Feasibility of Using Wearable Sensors and Artificial Intelligence for Carbohydrate Counting in Chinese Americans with Type 2 Diabetes
Brief description of study
Chinese Americans are one of the fastest growing racial/ethnic groups in the United States. Compared with non-Hispanic White, Chinese Americans have a higher incidence of type 2 diabetes (T2D) with lower body mass index, poor health status, worse glycemic control, and less compliance with dietary guidelines. With increases in the prevalence of T2D and unhealthy eating in Chinese Americans, strategies to improve dietary management in this population are critically needed. Carbohydrate counting (CC) is effective for reducing caloric intake, weight loss, and, glycemic control; however, the current approaches for CC are inaccurate, burdensome, and challenging. The cultural foods that Chinese Americans eat further complicate CC. A wearable imaging device (i.e., eButton) requires no volitional effort for food measuring, recording, or nutrient calculating, which helps to overcomes CC challenges for Chinese Americans. We will conduct a one-group pilot study (N=20) to evaluate the accuracy and acceptability of using eButton with a closed-loop feedback system for CC in Chinese Americans with T2D. The eButton will automatically record food data by taking food pictures during a meal. The recorded food picture data are processed by the artificial intelligence (AI) to determine food names, volume/portion size, and nutrient value (e.g., carbohydrate grams). We will use the eButton with a closed-loop feedback system, in which continuous glucose monitoring (CGM) will be applied to provide powerful feedback to evaluate CC, aiming to reduce the error and bias of estimation of carbohydrate grams. Aim 1 will evaluate the accuracy of CC using eButton for Chinese Americans by comparing its results with weighing food by registered dietitian nutritionist (“gold standard”) and food diary by participants. Aim 2 will examine the acceptability of using eButton for CC in Chinese Americans with T2D, which will be assessed by rate of use, surveys, and individual interviews. This study is of great importance because it targets a high-risk, rapidly growing, yet understudied minority population in the United States. If found to be feasible and acceptable, using a wearable imaging device, AI algorithms, and CGM-device enabled closed-loop control to automatically estimate carbohydrate grams will potentially also benefit other groups of adults with type 2 diabetes and aid in reducing their burden of dietary management. The proposed approach will advance dietary management in clinical practice. The findings will provide preliminary data for the subsequent R01 application for NIH/NIDDK R01s for PA-19-056 or PAS-20-160.
If you are registered as a volunteer, please log in to contact the study team/express interest in this study.