 months of age (n = 70). A total of 91% (n = 64) of surrogate reporters used the baby mFR to record their infants' eating occasions. The mean number of images submitted daily per participant via the mFR was 4.2 (SD 0.2). A majority of submitted images contained the fiducial marker and the food and/or beverage was completely visible. The mFR was found to be easy to use; however, suggestions were provided to increase utility of the application such as the inclusion of a bottle button and reminders. An image-based dietary assessment method using a mobile app was found to be feasible for surrogate reporters to record an infant's food and beverage intake throughout the day.

Corrigendum: Schisandrin B Antagonizes Cardiotoxicity Induced by Pirarubicin by Inhibiting Mitochondrial Permeability Transition Pore (mPTP) Opening and Decreasing Cardiomyocyte Apoptosis.

[This corrects the article DOI: 10.3389/fphar.2021.733805.].

Accuracy and Cost-effectiveness of Technology-Assisted Dietary Assessment Comparing the Automated Self-administered Dietary Assessment Tool, Intake24, and an Image-Assisted Mobile Food Record 24-Hour Recall Relative to Observed Intake: Protocol for a Randomized Crossover Feeding Study.

BACKGROUND: The assessment of dietary intake underpins population nutrition surveillance and nutritional epidemiology and is essential to inform effective public health policies and programs. Technological advances in dietary assessment that use images and automated methods have the potential to improve accuracy, respondent burden, and cost; however, they need to be evaluated to inform large-scale use. OBJECTIVE: The aim of this study is to compare the accuracy, acceptability, and cost-effectiveness of 3 technology-assisted 24-hour dietary recall (24HR) methods relative to observed intake across 3 meals. METHODS: Using a controlled feeding study design, 24HR data collected using 3 methods will be obtained for comparison with observed intake. A total of 150 healthy adults, aged 18 to 70 years, will be recruited and will complete web-based demographic and psychosocial questionnaires and cognitive tests. Participants will attend a university study center on 3 separate days to consume breakfast, lunch, and dinner, with unobtrusive documentation of the foods and beverages consumed and their amounts. Following each feeding day, participants will complete a 24HR process using 1 of 3 methods: the Automated Self-Administered Dietary Assessment Tool, Intake24, or the Image-Assisted mobile Food Record 24-Hour Recall. The sequence of the 3 methods will be randomized, with each participant exposed to each method approximately 1 week apart. Acceptability and the preferred 24HR method will be assessed using a questionnaire. Estimates of energy, nutrient, and food group intake and portion sizes from each 24HR method will be compared with the observed intake for each day. Linear mixed models will be used, with 24HR method and method order as fixed effects, to assess differences in the 24HR methods. Reporting bias will be assessed by examining the ratios of reported 24HR intake to observed intake. Food and beverage omission and intrusion rates will be calculated, and differences by 24HR method will be assessed using chi-square tests. Psychosocial, demographic, and cognitive factors associated with energy misestimation will be evaluated using chi-square tests and multivariable logistic regression. The financial costs, time costs, and cost-effectiveness of each 24HR method will be assessed and compared using repeated measures analysis of variance tests. RESULTS: Participant recruitment commenced in March 2021 and is planned to be completed by the end of 2021. CONCLUSIONS: This protocol outlines the methodology of a study that will evaluate the accuracy, acceptability, and cost-effectiveness of 3 technology-enabled dietary assessment methods. This will inform the selection of dietary assessment methods in future studies on nutrition surveillance and epidemiology. TRIAL REGISTRATION: Australian New Zealand Clinical Trials Registry ACTRN12621000209897; https://tinyurl.com/2p9fpf2s. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): DERR1-10.2196/32891.

Motion Artifact Reduction In Photoplethysmography For Reliable Signal Selection.

Photoplethysmography (PPG) is a non-invasive and economical technique to extract vital signs of the human body. Although it has been widely used in consumer and research grade wrist devices to track a user's physiology, the PPG signal is very sensitive to motion which can corrupt the signal's quality. Existing Motion Artifact (MA) reduction techniques have been developed and evaluated using either synthetic noisy signals or signals collected during high-intensity activities - both of which are difficult to generalize for real-life scenarios. Therefore, it is valuable to collect realistic PPG signals while performing Activities of Daily Living (ADL) to develop practical signal denoising and analysis methods. In this work, we propose an automatic pseudo clean PPG generation process for reliable PPG signal selection. For each noisy PPG segment, the corresponding pseudo clean PPG reduces the MAs and contains rich temporal details depicting cardiac features. Our experimental results show that 71% of the pseudo clean PPG collected from ADL can be considered as high quality segment where the derived MAE of heart rate and respiration rate are 1.46 BPM and 3.93 BrPM, respectively. Therefore, our proposed method can determine the reliability of the raw noisy PPG by considering quality of the corresponding pseudo clean PPG signal.

