The accuracy of portion size estimation using food images and textual descriptions of portion sizes: an evaluation study

• Published in: Journal of human nutrition and dietetics Vol. 34; no. 6; pp. 945 - 952
• Main Authors: Lucassen, Desiree A., Willemsen, Romy F., Geelen, Anouk, Brouwer‐Brolsma, Elske M., Feskens, Edith J.M.
• Format: Journal Article
• Language: English
• Published: England Blackwell Publishing Ltd 01-12-2021; John Wiley and Sons Inc
• Subjects: Beverages; Descriptions; Diet; Dietary assessment; Dietary intake; Energy Intake; Error analysis; Error reduction; Food; Food images; Food intake; Household measures; Humans; Lunch; Lunches; Nutrition Assessment; Nutritional Assessment; Portion Size; Portion size estimation; PSEA; Self Report; Size Perception; Standard portion sizes; Dietary assessment; Household measures; PSEA; Standard portion sizes; Portion size estimation; Food images
• ISSN: 0952-3871; 1365-277X
• DOI: 10.1111/jhn.12878

Background
Inaccurate self‐report of portion sizes is a major cause of measurement error in dietary assessment. To reduce this error, different portion size estimation aids (PSEAs) have been developed, including food images (image based, IB‐PSE) and textual descriptions of portion sizes (text‐based, TB‐PSE). We assessed the accuracy of portion size estimation by IB‐PSE and TB‐PSE.
Methods
True intake of one lunch was ascertained in forty participants. Self‐reported portion sizes were assessed after 2 and 24 hours by means of TB‐PSE and IB‐PSE, in random order. Wilcoxon's tests were used to compare mean true intakes to reported intakes. Moreover, proportions of reported portion sizes within 10% and 25% of true intake were assessed. An adapted Bland‐Altman approach was used to assess agreement between true and reported portion sizes. Analyses were conducted for all foods and drinks combined and for predetermined food types.
Results
No significant differences were observed between reported portion sizes at 2 and 24 hours after lunch. Combining median relative errors of all foods items resulted in an overall 0% error rate for TB‐PSE and 6% error rate for IB‐PSE. Comparing reported portion sizes within 10% (31% vs. 13%) and 25% (50% vs. 35%) of the true intake showed a better performance for TB‐PSE compared to IP‐PSE, respectively. Bland‐Altman plots indicated a higher agreement between reported and true intake for TB‐PSE compared to IB‐PSE.
Conclusions
Although the use of TB‐PSE still results in measurement error, our results suggest a more accurate dietary intake assessment with TB‐PSE than IB‐PSE.
Inaccurate self‐report of portion sizes is a major cause of measurement error in dietary assessment. To reduce this error, different portion size estimation aids have been developed. We assessed the accuracy of portion size estimation by using food images and textual descriptions of portion sizes.