A Scoping Review of Sensor-Based Capture of Eating and Drinking Occasions That Could Be Used for Enhancing Personalized Nutrition Interventions in Real Time.

BACKGROUND

Traditional dietary assessment methods used in nutrition research and practice are self-reported, burdensome, and prone to error, limiting utility. In recent years, sensor-based devices and machine learning approaches have emerged as promising tools for automating eating behavior detection and initiating different approaches to assessing intake. These technologies have potential to enhance dietary assessment and its accuracy, support personalized dietary interventions through real time, context-aware feedback, and reduce burden on respondents and practitioners. A prior 2021 review by the authors concluded that existing devices are not yet feasible for dietetic practice.

OBJECTIVES

This study aims to conduct a scoping review of sensor-based devices capable of detecting eating and drinking and to evaluate whether recent advancements have improved their feasibility for use in real-world nutrition applications.

METHODS

A scoping review was conducted using the Preferred Reporting Items for Systematic reviews and Meta-Analyses for Scoping Reviews framework. Studies published between January 2022 and September 2025 that evaluated the performance of sensor-based devices in identifying food and/or beverage intake were included. Devices were evaluated against 6 feasibility criteria to assess real-world applicability: ≥80% accuracy, freedom in food and beverage selection; social acceptability and comfort; long battery life; real-time detection; and ability to detect both eating and drinking.

RESULTS

Fifty studies (52 devices) were included: 19 wrist-worn, 8 neck-worn, 7 ear-worn, 7 glasses-type, 6 in the "other" category, and 5 multiposition devices. None met all 6 feasibility criteria. The most common unmet criterion was adequate battery life (n = 43), followed by real-time processing (n = 37), variety of foods or behaviors in testing (n = 31), detection of both eating and drinking (n = 31), social acceptability and comfort (n = 15), and accuracy (n = 10).

CONCLUSIONS

Although no sensor-based devices met all criteria for real-world feasibility, recent advancements suggest meaningful progress in areas of social acceptability and computational efficiency. These improvements signal a shift toward more practical, user-friendly designs that may soon be capable of supporting automated dietary assessment and individualized nutrition care.