Difference between revisions of "KHealth - Diabetes"
Latest revision as of 11:42, 11 October 2021
(mHealth to Improve Carbohydrate Counting Accuracy in Pediatric Type 1 Diabetes)
The incidence of type 1 diabetes (T1D) in children < 20 years is increasing in the U.S. with a 1.9% increase per year between 2002 and 2015. Standards of care in the pharmacologic management of youth with T1D include estimating the carbohydrate (CHO) content of foods consumed at meal and snack time and tailoring the insulin dose to match the amount of CHOs consumed (typically using an insulin to carbohydrate rate). Achieving consistent daily compliance and the ability of caregivers and children to accurately estimate CHO content represent challenges in the pediatric T1D population.
An indirect correlation between the CHO counting proficiency of the caregiver/patient and hemoglobin A1c (HbA1c) values have been demonstrated in numerous studies. Hence, in order to facilitate calculation of a insulin dose, T1D patients are currently required to approximate the portion size of a food they are eating either by simple visual estimation or by manually measuring the serving size with a measuring cup or scale (with the latter being preferred but not always feasible during day-to-day life).
kHealth Diabetes is a mobile health framework to manage type-I diabetes in pediatric patients by remote monitoring their dietary intake. The patients can log their meals into the mobile application using images or text. The image will be fed to the image recognition classifier to identify the meal. Once the meal is identified, the volume will be estimated using our algorithm and the nutritional breakdown will be calculated. Edamam, one of the largest nutritional databases in the world, has extended their collaboration to our project giving access to about nutritional information of 2 million recipes. The application will estimate the carbohydrate intake of the patients which facilitates the calculation of their insulin dosage. The first stage of the project involves testing the application with clinicians with gold standard nutrition estimation from dieticians before deploying it to the patients. The future work of this application will include personalized dietary recommendation based on patients health condition and food preference.
Prototype of nutrition management chatbot
- (i) Youtube - https://www.youtube.com/watch?v=MSjGwFVAKFk
- (ii)Wikipage - http://wiki.aiisc.ai/index.php/Nourich
OTHER RELATED PROJECTS
We have also developed a food logging system for post-surgical care for bariatric surgery patients. The patients found the application to be very useful and the average compliance towards using the application was more than 90%. For further information please visit kHealth Bariatrics.
Dr. Knight is a pediatric endocrinologist in Columbia, South Carolina and is affiliated with Prisma Health Richland Hospital. She is the clinical collaborator and the domain expert involved in this project.
Prof. Sheth is the founding director of the AI Institute at the University of South Carolina (AIISC), which has a large portfolio of translational research in healthcare (public health, augmented personalized health, epidemiology). Some of his perspectives and research reviews appear in the form of keynotes/talks in slideshare. Description (and video capture of demos) of three chatbots in development can be found at [link]. The kHealth diabetes mApps is based on our kHealth initiative for personalized digital health that has involved the development of mApps and chatbots for different medical conditions and purposes, several of which have involved patient evaluations under approved IRB. One such healthcare project where the team successfully identified personalized causes of asthma episodes is kHealth Asthma: Semantic Multisensory Mobile Approach to Personalized Asthma Care (PI Sheth, NICHDR01HD087132, ongoing). Other projects can be found at - http://wiki.aiisc.ai/index.php/Main_Page
Personal webpage - http://amit.aiisc.ai/
Revathy is a PhD student advised by Dr. Amit Sheth. She is working on food computation models, nutrition management and personalized food recommendation. With respect to this project, she focuses on designing a food logging system and extracting calorie information from food images. She has also worked on a food logging system for Bariatric surgery patients and hypertension patients, addressing the research challenges. She is also interested in personalized food recommendations based on the user's health condition and food preferences.
Dr. Pankesh Patel is a post-doctoral researcher at the AIISC. He has a strong background and experience in system design and development, he oversees the development of the application.
Linkedin - https://www.linkedin.com/in/pankeshpatel/