Development and validation of nutrient estimates based on a food-photographic record in Japan

Development, Relative Validity, and Reproducibility of a Short Food Frequency Questionnaire for the Japanese

We have developed a short food frequency questionnaire (FFQ) for use in assessing diet quality for Japan, with special reference to the prevention of non-communicable diseases. Here, we assessed the ranking performance of this FFQ and its reproducibility. We developed a 28-item (21 food groups and 7 beverage) FFQ with consideration to both Japanese dietary culture and evidence of disease prevention. Twenty-four university faculty members participated in the validation study. They completed 3-day photographic food record and answered the FFQ on the next day of the last food record (time 1) and a week later (time 2). We calculated Spearman correlation coefficients between intakes of food groups from photographic food records and the consumption frequency from the FFQs (ranking ability) and between the consumption frequency of food groups from the FFQs (time 1 and time 2) (reproducibility). Spearman correlation coefficients between the food records and FFQ (time 1) ranged from −0.12 to 0.86 (median 0.51). These values were comparable to those in comparison with FFQ (time 2). After energy adjustment of intakes from the food records, the corresponding values were somewhat weakened for many food groups. The correlation coefficients between two FFQs ranged from 0.14 to 0.96 (median 0.79). The short FFQ showed acceptable reproducibility and ability to rank the consumption of most food groups.

Usefulness of Simple Meal Screening Using Dietary Assessment Sheets for Evacuation Shelters

To maintain the health of evacuees, meals provided at emergency shelters should be monitored and improvement measures should be taken where necessary. Time and human resources are limited during devastating disasters; thus, the authors have developed the Dietary Assessment Sheets for Evacuation Shelters. This study examines whether the quality of meals can be assessed using simple check items on the Dietary Assessment Sheets.

A total of 28 Dietary Assessment Sheets for 98 meals provided 16-19 days after the heavy rains in Kumamoto in July 2020 at 12 shelters were used in the analysis. Dietitians working for shelters were requested to fill the Dietary Assessment Sheets and food record sheets provided by the Prefectural Government. If any subitem was checkmarked, it was considered "checked," and if left blank, it was considered "not checked" and divided into categorical groups. Based on the food record sheets and meal photos, contents of energy, protein, and vitamins B₁, B₂, and C were calculated for each meal. Energy and nutrient contents were compared between the categorical groups.

The study revealed that nutritionally poor meals were characterized as those comprising only "Grain dishes," those that were not "Boxed meals," and those involving "Self-Defense Forces," "Dietitians," or "Others." Their energy, protein, and vitamin B₁, B₂, and C content was significantly lower than that of meals with marks on other checkboxes in the same category.

Shelters that provide these meals should be given priority for nutrition assistance due to their poor diets. The results imply that the Dietary Assessment Sheets could serve as a simple tool to determine the shelters that require nutritional assistance.

A Cross-Sectional Reproducibility Study of a Standard Camera Sensor Using Artificial Intelligence to Assess Food Items: The FoodIntech Project

Having a system to measure food consumption is important to establish whether individual nutritional needs are being met in order to act quickly and to minimize the risk of undernutrition. Here, we tested a smartphone-based food consumption assessment system named FoodIntech. FoodIntech, which is based on AI using deep neural networks (DNN), automatically recognizes food items and dishes and calculates food leftovers using an image-based approach, i.e., it does not require human intervention to assess food consumption. This method uses one-input and one-output images by means of the detection and synchronization of a QRcode located on the meal tray. The DNN are then used to process the images and implement food detection, segmentation and recognition. Overall, 22,544 situations analyzed from 149 dishes were used to test the reliability of this method. The reliability of the AI results, based on the central intra-class correlation coefficient values, appeared to be excellent for 39% of the dishes (n = 58 dishes) and good for 19% (n = 28). The implementation of this method is an effective way to improve the recognition of dishes and it is possible, with a sufficient number of photos, to extend the capabilities of the tool to new dishes and foods.

Accuracy of estimates of serving size using digitally displayed food photographs among Japanese adults

We evaluated the accuracy of the estimated serving size using digital photographs in a newly developed food atlas. From 209 food items in the food atlas, we selected 14 items with various appearances for evaluation. At the study site, fifty-four participants aged 18-33 years were served fourteen foods in the amount they usually ate. After they left, each food item was weighed by a researcher. The following day, the participants estimated the quantity of each food they served based on food photographs using a web-based questionnaire. We compared the weights of the foods the participants served (true serving sizes) and those determined based on the photographs (estimated serving sizes). For ten of the fourteen food items, significant differences were observed between the estimated and true serving sizes, ranging from a 29⋅8 % underestimation (curry sauce) to a 34⋅0 % overestimation (margarine). On average, the relative difference was 8⋅8 %. Overall, 51⋅6 % of the participants were within ±25 % of the true serving size, 81⋅9 % were within ±50 % and 93⋅4 % were within ±75 %. Bland-Altman plots showed wide limits of agreement and increased variances with larger serving sizes for most food items. Overall, no association was found between estimation errors and participant characteristics. The food atlas has shown potential for assessment of portion size estimation. Further development, refinement and testing are needed to improve the usefulness of the digital food photographic atlas as a portion size estimation aid.