Long-Term Glucose Forecasting for Open-Source Automated Insulin Delivery Systems: A Machine Learning Study with Real-World Variability Analysis

Abstract

Glucose forecasting serves as a backbone for several healthcare applications, including real-time insulin dosing in people with diabetes and physical activity optimization. This paper presents a study on the use of machine learning (ML) and deep learning (DL) methods for predicting glucose variability (GV) in individuals with open-source automated insulin delivery systems (AID). A three-stage experimental framework is employed in this work to systematically implement and evaluate ML/DL methods on a large-scale diabetes dataset collected from individuals with open-source AID. The first stage involves data collection, the second stage involves data preparation and exploratory analysis, and the third stage involves developing, fine-tuning, and evaluating ML/DL models. The performance and resource costs of the models are evaluated alongside relative and proportional errors for 17 GV metrics. Evaluation of fine-tuned ML/DL models shows considerable accuracy in glucose forecasting and variability analysis up to 48 h in advance. The average MAE ranges from 2.50 mg/dL for long short-term memory models (LSTM) to 4.94 mg/dL for autoregressive integrated moving average (ARIMA) models, and the RMSE ranges from 3.7 mg/dL for LSTM to 7.67 mg/dL for ARIMA. Model execution time is proportional to the amount of data used for training, with long short-term memory models having the lowest execution time but the highest memory consumption compared to other models. This work successfully incorporates the use of appropriate programming frameworks, concurrency-enhancing tools, and resource and storage cost estimators to encourage the sustainable use of ML/DL in real-world AID systems.

Machine Learning and Deep Learning Models for Nocturnal High- and Low-Glucose Prediction in Adults with Type 1 Diabetes

“…In many studies, the authors focused on forecasting events such as nocturnal hypoglycemia [11][12][13][14][15][16][17]. Other studies predicted interstitial glucose values [21][22][23][24][25][26][27][28]. In this study, we proposed a different approach for glucose prediction by classifying the predicted values into three ranges.…”

“…In studies forecasting nocturnal hypoglycemia, the values of ROC-AUC exceeded 70%, indicating an acceptable sensitivity and specificity [11][12][13][14][15][16][17]. In studies predicting glucose levels, the root mean squared error varied from 0.36 to 1.95 mmol/L (6.45-35.10 mg/dL) at PH values up to 120 minutes [21][22][23][24][25][26][27][28]. In the aforementioned study by Guemes et al, which addressed the problem of classifying future glucose levels into the target and non-target ranges, the model was able to predict the quality of overnight glycemic control with reasonable accuracy (AUC-ROC = 0.7) [41].…”

“…Advanced glucose prediction models based on CGM data and ML or DL algorithms are generally considered to be promising elements for closed-loop automatic insulin delivery systems. Zafar et al successfully incorporated ML-based and DL-based methods of glucose prediction in individuals with open-source automated insulin delivery systems [28].…”

Diabetic Drug ontology mapping for individual Diabetic Person and predict Insulin Dosage on Daily Basis

“…The author predicted insulin dosage for gestational diabetes women through a logistic regression model [25]. The author designed automated insulin delivery (AID) system to predict glucose variability in individuals using long short-term memory (LSTM) and an autoregressive model [26]. Using the Kalman filter, the author developed an onboard insulin model to predict insulin in real-time [27].…”