Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
| Name | Class |
|---|---|
| University of Warwick | OTHER |
Not provided
Not provided
Not provided
Not provided
Paediatric Type 1 Diabetes (T1D) patients are at greater risk for developing severe hypo and hyperglycaemic events due to poor glycaemic control and incorrect Insulin administration. To reduce the risk of adverse events, patients need to achieve the best possible glycaemic control through frequent blood glucose monitoring with finger prick or Continuous Glucose Monitoring (CGM) systems. However, several non-invasive techniques have been proposed aiming at exploiting changes in physiological parameters based on glucose levels. The overall objective of this study is to validate a deep learning algorithm to detect glycaemic events using electrocardiogram (ECG) signals collected through non-invasive device.
This observational single-arm study will enrol participants with T1D aged less than 18 years old who already use CGM device. Participants will wear an additional non-invasive wearable device, for recording physiological data (e.g. ECG, breathing waveform, 3-axis acceleration) for three days. ECG variables (e.g. heart rate variability features), respiratory rate, physical activity, posture and glycaemic measurements driven through ECG variables and other physiological signals (e.g. the frequency of hypo or hyperglycaemic events, the time spent in hypo- or hyperglycaemia and the time in range) are the main outcomes. A quality-of-life questionnaire will be administered to collect secondary outcomes. Data collected will be used to design, develop and validate the personalised and generalized classifiers based on a deep-learning artificial intelligence (AI) algorithm developed during the pilot study, able to automatically detect hypoglycaemic events by using few ECG heartbeats recorded with wearable devices.
This study is a validation study that will carry out additional tests on a larger diabetes sample population, to validate the previous promising pilot results that were based on four healthy adult subjects. Therefore, this study will provide evidence on the reliability of the deep-learning artificial intelligence algorithms investigators developed, in detecting glycaemic events in paediatric diabetic patients in free-living conditions. Additionally, this study aims to develop the generalized AI model for the automated glycaemic events detection on real-time ECG.
As per inclusion criteria, the study participants continue to use their CGM device they are already using. During their routine diabetes hospital visit, the participants are asked to wear an additional wearable device, Medtronic Zephyr BioPatch, for recording the physiological data for a period of up to three days. After receiving the training session and relevant information about the study, the participants are allowed to return home with the wearable device attached. During the hospital visit, the quality of life questionnaire for paediatric patients (PEdsQL) is submitted to recruited patients. They are asked to answer questions on how T1D affects their daily activities.
During the monitoring days, patients can continue their daily activities undisturbed, without any changes in either physical activities or diet. In this way, data gathered from free-living conditions are obtained. They should wear the sensor during the day and the night and remove it while showering. The device should be approximately charged every 12-hours. For this reason, patients were provided with two devices. While wearing the second device the one used during the day should be recharged and vice versa. Patients receive regular contact from the research team not only to check on their safety and wellbeing, but also to ensure the data collection is successful. At the end of the third day, patients should return the devices to the hospital.
