This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mhealth and uhealth, is properly cited. The complete bibliographic information, a link to the original publication on http://mhealth.jmir.org/, as well as this copyright and license information must be included.
Dietary management is important for personal health. However, it is challenging to record quantified food information in an efficient, accurate, and sustainable manner, particularly for the consumption of Chinese food.
The objective of this study was to develop a dietary management system to record information on consumption of Chinese food, which can help in assessing individuals’ dietary intake and maintaining healthy eating behaviors. We proposed to use plates embedded with radio-frequency identification chips to carry Chinese foods and collect food consumption data.
We obtained food composition and nutrient (eg, carbohydrate, fat, fiber) data from the Chinese Recipe Database and China Food Composition Database. To test the feasibility of the dietary management system at a population level, we applied it to collect data on 489 Chinese foods that were consumed at lunchtime across 7 weeks by 10,528 individuals. To test individual-level output, we selected an individual participant with completed 20-day dietary data for analysis. We examined the system’s nutrient calculation performance by comparing the nutrient values of 3 selected Chinese dishes calculated by our method with the results of chemical measurements.
We collected the dietary intake for a group of 10,528 individuals aged from 20 to 40 years having lunch in a restaurant across 7 weeks. A total of 489 Chinese dishes were identified. We analyzed a specified customer’s diet recordings and broke his or her 20 lunch diet recordings down to ingredients and then to nutrient intake. We compared the nutrient value of a given Chinese dish (eg, garlic puree cooked pork leg) calculated by our method with the results of chemical measurements. The mean absolute percentage deviation showed that our method enabled collection of dietary intake for Chinese foods.
This preliminary study demonstrated the feasibility of radio-frequency identification–based dietary management for Chinese food consumption. In future, we will investigate factors such as preparation method, weight of food consumed, and auxiliary ingredients to improve dietary assessment accuracy.
Diet is an important risk factor or prevention-related factor for health management and disease treatment. Healthy eating behaviors can help decrease the burden of chronic diseases such as overweight, cardiovascular disease, liver disease, and diabetes [
Commonly used dietary assessment methods include dietary records, 24-hour dietary recall, food frequency questionnaires, and brief dietary assessment instruments [
Technical advances in information and communication technology have improved the automatic collection of self-reported dietary data by means such as standardized question sequencing, fasting data processes, and increased flexibility [
In this study, we proposed to apply a sensor technology, radio-frequency identification (RFID), to collect information on consumption of Chinese food.
RFID technology enables information interaction or exchange without human intervention and awareness [
In Chinese diets, foods are cooked according to numerous recipes, and are classified as staple foods (eg, rice, steamed buns), cooked dishes (eg, cooked tomato with eggs), and soups (eg, egg drop soup). In this study, the RFID plates were used to carry the Chinese foods. We developed our dietary management system to automatically collect food consumption information (ie, food names and food frequency) but no personal information of the consumers (eg, sex, age, and smoking habits). The system further provided functions of food frequency statistics, food composition overview, and proxy indicators of dietary intake.
A total of 10,528 participants took part in this study. Participants’ ages ranged from 20 to 40 years, and all had access to a restaurant equipped with a dietary management system for lunch. Each of them was assigned a unique identification (ID) the first time they had meals in the restaurant.
We designed and implemented a dietary management system for the collection of Chinese food information (see
We used iPlate (Hangzhou Sovell Technology Development Company Limited), a research and development solution for dietary management, which provides an open platform supporting Chinese nutrient research [
We classified foods into 3 categories, put onto 3 types of plates: staple foods (eg, rice) are on plate 1, cooked dishes (eg, cooked tomato with eggs) are on plate 2, and soups (eg, egg drop soup) are on plate 3. The RFID chips embedded in the plates recorded information about the foods, including their name, price, and weight. When a diner checked out after choosing a plate of food, the RFID reader interacted with the plate, recording food consumption information by connecting the consumer’s ID with the food information.
Each plate contained 1 type of food or dish. The weight of food in each plate was set by the restaurant according to the food itself and the type of plate. Dietary information (cooked food price, food name, consumer ID, and time of meal: breakfast, lunch, or supper) was stored in the individual food consumption database.
To analyze the composition of the chosen foods, we developed 2 databases: a recipe database and a nutrient intake database.
The recipe database stored the Chinese food recipes, including the food names, main ingredients and their corresponding weights, and preparation methods. The function of the recipe database was to break down the Chinese foods on the RFID plates into their ingredients.
The function of the food composition database was to break down the ingredients into their nutrient components. We used the China Food Composition Database [
We selected 3 specific Chinese dishes for nutrient calculations. We first broke down their ingredients according to the recipe database. Then, we calculated the dishes’ total nutrients by summing the nutrients of each food ingredient according to the China Food Composition Database. We coded Chinese food in the food composition table. Our previous studies on data representation for food nutritional composition [
A radio-frequency identification (RFID)–based dietary management system for Chinese foods.
Data selection process for dietary records.
