Prospective Independent Evaluation of the Carbohydrate Counting Accuracy of Two Smartphone Applications

Abstract

Introduction: Smartphone applications (apps) have been designed that help patients to accurately count their carbohydrate intake in order to optimize prandial insulin dose matching. Our aim was to evaluate the accuracy of two carbohydrate (carb) counting apps. Methods: Medical students, in the role of mock patients, evaluated meals using two smartphone apps: Foodvisor Ò (which uses automatic food photo recognition technology) and Glucicheck Ò (which requires the manual entry of carbohydrates with the help of a photo gallery). The macronutrient quantifications obtained with these two apps were compared to a reference quantification. Results: The carbohydrate content of the entire meal was underestimated with Foodvisor Ò (Foodvisor Ò quantification minus gold standard quantification = -7.2 ± 17.3 g; p \ 0.05) but reasonably accurately estimated with Glucicheck Ò (Glucicheck Ò quantification minus gold standard quantification = 1.4 ± 13.4 g; ns). The percentage of meals with an absolute error in carbohydrate quantification above 20 g was greater for Foodvisor Ò compared to Glucicheck Ò (30% vs 14%; p \ 0.01). Conclusion:The carb counting accuracy was slightly better when using Glucicheck Ò compared to Foodvisor Ò . However, both apps provided a lower mean absolute carb counting error than that usually made by T1D patients in everyday life, suggesting that such apps may be a useful adjunct for estimating carbohydrate content.

Uso de aplicativos para contagem de carboidratos como ferramenta de auxílio no autogerenciamento do diabetes mellitus tipo 1: uma revisão sistemática

“…Em contrapartida, pesquisas realizadas com o intuito de avaliar a precisão da CCHO por diferentes aplicativos, obtiveram como resultado um erro absoluto, ou seja, a diferença entre o valor medido e o valor real, na contagem em torno 10-15g, a depender do sistema utilizado (Chotwanvirat et al, 2021;Joubert et al, 2021). Entretanto, quando comparado à estimativa realizada por pacientes, foram encontrados valores de erro absoluto no teor de CHO de 27,9g, sendo constatado que, quanto maior o erro absoluto do teor de CHO, maior será a variabilidade glicêmica e menor será o TIR, de tal forma que estes fatores podem ocasionar complicações metabólicas (Rhyner et al, 2016).…”

Glycaemic control and novel technology management strategies in pregestational diabetes mellitus

“…Another innovation which has shown promise in the management of type 1 diabetes is the use of applications which facilitate carbohydrate counting and bolus calculation. A number of applications are available to facilitate carbohydrate counting, however more recent developments include the launch of food identification software that calculates the carbohydrate content of food using photographs ( 115 , 116 ). This software can improve carbohydrate counting and HbA1c in a population of young adults, although its use is limited and it remains unvalidated in pregnancy ( 117 ).…”

Mobile game model for monitoring Malaysian food calories intake using image recognition

“…An image classification model may learn to predict whether fresh photos belong to any classes it has been trained in given enough training data (typically hundreds or thousands of images per label), and this prediction process is called inference. TensorFlow image classification API were used to conduct image classification tests with flowers [20], wildlife animals [22], airplanes and birds, horses, dogs, cats, and humans [23], and common garbage [24]. The test accuracies for the investigations ranged from 87.2% to 99%, which is almost accurate.…”