Carbohydrate Estimation Supported by the GoCARB System

The standard method for determining the carbohydrate content of a meal in patients with diabetes mellitus is the weighing of individual foods. However, in daily life, the weighing is not practical at all times. Inaccurate estimation of meal's CHO content, leads to wrong insulin doses and consequently to poor postprandial glucose control. Fact is that even well trained diabetic individuals find it difficult to estimate CHO precisely and that especially meals served on a plate are prone to false estimations underlining an emergent need for novel approaches to CHO estimation. GoCarb is a computer vision-based system for calculating the carbohydrate content of meals. In a typical scenario, the user places a credit card-sized reference object next to the meal and acquires two images using his/her smartphone. A series of computer vision modules follows: the plate is detected and the different food items on the plate are automatically segmented and recognized, while their 3D shape is reconstructed. On the basis of the shape, the segmentation results and the reference card, the volume of each item is then estimated. The CHO content is calculated by combining the food types with its volumes, and by using the USDA nutritional database. Finally, the results are displayed to the user. A preclinical study using the GoCarb system indicates that the system is able to estimate the meal's CHO content with higher accuracy than individuals with T1D. Furthermore, the feedback gathered by the participants showed that the system is easy to use even for non-smartphone users. The aim of this randomized, cross-over pilot study is to investigate the benefits of an automated determination of the carbohydrate content of meals on glycemic control in subjects with type 1 diabetes mellitus with sensor-augmented insulin pump therapy.