Schisandrin B Antagonizes Cardiotoxicity Induced by Pirarubicin by Inhibiting Mitochondrial Permeability Transition Pore (mPTP) Opening and Decreasing Cardiomyocyte Apoptosis.

Objective: Pirarubicin (THP), one of the anthracycline anticancer drugs, is widely used in the treatment of various cancers, but its cardiotoxicity cannot be ignored. Schisandrin B (SchB) has the ability to upregulate cellular antioxidant defense mechanism and promote mitochondrial function and antioxidant status. However, it has not been reported whether it can resist THP-induced cardiotoxicity. The aim of this study was to investigate the effect of SchB on THP cardiotoxicity and its mechanism. Methods: The rat model of cardiotoxicity induced by THP was established, and SchB treatment was performed at the same time. The changes of ECG, cardiac coefficient, and echocardiogram were observed. The changes of myocardial tissue morphology were observed by H&E staining. Apoptosis was detected by TUNEL. The levels of LDH, BNP, CK-MB, cTnT, SOD, and MDA in serum were measured to observe the heart damage and oxidative stress state of rats. The expression of cleaved-caspase 9, pro/cleaved-caspase 3, Bcl-2/Bax, and cytosol and mitochondrial Cyt C and Bax was evaluated by western blot. H9c2 cardiomyocytes were cocultured with THP, SchB, and mPTP inhibitor CsA to detect the production of ROS and verify the above signaling pathways. The opening of mPTP and mitochondrial swelling were detected by mPTP kit and purified mitochondrial swelling kit. Results: After 8 weeks, a series of cardiotoxicity manifestations were observed in THP rats. These adverse effects can be effectively alleviated by SchB treatment. Further studies showed that SchB had strong antioxidant and antiapoptotic abilities in THP cardiotoxicity. Conclusion: SchB has an obvious protective effect on THP-induced cardiotoxicity. The mechanism may be closely related to the protection of mitochondrial function, inhibition of mPTP opening, and alleviation of oxidative stress and apoptosis of cardiomyocytes.

Dietary Aspects to Incorporate in the Creation of a Mobile Image-Based Dietary Assessment Tool to Manage and Improve Diabetes.

Diabetes is the seventh leading cause of death in United States. Dietary intake and behaviors are essential components of diabetes management. Growing evidence suggests dietary components beyond carbohydrates may critically impact glycemic control. Assessment tools on mobile platforms have the ability to capture multiple aspects of dietary behavior in real-time throughout the day to inform and improve diabetes management and insulin dosing. The objective of this narrative review was to summarize evidence related to dietary behaviors and composition to inform a mobile image-based dietary assessment tool for managing glycemic control of both diabetes types (type 1 and type 2 diabetes). This review investigated the following topics amongst those with diabetes: (1) the role of time of eating occasion on indicators of glycemic control; and (2) the role of macronutrient composition of meals on indicators of glycemic control. A search for articles published after 2000 was completed in PubMed with the following sets of keywords "diabetes/diabetes management/diabetes prevention/diabetes risk", "dietary behavior/eating patterns/temporal/meal timing/meal frequency", and "macronutrient composition/glycemic index". Results showed eating behaviors and meal macronutrient composition may affect glycemic control. Specifically, breakfast skipping, late eating and frequent meal consumption might be associated with poor glycemic control while macronutrient composition and order of the meal could also affect glycemic control. These factors should be considered in designing a dietary assessment tool, which may optimize diabetes management to reduce the burden of this disease.

Utility of self-rated adherence for monitoring dietary and physical activity compliance and assessment of participant feedback of the Healthy Diet and Lifestyle Study pilot.

