Incorporating Glucose Variability into Glucose Forecasting Accuracy Assessment Using the New Glucose Variability Impact Index and the Prediction Consistency Index: An LSTM Case Example DOI Open Access
Clara Mosquera-Lopez, Peter G. Jacobs

Journal of Diabetes Science and Technology, Journal Year: 2021, Volume and Issue: 16(1), P. 7 - 18

Published: Sept. 7, 2021

Background: In this work, we developed glucose forecasting algorithms trained and evaluated on a large dataset of free-living people with type 1 diabetes (T1D) using closed-loop (CL) sensor-augmented pump (SAP) therapies; demonstrate how variability impacts accuracy. We introduce the impact index (GVII) prediction consistency (GPCI) to assess accuracy algorithms. Methods: A long-short-term-memory (LSTM) neural network was designed predict up 60 minutes in future continuous measurements insulin data collected from 175 T1D (41,318 days) 75 (11,333 Tidepool Big Data Donation Dataset. LSTM compared two naïve as well Ridge linear regression random forest root-mean-square error (RMSE). Parkes grid quantified clinical Regression analysis used derive GVII GPCI. Results: The had highest best RMSE for CL 19.8 ± 3.2 33.2 5.4 mg/dL 30- 60-minute horizons, respectively. SAP 19.6 3.8 33.1 7.3 respectively; 99.6% 97.6% predictions were within zones A+B at Glucose strongly correlated (R≥0.64, P < 0.001); GPCI demonstrated means compare across datasets different variability. Conclusions: model accurate real-world dataset. should be considered when assessing indices such

Language: Английский

An overview of methods and techniques in multimodal data fusion with application to healthcare DOI
Siwar Chaabene, Amal Boudaya, Bassem Bouaziz

et al.

International Journal of Data Science and Analytics, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 10, 2025

Language: Английский

Citations

3
Recent applications of machine learning and deep learning models in the prediction, diagnosis, and management of diabetes: a comprehensive review DOI Creative Commons
Elaheh Afsaneh,

Amin Sharifdini,

Hadi Ghazzaghi

et al.

Diabetology & Metabolic Syndrome, Journal Year: 2022, Volume and Issue: 14(1)

Published: Dec. 27, 2022

Abstract Diabetes as a metabolic illness can be characterized by increased amounts of blood glucose. This abnormal increase lead to critical detriment the other organs such kidneys, eyes, heart, nerves, and vessels. Therefore, its prediction, prognosis, management are essential prevent harmful effects also recommend more useful treatments. For these goals, machine learning algorithms have found considerable attention been developed successfully. review surveys recently proposed (ML) deep (DL) models for objectives mentioned earlier. The reported results disclose that ML DL promising approaches controlling glucose diabetes. However, they should improved employed in large datasets affirm their applicability.

Language: Английский

Citations

58
Personalized Blood Glucose Prediction for Type 1 Diabetes Using Evidential Deep Learning and Meta-Learning DOI
Taiyu Zhu, Kezhi Li, Pau Herrero

et al.

IEEE Transactions on Biomedical Engineering, Journal Year: 2022, Volume and Issue: 70(1), P. 193 - 204

Published: July 1, 2022

The availability of large amounts data from continuous glucose monitoring (CGM), together with the latest advances in deep learning techniques, have opened door to a new paradigm algorithm design for personalized blood (BG) prediction type 1 diabetes (T1D) superior performance. However, there are several challenges that prevent widespread implementation algorithms actual clinical settings, including unclear confidence and limited training T1D subjects. To this end, we propose novel framework, Fast-adaptive Confident Neural Network (FCNN), meet these challenges. In particular, an attention-based recurrent neural network is used learn representations CGM input forward weighted sum hidden states evidential output layer, aiming compute BG predictions theoretically supported model confidence. model-agnostic meta-learning employed enable fast adaptation subject data. proposed framework has been validated on three datasets. dataset 12 subjects T1D, FCNN achieved root mean square error 18.64±2.60 mg/dL 31.07±3.62 30 60-minute horizons, respectively, which outperformed all considered baseline methods significant improvements. These results indicate viable effective approach predicting levels T1D. well-trained models can be implemented smartphone apps improve glycemic control by enabling proactive actions through real-time alerts.

Language: Английский

Citations

53
Artificial Intelligence and Machine Learning for Improving Glycemic Control in Diabetes: Best Practices, Pitfalls, and Opportunities DOI Creative Commons
Peter G. Jacobs, Pau Herrero, Andrea Facchinetti

et al.

