Artificial Intelligence in Diabetes Treatment
Diabetes is a chronic condition that leads to high blood sugar levels. The CDC reports that, as of 2020, 34 million Americans (10.5% of the US population) have diabetes, with a steady rise in diabetes prevalence since the year 2000 [1]. As the general population and the number of diabetes patients grow, healthcare providers will need cutting-edge tools in order to improve outcomes, treatment, and prevention. In particular, academic researchers and medical practitioners are finding that artificial intelligence (AI) is an effective tool for improving diabetes care. Specific applications of AI include calculation of insulin dosage, the control of artificial pancreases, prediction of blood sugar levels, and lifestyle support for patients.
When diabetes patients self-administer insulin after meals, the process usually involves three steps. First, the patient takes a blood sugar reading. Second, the patient calculates the proper insulin dosage. Third, the patient self-injects insulin [2]. Researchers have proposed to use AI in the second step, calculating the proper dosage from blood sugar measurements, the nutritional content of recent meals, the amounts of recent doses of insulin, and other sources of data [3]. For example, one study conducted at the Imperial College in London found that an AI-based smartphone application for calculating dosages was safe and effective for type 1 diabetes patients over a six-week period [4][5].
Even with AI-assisted dosage calculations, the patient must measure and inject the dosage correctly. With recent technologies, it is possible to automate the entire process of insulin administration. For example, the artificial pancreas is a portable device that consists of an insulin pump and a blood glucose sensor. An internal AI controls how much insulin the pump should administer, given the monitor’s glucose readings [2]. Artificial pancreas research has gained plenty of attention in recent years, with researchers seeking the best AI algorithms and methods for controlling the devices [3][4]. An artificial pancreas, like an organic pancreas, should be extremely accurate and immune to patient error, compared to a manually operated insulin pump.
A more challenging problem is the prediction of blood sugar levels. In a literature review of 49 peer-reviewed studies to predict blood sugar from glucose sensor measurements and physiological data, researchers found that 76% of those studies could predict blood sugar levels up to one hour in advance [3]. Accurate predictions would identify if a patient’s treatment is working, and they might help improve insulin dosage calculations. In more extreme cases, patients could be notified to seek medical attention if the AI forecasts a sudden, dangerous fluctuation in blood sugar levels.
Perhaps the nearest-term impact of AI is on lifestyle support for diabetes patients. Most existing tools for diabetes management rely on general models that roughly characterize each patient based on data from a large group of patients. The added value of AI is to personalize diabetes management, based on an individual patient’s attributes and risk factors. For example, an AI-based application might encourage patients to adhere to their medication, exercise periodically, or visit their doctor. Wearable activity trackers, like those produced by Fitbit and Garmin, can give AI tools more accurate patient data, which will further personalize the patient’s experience. In doing so, the app could improve patient engagement with and responsiveness to their treatment [3].
In conclusion, diabetes treatment and management stand to benefit from the integration of artificial intelligence. Despite the expected benefits to personalization and patient outcomes, these data-driven methods must arrive with appropriate protections for patients. Data from an artificial pancreas, a blood sugar predictor, or a lifestyle assistant are deeply personal and private. Without technological and policy safeguards alongside these innovations, patients may be forced to sacrifice their privacy for improved medical outcomes.
References
[1] National Diabetes Statistics Report, 2020: Estimates of Diabetes and Its Burden in the United States. U.S. Centers for Disease Control and Prevention. 2020. https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf.
[2] A. Tenderich. Artificial Pancreas: What You Should Know. Healthline. March 31, 2020. https://www.healthline.com/diabetesmine/artificial-pancreas-what-you-should-know.
[3] Contreras I., and Vehi J. Artificial Intelligence for Diabetes Management and Decision Support: Literature Review. Journal of Medical Internet Research 2018; 20: 5. DOI:10.2196/10775.
[4] Rigla M., et al. Artificial Intelligence Methodologies and Their Application to Diabetes. Journal of Diabetes Science and Technology 2018; 12: 2. DOI:10.1177/1932296817710475.
[5] Reddy M., et al. Clinical Safety and Feasibility of the Advanced Bolus Calculator for Type 1 Diabetes based on Case-based Reasoning: A 6-week Nonrandomized Single-arm Pilot Study. Diabetes Technology and Therapeutics 2016; 18: 8. DOI:10.1089/dia.2015.0413.