BACKGROUND: We examined the utility of self-rated adherence to dietary and physical activity (PA) prescriptions as a method to monitor intervention compliance and facilitate goal setting during the Healthy Diet and Lifestyle Study (HDLS). In addition, we assessed participants' feedback of HDLS. HDLS is a randomized pilot intervention that compared the effect of intermittent energy restriction combined with a Mediterranean diet (IER + MED) to a Dietary Approaches to Stop Hypertension (DASH) diet, with matching PA regimens, for reducing visceral adipose tissue area (VAT). METHODS: Analyses included the 59 (98%) participants who completed at least 1 week of HDLS. Dietary and PA adherence scores were collected 8 times across 12 weeks, using a 0-10 scale (0 = not at all, 4 = somewhat, and 10 = following the plan very well). Adherence scores for each participant were averaged and assigned to high and low adherence categories using the group median (7.3 for diet, 7.1 for PA). Mean changes in VAT and weight from baseline to 12 weeks are reported by adherence level, overall and by randomization arm. Participants' feedback at completion and 6 months post-intervention were examined. RESULTS: Mean ± SE, dietary adherence was 6.0 ± 0.2 and 8.2 ± 0.1, for the low and high adherence groups, respectively. For PA adherence, mean scores were 5.9 ± 0.2 and 8.5 ± 0.2, respectively. Compared to participants with low dietary adherence, those with high adherence lost significantly more VAT (22.9 ± 3.7 cm2 vs. 11.7 ± 3.9 cm2 [95% CI, - 22.1 to - 0.3]) and weight at week 12 (5.4 ± 0.8 kg vs. 3.5 ± 0.6 kg [95% CI, - 3.8 to - 0.0]). For PA, compared to participants with low adherence, those with high adherence lost significantly more VAT (22.3 ± 3.7 cm2 vs. 11.6 ± 3.6 cm2 [95% CI, - 20.7 to - 0.8]). Participants' qualitative feedback of HDLS was positive and the most common response, on how to improve the study, was to provide cooking classes. CONCLUSIONS: Results support the use of self-rated adherence as an effective method to monitor dietary and PA compliance and facilitate participant goal setting. Study strategies were found to be effective with promoting compliance to intervention prescriptions. TRIAL REGISTRATION: ClinicalTrials.gov Identifier: NCT03639350 . Registered 21st August 2018-retrospectively registered.

Correction: Improving Nutrition and Activity Behaviors Using Digital Technology and Tailored Feedback: Protocol for the Tailored Diet and Activity (ToDAy) Randomized Controlled Trial.

[This corrects the article DOI: 10.2196/12782.].

Food Security and Diet Quality in Native Hawaiian, Pacific Islander, and Filipino Infants 3 to 12 Months of Age.

Food insecurity and other nutritional risks in infancy pose a lifelong risk to wellbeing; however, their effect on diet quality in Native Hawaiian, Pacific Islander, and Filipino (NHPIF) infants in Hawai'i is unknown. In this cross-sectional analysis, the association between various indicators of food security and NHPIF infant diet quality were investigated in 70 NHPIF infants aged 3-12 months residing on O'ahu, Hawai'i. The dietary assessments of the infants were collected using a mobile food recordTM. Foods consumed across four days were categorized into seven food groups. Indicators for food security were examined through an adapted infant food security index and other indicators. Data were analyzed using chi-square tests, independent sample t-tests, multinomial logistic regression, and linear regression models. In models adjusting for age and sex, infants defined as food insecure by the adapted index were found to consume foods from more food groups and consume flesh foods on a greater proportion of days. Of the indicators examined, the adapted index was shown to be the best indicator for food group consumption. Further work is needed on a more representative sample of NHPIF infants to determine the impact that food security has on nutritional status and other indicators of health.

Type, Timing, and Diversity of Complementary Foods Among Native Hawaiian, Pacific Islander, and Filipino Infants.

Prevention is the recommended strategy for addressing childhood obesity and may be particularly important for minority groups such as Native Hawaiians, Pacific Islanders, and Filipinos (NHPIF) who display poorer health outcomes than other race/ethnic groups. Complementary feeding is a critical milestone in the first 1,000 days of life and plays a critical role in growth and eating habit formation. This cross-sectional study recruited NHPIF infants between 3 - 12 months of age residing on O'ahu, Hawai'i to examine timing and types of complementary foods introduced first as well as the dietary diversity of those infants 6 - 12 months of age. Basic demographic information and early feeding practices were assessed via online questionnaire. Diet was evaluated using the image-based mobile food record completed over 4-days. Images were evaluated to derive the World Health Organization's minimum dietary diversity (MDD) score. Data were analyzed using descriptive statistics and linear regressions. Seventy participants completed the study with a majority being between the ages of 6 - 12 months (n=56). About half of the participants were provided a complementary food prior to 6 months of age with the most common first complementary food being poi (steamed, mashed taro). Grains were the most commonly reported food group while the high protein food groups was the least commonly reported. Approximately 25% of infants 6 - 12 months of age met MDD all four days. Meeting MDD was significantly associated with age. Findings illuminate opportunities for improvement (eg, delayed introduction) and for promotion (eg, cultural foods) in NHPIF complementary feeding.