Type 1 Diabetes Self-Management and Carbohydrate Counting: A Computer Vision based Approach

Patients with diabetes must be taught how to achieve glycaemic control by monitoring their glucose levels properly. They are medicated with either exogenous insulin or other drugs and encouraged to improve their diet and physical activity. Although studies have shown that planning meals and counting carbohydrates is of great importance for diabetic patients, even well trained diabetic patients find it difficult to estimate carbohydrates precisely.

The aim of the project is the design, development and evaluation of a system which will permit the automatic, near real-time recognition of the different types of foods on a plate and the estimation of their content of carbohydrates. The system will be based on the advanced analysis of colour images and will be composed of i) an Advanced Image Processing (AIP) module, including a camera for image capture, ii) a Carbohydrate Estimator (CE) and iii) a Data Base (DB). The AIP module will incorporate the entire image processing tools for acquisition, pre-processing, segmentation, feature detection, feature representation and selection, and classification. The CE will estimate the volume and the weight of the food, while the DB will contain a list of nutrients, along with the corresponding grams of carbohydrates. In a typical use scenario, the diabetic will take a picture of the incoming meal with the mobile phone camera. This image will be processed in order to estimate a set of characteristic features describing the type of nutrition and the corresponding grams of carbohydrate. In addition to dietary assessment, this information will be used to optimise the calculation of the bolus insulin dose.

The ultimate objective is to have an application running on a portable device which can be used in everyday life to support the diabetic patient during carbohydrate counting and insulin dose estimation in a precise, easy and flexible manner.