Dataset Information

Use of Different Food Image Recognition Platforms in Dietary Assessment: Comparison Study.

ABSTRACT:

Background

In the domain of dietary assessment, there has been an increasing amount of criticism of memory-based techniques such as food frequency questionnaires or 24 hour recalls. One alternative is logging pictures of consumed food followed by an automatic image recognition analysis that provides information on type and amount of food in the picture. However, it is currently unknown how well commercial image recognition platforms perform and whether they could indeed be used for dietary assessment.

Objective

This is a comparative performance study of commercial image recognition platforms.

Methods

A variety of foods and beverages were photographed in a range of standardized settings. All pictures (n=185) were uploaded to selected recognition platforms (n=7), and estimates were saved. Accuracy was determined along with totality of the estimate in the case of multiple component dishes.

Results

Top 1 accuracies ranged from 63% for the application programming interface (API) of the Calorie Mama app to 9% for the Google Vision API. None of the platforms were capable of estimating the amount of food. These results demonstrate that certain platforms perform poorly while others perform decently.

Conclusions

Important obstacles to the accurate estimation of food quantity need to be overcome before these commercial platforms can be used as a real alternative for traditional dietary assessment methods.

SUBMITTER: Van Asbroeck S

PROVIDER: S-EPMC7752530 | biostudies-literature | 2020 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Use of Different Food Image Recognition Platforms in Dietary Assessment: Comparison Study.

Van Asbroeck Stephanie S Matthys Christophe C

JMIR formative research 20201207 12

<h4>Background</h4>In the domain of dietary assessment, there has been an increasing amount of criticism of memory-based techniques such as food frequency questionnaires or 24 hour recalls. One alternative is logging pictures of consumed food followed by an automatic image recognition analysis that provides information on type and amount of food in the picture. However, it is currently unknown how well commercial image recognition platforms perform and whether they could indeed be used for dieta ...[more]

PMID: 33284118

Similar Datasets

Project description:BACKGROUND:A Web-based dietary assessment tool-RiksmatenFlex-was developed for the national dietary survey of adolescents in Sweden. OBJECTIVE:This study aimed to describe the Web-based method RiksmatenFlex and to test the validity of the reported dietary intake by comparing dietary intake with 24-hour dietary recalls (recall interviews), estimated energy expenditure, and biomarkers. METHODS:Adolescents aged 11-12, 14-15, and 17-18 years were recruited through schools. In total, 78 students had complete dietary information and were included in the study. Diet was reported a few weeks apart with either RiksmatenFlexDiet (the day before and a random later day) or recall interviews (face-to-face, a random day later by phone) in a cross-over, randomized design. At a school visit, weight and height were measured and blood samples were drawn for biomarker analyses. Students wore an accelerometer for 7 days for physical activity measurements. Dietary intake captured by both dietary methods was compared, and energy intake captured by both methods was compared with the accelerometer-estimated energy expenditure (EEest). Intake of whole grain wheat and rye and fruit and vegetables by both methods was compared with alkylresorcinol and carotenoid concentrations in plasma, respectively. RESULTS:The mean of the reported energy intake was 8.92 (SD 2.77) MJ by RiksmatenFlexDiet and 8.04 (SD 2.67) MJ by the recall interviews (P=.01). Intake of fruit and vegetables was 224 (169) g and 227 (150) g, and whole grain wheat and rye intake was 12.4 (SD 13.2) g and 12.0 (SD 13.1) g, respectively; the intakes of fruit and vegetables as well as whole grain wheat and rye did not differ between methods. Intraclass correlation coefficients ranged from 0.57 for protein and carbohydrates to 0.23 for vegetables. Energy intake by RiksmatenFlexDiet was overreported by 8% (P=.03) but not by the recall interviews (P=.53) compared with EEest. The Spearman correlation coefficient between reported energy intake and EEest was 0.34 (P=.008) for RiksmatenFlexDiet and 0.16 (P=.21) for the recall interviews. Spearman correlation coefficient between whole grain wheat and rye and plasma total alkylresorcinol homologs was 0.36 (P=.002) for RiksmatenFlexDiet and 0.29 (P=.02) for the recall interviews. Spearman correlations between intake of fruit and vegetables and plasma carotenoids were weak for both dietary tools. The strongest correlations were observed between fruit and vegetable intake and lutein/zeaxanthin for RiksmatenFlexDiet (0.46; P<.001) and for recall interviews (0.28; P=.02). CONCLUSIONS:RiksmatenFlexDiet provides information on energy, fruit, vegetables, and whole grain wheat and rye intake, which is comparable with intake obtained from recall interviews in Swedish adolescents. The results are promising for cost-effective dietary data collection in upcoming national dietary surveys and other studies in Sweden. Future research should focus on how, and if, new technological solutions could reduce dietary reporting biases.