Not provided
Not provided
Not provided
Not provided
Not provided
| Label | Type | Description | Intervention Names |
|---|---|---|---|
| type1diabetes patients who use CGM | Males and females diagnosed with T1D, aged less than 18 years old who are currently under the care of the Unit of Endocrinology and Diabetes of Bambino Gesù Children's Hospital, Rome, Italy and who already use continuous glucose monitoring (CGM) systems are eligible to be involved in the study. Participants will wear an additional non-invasive wearable device, Medtronic Zephyr BioPatch, for recording physiological data for three days. |
Not provided
| Measure | Description | Time Frame |
|---|---|---|
| Interval across different fiducial point | The interval across different fiducial points (millisecond) is one of the Heart Rate Variability Features (HRV) that are useful to quantify the difference in ECG signals for different glycaemic events. The glycaemic events can be determined non-invasively via ECG signals by the automated AI algorithm which are trained according to glucose measurements from the CGM. The difference in ECG signals for different glycaemic events can be quantified through the difference in the intervals across different fiducial points (five fiducial points (P.Q.R,S,T) and 9 intervals among them) calculated over three days of continued ECG signal registration. | three days |
| Slope across different fiducial points | The Slope across different fiducial points (mV/ms) is one of the Heart Rate Variability Features (HRV) that are useful to quantify the difference in ECG signals for different glycaemic events. The glycaemic events can be determined non-invasively via ECG signals by the automated AI algorithm which are trained according to glucose measurements from the CGM. The difference in ECG signals for different glycaemic events can be quantified through the difference in the slope across different fiducial points (five fiducial points (P.Q.R,S,T) and 9 intervals among them) calculated over three days of continued ECG signal registration. | three days |
| Absolute power | The absolute power (ms^2/Hz) is one of the Heart Rate Variability Features (HRV) that are useful to quantify the difference in ECG signals for different glycaemic events over three days of continued ECG signal registration.The signal energy can be determined for 5 minutes ECG excerpt within Ultra Low Frequency (ULF) (≤0.003 Hz), Very Low Frequency (VLF) (0.0033-0.04 Hz), Low Frequency (LF) (0.04-0.15 Hz) and High Frequency (HF) (0.15-0.4 Hz) | three days |
| Severe hypoglycaemic events detection | The severe hypoglycaemic events (identified by glycaemic values < 50mg/dl) will be indirectly detected non-invasively via ECG signals by the automated AI algorithm which are trained according to glucose measurements from the CGM. The deep-learning algorithm is able to automatically detect the severe hypoglycaemic events through the assessment of the ECG variables (heart rate (BPM), physical activity and posture (lying, standing, walking, running) and HRV features over three days of continued ECG and CGM signals registration. |
| Measure | Description | Time Frame |
|---|---|---|
| Health related quality of life | The Health related quality of life for pediatric patients is assessed through the Pediatric Quality of Life Inventory (PedsQL) questionnaire. The Pediatric Quality of Life Inventory (PedsQL) is a 23-item generic health status instrument with parent and child forms that assesses five domains of health (physical functioning, emotional functioning, psychosocial functioning, social functioning, and school functioning) in children and adolescents ages 2 to 18. the minimum and maximum values: 0, 100 higher scores mean a better outcome |
Not provided
Inclusion Criteria:
Exclusion Criteria:
Not provided
Not provided
Males and females diagnosed with T1D, aged less than 18 years old who are currently under the care of the Unit of Endocrinology and Diabetes of Bambino Gesù Children's Hospital, Rome, Italy and who already use continuous glucose monitoring (CGM) systems are eligible to be involved in the study.
| Name | Role | Phone | Extension | |
|---|---|---|---|---|
| Martina Andellini, PhDstudent | Contact | +393357625261 | martina.andellini@opbg.net |
| Name | Affiliation | Role |
|---|---|---|
| Matteo Ritrovato, PhD | Bambino Gesù Children's Hospital | Principal Investigator |
| Facility | Status | City | State | ZIP | Country | Contacts |
|---|---|---|---|---|---|---|
| Bambino Gesù Children's Hospital | Recruiting | Rome | 00165 | Italy |
| PubMed Identifier | Type | Citation | Retractions |
|---|---|---|---|
| 31932608 | Background | Porumb M, Stranges S, Pescape A, Pecchia L. Precision Medicine and Artificial Intelligence: A Pilot Study on Deep Learning for Hypoglycemic Events Detection based on ECG. Sci Rep. 2020 Jan 13;10(1):170. doi: 10.1038/s41598-019-56927-5. | |
