Evaluation of an Artificial Intelligence-Assisted, Image-Based Dietary Assessment Tool in the Framingham Heart Study

Evaluation of an Artificial Intelligence-Assisted, Image-Based Dietary Assessment Tool in the Framingham Heart Study: A Block Randomized Controlled Trial

Assessment of dietary intake in large, free-living populations is inherently challenging due to the complex nature of human diet. Advancements in traditional methods of dietary assessment (i.e., web-based dietary recalls or records) have aimed at improving data accuracy while reducing participant burden. Further utilizing food recognition technologies to capture real-time food intake may aid in overcoming limitations of existing methods. Keenoa, an artificial intelligence-enhanced, image-assisted tool, is a newly designed mobile application that may facilitate collection of dietary data.

Primarily, the investigators will assess acceptability and usability of Keenoa compared with the traditional, web-based Automatic Self-Administered 24-Hour (ASA24) Dietary Assessment Tool in the Framingham Heart Study Third Generation-based cohorts at examination 4. The investigators will also determine the proportion of participants who complete all three days of dietary assessment, either through Keenoa or ASA24. Further, the investigators will relate dietary determinants of glycemic variability (e.g., percent carbohydrate, fiber intake, etc.), obtained from each dietary assessment tool, to the continuous glucose monitor (CGM)-derived outcomes.

With a randomized block design, this study will take place as part of the Framingham Heart Study (FHS) glucose study (R01 DK129305). Currently participants from the Third Generation-based cohorts are asked at their fourth examination to wear Dexcom G6 Pro continuous glucose monitor on either their arm or abdomen for a duration of at least 4 days. During this time, participants are asked to complete 3 consecutive days of dietary record through ASA24. For this trial, the investigators will randomize the dietary assessment tool weekly between ASA24 and Keenoa, therefore, depending on the week of administration, participants will be randomized to either a 3 days dietary record via ASA24 or a 3-day dietary record through Keenoa. This trial will last a total of 6 weeks.

Inclusion Criteria:

• FHS participants from the Gen 3 cohort who will attend the Exam 4 cycle

Exclusion Criteria:

• Participants with cognitive impairment which may prevent them from being able to complete diet records and follow instructions for continuous glucose monitor device wear.