Deep learning framework for detection of hypoglycemic episodes in children with type 1 diabetes

San, Phyo Phyo; Ling, Sai Ho; Nguyen, Hung T.

doi:10.1109/embc.2016.7591483

Cited by 37 publications

(35 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Studies co-authored by Prof. Hung T. Nguyen 41,42,44,53,54,75,76 involved nocturnal hypoglycemia detection in 15 type 1 diabetic children using different machine learning techniques (extreme learning 41 , hybrid swarm optimization 53 , neural networks 54 , genetic algorithms 44 , and a few others), using as inputs different ECG parameters computed from 5 or 10 minutes ECG excerpts, and achieving promissing sensitivity and specificity. For example, the more recent studies, proposed models based on a neural logic approach 76 , obtaining 79.07% sensitivity and 53.64% specificity, deep belief network approach 77 , achieving 80% sensitivity and 50% specificity, models based on extreme learning approach, obtaining 78% sensitivity and 60% specificity. As we already emphasised, direct performance comparison with those studies is not viable as the model we proposed is person-specific, which, in our opinion, explains why the proposed method outperforms the results achieved in previous studies.…”

Section: Discussionmentioning

confidence: 99%

Precision Medicine and Artificial Intelligence: A Pilot Study on Deep Learning for Hypoglycemic Events Detection based on ECG

Porumb

Stranges

Pescapè

et al. 2020

Sci Rep

142

View full text Add to dashboard Cite

Tracking the fluctuations in blood glucose levels is important for healthy subjects and crucial diabetic patients. Tight glucose monitoring reduces the risk of hypoglycemia, which can result in a series of complications, especially in diabetic patients, such as confusion, irritability, seizure and can even be fatal in specific conditions. Hypoglycemia affects the electrophysiology of the heart. However, due to strong inter-subject heterogeneity, previous studies based on a cohort of subjects failed to deploy electrocardiogram (ECG)-based hypoglycemic detection systems reliably. The current study used personalised medicine approach and Artificial Intelligence (AI) to automatically detect nocturnal hypoglycemia using a few heartbeats of raw ECG signal recorded with non-invasive, wearable devices, in healthy individuals, monitored 24 hours for 14 consecutive days. Additionally, we present a visualisation method enabling clinicians to visualise which part of the ECG signal (e.g., T-wave, ST-interval) is significantly associated with the hypoglycemic event in each subject, overcoming the intelligibility problem of deep-learning methods. These results advance the feasibility of a real-time, non-invasive hypoglycemia alarming system using short excerpts of ECG signal.Tracking the fluctuations in the blood glucose level is relevant for both healthy individuals and diabetic patients. High glucose levels (hyperglycemia) result in long-term complications and can damage the kidneys, nerves, blood vessels in the eye and can bring many other complications 14 . Low blood glucose levels (hypoglycemia) may result in acute short-term alterations of health status such as confusion, irritability, palpitations, feeling shaky and sweaty and can even result in severe loss of attention, coma or death 15 . In fact, hypoglycemia can be particularly dangerous during specific activities requiring great attention (e.g. while driving, performing complicated surgeries). Thus, technologies for non-invasive, continuous monitoring of glucose concentration aiming at early-detecting hypoglycemic events are highly required.The most diffuse methods for blood glucose testing consist of analysing a drop of blood resulted from a finger prick. However, this method does not allow continuous monitoring, is invasive, cumbersome, expensive, and it has been demonstrated that it affects patient compliance with the glucose measurements 16 . As an alternative, Continuous Glucose Monitoring Devices (CGMs) can infer the blood glucose levels in real-time based on the glucose in the interstitial fluid. These devices significantly empowered diabetic patients, but still, they present some limitations that make them unattractive for pre-diabetic patients and diabetics. Commercially available CGMs can be worn for a limited number of days, usually between 7 and 14 days. Most of the CGMs require finger prick calibration; some studies reported that the reliability of CGMs is limited 17-19 during low blood glucose level events and they sample from the interstitial fluid whi...

show abstract

Section: Discussionmentioning

confidence: 99%

Precision Medicine and Artificial Intelligence: A Pilot Study on Deep Learning for Hypoglycemic Events Detection based on ECG

Porumb

Stranges

Pescapè

et al. 2020

Sci Rep

142

View full text Add to dashboard Cite

show abstract

“…For T1DM diagnosed patients, the use of Continuous Glucose Monitoring (CGM) devices in combination with machine learning techniques has been widely used for predicting near future glycemic events. The research study in [ 11 ] used a Deep Believe Network (DBN) model and Electrocardiogram (ECG) signal to detect the natural occurrence of nocturnal hypoglycemia, using 15 children with T1DM who were monitored for 10 h overnight at the Princess Margaret Hospital for Children in Perth, Western Australia. Bertachi et al [ 12 ] also investigated the feasibility of a machine-learning-based prediction model to anticipate Nocturnal Hypoglycemia (NH) in T1DM patients, using Continuous Glucose Monitoring (CGM) devices and physical activity trackers under free-living conditions at home.…”