| Background | Porumb M, Griffen C, Hattersley J, Pecchia L. Nocturnal low glucose detection in healthy elderly from one-lead ECG using convolutional denoising autoencoders. Biomedical Signal Processing and Control. 2020;62:102054. | ||
| 36761922 |
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
Not provided
| three days |
| Hypoglycaemic events detection | The hypoglycaemic events (identified by glycaemic values between 50mg/dl and 70mg/dl) will be indirectly detected non-invasively via ECG signals by the automated AI algorithm which are trained according to glucose measurements from the CGM. The deep-learning algorithm is able to automatically detect the hypoglycaemic events through the assessment of the ECG variables (heart rate (BPM), physical activity and posture (lying, standing, walking, running) and HRV features over three days of continued ECG and CGM signals registration. | three days |
| Hyperglycaemic events detection | The hyperglycaemic events (identified by glycaemic values between 180mg/dl and 240mg/dl) will be indirectly detected non-invasively via ECG signals by the automated AI algorithm which are trained according to glucose measurements from the CGM. The deep-learning algorithm is able to automatically detect the hyperglycaemic events through the assessment of the ECG variables (heart rate (BPM), physical activity and posture (lying, standing, walking, running) and HRV features over three days of continued ECG and CGM signals registration. | three days |
| Severe hyperglycaemic events detection | The severe hyperglycaemic events (identified by glycaemic values > 240mg/dl) will be indirectly detected non-invasively via ECG signals by the automated AI algorithm which are trained according to glucose measurements from the CGM. The deep-learning algorithm is able to automatically detect the severe hyperglycaemic events through the assessment of the ECG variables (heart rate (BPM), physical activity and posture (lying, standing, walking, running) and HRV features over three days of continued ECG and CGM signals registration. | three days |
| one month |
| Glycated haemoglobin level (HbA1c) | Glycated haemoglobin level (percent) is a measure of the previous three-months average blood sugar level. | three months |
| Glycaemic variability (GV) | Glycaemic variability (mg/dl) is a measure of the fluctuations of glucose over three days. | three days |
| Frequency of severe hypoglycaemic events | the frequency of severe hypoglycaemic events (Frequency (percent) is measured as the ratio between the number of severe hypoglycaemic events (glucose level < 50 mg/dl) and the total number of glucose measurements over three days. | three days |
| Frequency of hypoglycaemic events | The frequency of hypoglycaemic events (Frequency (percent) is measured as the ratio between the number of hypoglycaemic events (50 mg/dl < glucose level < 70 mg/dl) and the total number of glucose measurements over three days. | three days |
| Frequency of hyperglycaemic events | The frequency of hyperglycaemic events (Frequency (percent)) is measured as the ratio of the number of hyperglycaemic events (180 mg/dl < glucose level < 240 mg/dl) and the total number of glucose measurements over three days. | three days |
| Frequency of severe hyperglycaemic events | The frequency of severe hyperglycaemic events (Frequency (percent) is measured as the ratio between the number of severe hyperglycaemic events (glucose level > 240 mg/dl) and the total number of glucose measurements over three days. | three days |
| Time in range | Time in Range (percent) is the percentage of time that a person spends with their blood glucose levels between 70 mg/dl and 180 mg/dl. | three days |
| Time in severe hypoglycaemia | Time in severe hypoglycaemia (percent) is the percentage of time that a person spends with their blood glucose levels less than 50 mg/dl. | three days |
| Time in hypoglycaemia | Time in in hypoglycaemia (percent) is the percentage of time that a person spends with their blood glucose levels between 50 mg/dl and 70 mg/dl. | three days |
| Time in hyperglycaemia | Time in hyperglycaemia (percent) is the percentage of time that a person spends with their blood glucose levels between 180 mg/dl and 240 mg/dl. | three days |
| Time in severe hyperglycaemia | Time in severe hyperglycaemia (percent) is the percentage of time that a person spends with their blood glucose levels more than 240 mg/dl. | three days |
| Derived |
| Andellini M, Haleem S, Angelini M, Ritrovato M, Schiaffini R, Iadanza E, Pecchia L. Artificial intelligence for non-invasive glycaemic-events detection via ECG in a paediatric population: study protocol. Health Technol (Berl). 2023;13(1):145-154. doi: 10.1007/s12553-022-00719-x. Epub 2023 Jan 23. |