IEEE Reviews in Biomedical Engineering, Journal Year: 2023, Volume and Issue: 17, P. 19 - 41

Published: Nov. 9, 2023

Artificial intelligence and machine learning are transforming many fields including medicine. In diabetes, robust biosensing technologies automated insulin delivery therapies have created a substantial opportunity to improve health. While the number of manuscripts addressing topic applying diabetes has grown in recent years, there been lack consistency methods, metrics, data used train evaluate these algorithms. This manuscript provides consensus guidelines for practitioners field best practice recommended approaches warnings about pitfalls avoid.

Language: Английский

Citations

42
Enhancing the Capabilities of Continuous Glucose Monitoring With a Predictive App DOI
Pau Herrero, Magí Andorrà,

Nils Babion

et al.

Journal of Diabetes Science and Technology, Journal Year: 2024, Volume and Issue: 18(5), P. 1014 - 1026

Published: Aug. 19, 2024

Despite abundant evidence demonstrating the benefits of continuous glucose monitoring (CGM) in diabetes management, a significant proportion people using this technology still struggle to achieve glycemic targets. To address challenge, we propose Accu-Chek

Language: Английский

Citations

11
Modeling risk of hypoglycemia during and following physical activity in people with type 1 diabetes using explainable mixed-effects machine learning DOI
Clara Mosquera-Lopez, Katrina Ramsey,

Valentina Roquemen-Echeverri

et al.

Computers in Biology and Medicine, Journal Year: 2023, Volume and Issue: 155, P. 106670 - 106670

Published: Feb. 11, 2023

Language: Английский

Citations

18
Measurement of multimodal physiological signals for stimulation detection by wearable devices DOI
Gloria Cosoli, Angelica Poli, Lorenzo Scalise

et al.

Measurement, Journal Year: 2021, Volume and Issue: 184, P. 109966 - 109966

Published: Aug. 5, 2021

Language: Английский

Citations

38
Artificial intelligence and diabetes technology: A review DOI
Thibault Gautier,

L. B. Ziegler,

Matthew S. Gerber

et al.

Metabolism, Journal Year: 2021, Volume and Issue: 124, P. 154872 - 154872

Published: Sept. 1, 2021

Language: Английский

Citations

37
Artificial Intelligence Algorithms for Treatment of Diabetes DOI Creative Commons
Mudassir Rashid, Mohammad Reza Askari, Canyu Chen

et al.

Algorithms, Journal Year: 2022, Volume and Issue: 15(9), P. 299 - 299

Published: Aug. 26, 2022

Artificial intelligence (AI) algorithms can provide actionable insights for clinical decision-making and managing chronic diseases. The treatment management of complex diseases, such as diabetes, stands to benefit from novel AI analyzing the frequent real-time streaming data occasional medical diagnostics laboratory test results reported in electronic health records (EHR). Novel are needed develop trustworthy, responsible, reliable, robust techniques that handle imperfect imbalanced EHRs inconsistencies or discrepancies with free-living self-reported information. challenges applications two problems healthcare domain were explored this work. First, we introduced designed be fair unbiased while accommodating privacy concerns predicting treatments outcomes. Then, studied innovative approach using machine learning improve automated insulin delivery systems through information wearable devices historical identify informative trends patterns data. Application examples diabetes demonstrate benefits tools informatics.

Language: Английский

Citations

20
An overview of advancements in closed-loop artificial pancreas system DOI Creative Commons
Doni Dermawan, Muhammad Abiyyu Kenichi Purbayanto

Heliyon, Journal Year: 2022, Volume and Issue: 8(11), P. e11648 - e11648

Published: Nov. 1, 2022

Type 1 diabetes (T1D) is one of the world's health problems with a prevalence 1.1 million for children and young adults under age 20. T1D problem characterized by autoimmunity destruction pancreatic cells that produce insulin. The available treatment to maintain blood glucose within desired normal range. To meet bolus basal requirements, patients may receive multiple daily injections (MDI) fast-acting long-acting insulin once or twice daily. In addition, pumps can deliver doses day without causing injection discomfort in individuals T1D. have also monitored their levels along replacement using continuous monitor (CGM). However, this CGM has some drawbacks, like sensor needs be replaced after being inserted skin seven days calibrated (for CGMs). treatments monitoring devices mentioned creating lot workloads Therefore, overcome these problems, closed-loop artificial pancreas (APD) are widely used manage patients. Closed-loop APD consists sensor, an infusion device, control algorithm. This study reviews progress systems from perspective device properties, uses, testing procedures, regulations, current market conditions.

Language: Английский

Citations

20