Section: Related Workmentioning

confidence: 99%

Deep Physiological Model for Blood Glucose Prediction in T1DM Patients

Muñoz-Organero

2020

Sensors

View full text Add to dashboard Cite

Accurate estimations for the near future levels of blood glucose are crucial for Type 1 Diabetes Mellitus (T1DM) patients in order to be able to react on time and avoid hypo and hyper-glycemic episodes. Accurate predictions for blood glucose are the base for control algorithms in glucose regulating systems such as the artificial pancreas. Numerous research studies have already been conducted in order to provide predictions for blood glucose levels with particularities in the input signals and underlying models used. These models can be categorized into two major families: those based on tuning glucose physiological-metabolic models and those based on learning glucose evolution patterns based on machine learning techniques. This paper reviews the state of the art in blood glucose predictions for T1DM patients and proposes, implements, validates and compares a new hybrid model that decomposes a deep machine learning model in order to mimic the metabolic behavior of physiological blood glucose methods. The differential equations for carbohydrate and insulin absorption in physiological models are modeled using a Recurrent Neural Network (RNN) implemented using Long Short-Term Memory (LSTM) cells. The results show Root Mean Square Error (RMSE) values under 5 mg/dL for simulated patients and under 10 mg/dL for real patients.

show abstract

“…Regarding the BG data, various preprocessing approaches had been used including differencing (derivative) BG values [27,28], CGM data reconstruction, or smoothing using different methods such as spline interpolation [29-33], a rough feature elimination, such as fast separability and correlation analysis algorithm [28,29], representing BG temporal change information [34], feature selection and feature ranking [35], filtering using Pearson’s correlation coefficient (PCC) and the t test, and the wrapper approach using greedy backward elimination [33]. The other physiological parameters (heart rate, ECG, skin impedance, and others) had been preprocessed using different methods such as normalization [36-38], feature extraction and selection [39,40], feature extraction using fast Fourier transform (FFT) [41], unsupervised restricted Boltzmann machine–based feature representation [42], filtering techniques such as Infinite impulse response high pass filter [41,43], correlation analysis [44-46], and transformation of frequency domain into time domain (FFT) [47].…”

Section: Resultsmentioning

confidence: 99%

Data-Driven Blood Glucose Pattern Classification and Anomalies Detection: Machine-Learning Applications in Type 1 Diabetes

Woldaregay¹,

Årsand²,

Botsis³

et al. 2019

J Med Internet Res

112

View full text Add to dashboard Cite

Background Diabetes mellitus is a chronic metabolic disorder that results in abnormal blood glucose (BG) regulations. The BG level is preferably maintained close to normality through self-management practices, which involves actively tracking BG levels and taking proper actions including adjusting diet and insulin medications. BG anomalies could be defined as any undesirable reading because of either a precisely known reason (normal cause variation) or an unknown reason (special cause variation) to the patient. Recently, machine-learning applications have been widely introduced within diabetes research in general and BG anomaly detection in particular. However, irrespective of their expanding and increasing popularity, there is a lack of up-to-date reviews that materialize the current trends in modeling options and strategies for BG anomaly classification and detection in people with diabetes. Objective This review aimed to identify, assess, and analyze the state-of-the-art machine-learning strategies and their hybrid systems focusing on BG anomaly classification and detection including glycemic variability (GV), hyperglycemia, and hypoglycemia in type 1 diabetes within the context of personalized decision support systems and BG alarm events applications, which are important constituents for optimal diabetes self-management. Methods A rigorous literature search was conducted between September 1 and October 1, 2017, and October 15 and November 5, 2018, through various Web-based databases. Peer-reviewed journals and articles were considered. Information from the selected literature was extracted based on predefined categories, which were based on previous research and further elaborated through brainstorming. Results The initial results were vetted using the title, abstract, and keywords and retrieved 496 papers. After a thorough assessment and screening, 47 articles remained, which were critically analyzed. The interrater agreement was measured using a Cohen kappa test, and disagreements were resolved through discussion. The state-of-the-art classes of machine learning have been developed and tested up to the task and achieved promising performance including artificial neural network, support vector machine, decision tree, genetic algorithm, Gaussian process regression, Bayesian neural network, deep belief network, and others. Conclusions Despite the complexity of BG dynamics, there are many attempts to capture hypoglycemia and hyperglycemia incidences and the extent of an individual’s GV using different approaches. Recently, the advancement of diabetes technologies and continuous accumulation of self-collected health data have paved the way for popularity of machine learning in these tasks. According to the review, most of the identified studies used a theoretical threshold, which suffers from inter- and intrapatient variation. Therefore, future studies should consider the differenc...

show abstract

Deep learning framework for detection of hypoglycemic episodes in children with type 1 diabetes

Cited by 37 publications

References 19 publications

Precision Medicine and Artificial Intelligence: A Pilot Study on Deep Learning for Hypoglycemic Events Detection based on ECG

Precision Medicine and Artificial Intelligence: A Pilot Study on Deep Learning for Hypoglycemic Events Detection based on ECG

Deep Physiological Model for Blood Glucose Prediction in T1DM Patients

Data-Driven Blood Glucose Pattern Classification and Anomalies Detection: Machine-Learning Applications in Type 1 Diabetes

Contact Info

Product

Resources

About