Project description:BackgroundComplying with individual privacy perceptions is essential when processing personal information for research. Our specific research area is performance development of elite athletes, wherein nutritional aspects are important. Before adopting new automated tools that capture such data, it is crucial to understand and address the privacy concerns of the research subjects that are to be studied. Privacy as contextual integrity emphasizes understanding contextual sensitivity in an information flow. In this study, we explore privacy perceptions in image-based dietary assessments. This research field lacks empirical evidence on what will be considered as privacy violations when exploring trends in long-running studies. Prior studies have only classified images as either private or public depending on their basic content. An assessment and analysis are thus needed to prevent unwanted consequences of privacy breach and other issues perceived as sensitive when designing systems for dietary assessment by using food images.ObjectiveThe aim of this study was to investigate common perceptions of computer systems using food images for dietary assessment. The study delves into perceived risks and data-sharing behaviors.MethodsWe investigated the privacy perceptions of 105 individuals by using a web-based survey. We analyzed these perceptions along with perceived risks in sharing dietary information with third parties.ResultsWe found that understanding the motive behind the use of data increases its chances of sharing with a social group.ConclusionsIn this study, we highlight various privacy concerns that can be addressed during the design phase. A system design that is compliant with general data protection regulations will increase participants' and stakeholders' trust in an image-based dietary assessment system. Innovative solutions are needed to reduce the intrusiveness of a continuous assessment. Individuals show varying behaviors for sharing metadata, as knowing what the data is being used for, increases the chance of it being shared.

Project description:BackgroundAccurate dietary assessment is needed in studies that include analysis of nutritional intake. Image-based dietary assessment apps have gained in popularity for assessing diet, which may ease researcher and participant burden compared to traditional pen-to-paper methods. However, few studies report the validity of these apps for use in research. Keenoa is a smartphone image-based dietary assessment app that recognizes and identifies food items using artificial intelligence and permits real-time editing of food journals.ObjectiveThis study aimed to assess the relative validity of an image-based dietary assessment app - Keenoa - against a 3-day food diary (3DFD) and to test its usability in a sample of healthy Canadian adults.MethodsWe recruited 102 participants to complete two 3-day food records. For 2 weeks, on 2 non-consecutive days and 1 weekend day, in random order, participants completed a traditional pen-to-paper 3DFD and the Keenoa app. At the end of the study, participants completed the System Usability Scale. The nutrient analyses of the 3DFD and Keenoa data before (Keenoa-participant) and after they were reviewed by dietitians (Keenoa-dietitian) were analyzed using analysis of variance. Multiple tests, including the Pearson coefficient, cross-classification, kappa score, % difference, paired t test, and Bland-Altman test, were performed to analyze the validity of Keenoa (Keenoa-dietitian).ResultsThe study was completed by 72 subjects. Most variables were significantly different between Keenoa-participant and Keenoa-dietitian (P<.05) except for energy, protein, carbohydrates, fiber, vitamin B1, vitamin B12, vitamin C, vitamin D, and potassium. Significant differences in total energy, protein, carbohydrates, % fat, saturated fatty acids, iron, and potassium were found between the 3DFD and Keenoa-dietitian data (P<.05). The Pearson correlation coefficients between the Keenoa-dietitian and 3DFD ranged from .04 to .51. Differences between the mean intakes assessed by the 3DFD and Keenoa-dietitian were within 10% except for vitamin D (misclassification rate=33.8%). The majority of nutrients were within an acceptable range of agreement in the Bland-Altman analysis; no agreements were seen for total energy, protein, carbohydrates, fat (%), saturated fatty acids, iron, potassium, and sodium (P<.05). According to the System Usability Scale, 34.2% of the participants preferred using Keenoa, while 9.6% preferred the 3DFD.ConclusionsThe Keenoa app provides acceptable relative validity for some nutrients compared to the 3DFD. However, the average intake of some nutrients, including energy, protein, carbohydrates, % fat, saturated fatty acids, and iron, differed from the average obtained using the 3DFD. These findings highlight the importance of verifying data entries of participants before proceeding with nutrient analysis. Overall, Keenoa showed better validity at the group level than the individual level, suggesting it can be used when focusing on the dietary intake of the general population. Further research is recommended with larger sample sizes and objective dietary assessment approaches.

Dataset Information

Use of Different Food Image Recognition Platforms in Dietary Assessment: Comparison Study.

Background

Objective

Methods

Results

Conclusions

Publications

Use of Different Food Image Recognition Platforms in Dietary Assessment: Comparison Study.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets