Integrated Sensors and Software for Dietary Intake Assessment

Abstract

Dietary intake information is valuable for understanding the relationship between diet and health, and it is becoming increasingly important as obesity rates soar. However, obtaining valid measurements of dietary intake for free-living persons is one of the most challenging problems of nutrient studies. Traditional paper-based dietary assessment methods have limitations due to bias, user burden, and cost, so improved electronic methods are needed to address important hypotheses related to diet and health. Among various approaches, image-based systems are emerging and becoming more widely used for dietary assessment with the increased popularity of cameras and algorithms. They are especially useful for food volume estimation, which is the largest source of error in calculating dietary intake. In the thesis, the Diet Data Recorder System (DDRS) is described, where integrated sensors and software are used for the objective measurement of dietary intake for free-living persons. The DDRS has a client-server architecture, which can efficiently collect, process, and store dietary data. The client consists of a mobile sensor package and software for data collection, while the server runs the algorithms for image processing, the database, and the interface for data storage and management. The sensor package consists of a Mobile Structured Light System (MSLS) that acquires structured light data for food volume estimation. The hardware design for the MSLS integrates a smartphone and a laser attachment. The laser attachment projects patterns onto the food, and the phone's camera acquires video clips that include the laser patterns. After presenting the entire DDRS design, data analysis that estimates the food volume is emphasized. Algorithms that extract food regions in the video, identify the laser pattern in the video frames, localize the intersection points, and generate depth maps for the corresponding images are applied. The estimated food volume, together with direct entries of food questionnaires and voice recordings, could provide dietitians with more complete food descriptions and more accurate food portion sizes. The results prove the feasibility of the proposed approach for assessing dietary intake with the DDRS. Future work will focus on automating and integrating the entire processing pipeline.