Introduction: Carbohydrate counting is an important part of diabetes management, but it is difficult and can negatively impact quality of life. To address this challenge, a novel mobile application (app) called iSpy that uses computer vision and artificial intelligence to identify foods, estimate their carbohydrate content, and provide real-time feedback to users, was developed.

Objective: To evaluate iSpy’s functionality among youth aged 10 to 17 years (mean age = 14 years) with type 1 diabetes mellitus (T1DM).

Design/Methods: A pilot study among 44 youth (22 each randomized to intervention (iSpy) or control) was conducted to assess: (1) whether iSpy leads to improved carbohydrate counting ability (accuracy and speed) measured using standardized meal plates at baseline and at the end of the 12-week study; and (2) implementation outcomes such as app acceptability and technical problems.

Results: Compared to the control group, the intervention group improved its carbohydrate counting accuracy (total grams per meal) over the 12-week study period (p=0.007). Moreover, the percentage of times the carbohydrate estimate of a particular item was off by more than 10g from the true value was reduced by 9.9% (p=0.047). There was no statistically significant difference in total time to count carbohydrates. Qualitative interviews and app acceptability scale scores across 7 domains were primarily positive. 43% of iSpy participants were still using the app and logged at least once per 2 weeks at the end of the study. No major technical challenges were identified.

Conclusion(s): Participants using iSpy demonstrated improved carbohydrate counting accuracy and positive acceptability. iSpy has the potential to improve diabetes management, which should be further tested in a formal randomized controlled trial.


E. Choi: Other Relationship; Self; Inner Analytics Inc. J. Alfonsi: Other Relationship; Self; Inner Analytics Inc. T. Arshad: None. S. Sammott: None. V. Pais: None. J. Stinson: None. M. Palmert: None.


Physician Services Incorporated Foundation

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at