We used the following criteria to select RFID-collected data for further analysis, as
To test the feasibility of the dietary management system, we applied it to collect data on 489 kinds of Chinese foods, which were consumed at lunchtime by our study participants from January 5 to February 15, 2016. We used these examples to test our system output at the population level, including consumed food distribution and preparation method distribution. To test our system at an individual level, we selected one participant who completed 20-day dietary data.
We obtained 489 kinds of unique Chinese dishes eaten by diners at lunchtime. The system monitored the distribution of the ingredients of all the dishes provided, as
Following Chinese dietary habits, we divided the 489 dishes of consumed Chinese foods into 3 types: staple food (83/489, 17.0%), cooked food (391/489, 80.0%), and porridge and soup (15/489, 3.0%).
The proportion of cooked food is 80% (n=391), which means it is an indispensable component of Chinese food. Based on the cooked food recorded, we analyzed the distribution of preparation methods at the population level (see
To demonstrate the individual-level system output, we selected an individual with completed 20-day dietary data for analysis.
Example of system output: distribution of consumed foods.
Example of system output: distribution of preparation methods.
Example of system output: frequency of foods consumed by 1 sample participant.
Example of system output: a sample individual’s preferred food composition (left) and preparation methods (right).
Example of system output: energy (left) and nutrient calculation results (right) for a sample individual. Blue: carbohydrate; orange: fat; gray: protein; green: fiber; red: energy (kcal).
To examine our nutrient calculation performance, we selected 3 Chinese food dishes for comparison analysis. The 3 dishes were garlic puree cooked pork leg, dry-fried string beans, and roast lamb, which were measured by chemical methods in a previous study [
Nutrient calculation results compared with chemical measurements.
Chinese food dishes, ingredients and preparation method | Nutrient | Nutrient content measure | |||
MAPDc (%) | |||||
Pork, garlic, scallions, pepper; steaming | Energy (kcal) | 288.1 | 253 | 13.87 | |
Protein (g) | 25 | 22 | 13.64 | ||
Fat (g) | 20.9 | 18.3 | 14.21 | ||
Carbohydrate (g) | 0 | 0 | — | ||
String beans, minced pork, dried chilies, |
Energy (kcal) | 210.7 | 229 | 7.99 | |
Protein (g) | 9.25 | 10.8 | 14.35 | ||
Fat (g) | 18.2 | 16.7 | 8.98 | ||
Carbohydrate (g) | 5.73 | 9 | 36.33 | ||
Lamb, onion, cumin, oil; roasting | Energy (kcal) | 133.4 | 162 | 17.65 | |
Protein (g) | 19.4 | 21.5 | 9.77 | ||
Fat (g) | 6.2 | 8.1 | 23.46 | ||
Carbohydrate (g) | 0 | 0.8 | 100.00 |
aNutrient content calculated in this study.
bMeasurement obtained by chemical methods (Kjeldahl nitrogen method for protein content and acid hydrolysis method for fat content).
cMAPD: mean absolute percentage deviation.
We developed a dietary management system to collect information on Chinese food using RFID technology. We applied the system to process real-world dietary data to test its feasibility. As shown in the preliminary results, the system can automatically record the Chinese foods carried by RFID-embedded plates and generate food frequency reports. With the support of the Chinese Recipe Database and China Food Composition Database, our dietary system broke down the cooked Chinese foods into their ingredients and nutrients. The system outputs included data on the frequency of the foods consumed and an overview of the composition of chosen foods, at both a population level and an individual level.
We compared the nutrients calculated for 3 selected Chinese foods with those measured by chemical methods. The result showed that the deviation was less than 20%. This deviation in calculated nutrients may have been caused by the instability in ingredient proportions and the change in food ingredient compositions according to location, season, and weather. The estimated weight of each dish and the ingredients may also result in deviation. However, comparison of 3 samples was insufficient to established a solid conclusion. More samples should be included for calculating
Our study had 3 main limitations. First, we estimated the weight of the food on each kind of plate by the type of plate and the food itself. We did not weigh exactly the consumed part on each plate. This may have affected the accuracy of our nutrient intake calculation. Second, the preparation methods used in Chinese cuisine affected the nutrient calculation results. Recent studies investigated the effects of cooking methods on specific foods [
In this study, we used an RFID-based system to track and record the foods and their corresponding weight eaten by our study group. We used 2 databases, a recipe database and a food composition database, to obtain the distribution of ingredients consumed by the group and to validate the calculated nutrients.
In future research, we will consider more factors (eg, preparation method, weight of the consumed portion, and auxiliary ingredients) in the nutrition calculation and improve the accuracy of our results. Additionally, we will combine the nutrition analysis with the consumer’s information, including health condition and demographic information such as age, sex, and birthplace. All of these efforts aim to provide more personalized dietary management.
identification
mean absolute percentage deviation
radio-frequency identification
This work was supported by the National Natural Science Foundation of China (grant #81601573), National Social Science Foundation of China (grant #14BTQ032), National Population and Health Scientific Data Sharing Program of China, the National Key Research and Development Program of China (grant #2016YFC0901901), and the Key Laboratory of Medical Information Intelligent Technology Chinese Academy of Medical Sciences. The authors thank Hangzhou Sovell Technology Development Company Limited for providing the iPlate system solution and valuable discussion on the study.
None declared.