Security Mechanisms and Communication Strategies for the Adaptive Partition of Remote Electrocardiogram (ECG) Diagnosis Between Wearable Sensor Net and Cloud
Presenter Type
UNO Graduate Student (Doctoral)
Major/Field of Study
<--Please Select Department-->
Other
Computer Engineering
Advisor Information
Dr. Dongming Peng
Location
CEC RM #127
Presentation Type
Oral Presentation
Start Date
22-3-2024 10:30 AM
End Date
22-3-2024 11:45 AM
Abstract
Smart ECG healthcare system consists of multiple processing layers: wearable sensors, edge devices, and cloud servers that form a distributed network. In this network, wearable ECG sensors are mostly dependent on the cloud or other powerful devices for remote diagnosis and other computations because of their limited resources. All-time dependency on the cloud may slow down the advancement of wearable devices in many areas and would directly affect the versatility of wearable sensors. These areas include special diagnosis algorithms and efficient communication strategies for resource-constrained devices. In this work, we propose a novel approach to partitioning the ECG diagnosis algorithm among wearable devices, edge devices, and medical servers so that even wearable devices can play a vital role to diagnose the patient. In this way, wearable ECG sensors can present their pre-diagnosis result to medical servers, and experts and even warn the patient themselves. In addition, we propose a novel communication strategy: open-loop and closed-loop switch mode algorithms for wearable devices and edge devices. Open-loop switch mode is initiated in less severe conditions that conserve battery power whereas closed-loop switch mode is initiated in severe conditions with a feedback channel. Along with the partition of ECG diagnosis and communication strategies we have addressed the need for a secure platform to document ECG medical sessions and other information by proposing a novel Medical Virtual Chain (MVC). The proposed MVC contributes to the authorization and authentication of the incoming processing layer in the network via entrance application and layer signature. For data confidentiality and patients’ anonymity, the combination of multilayer encryption and sparse embedding is proposed. The robustness of the sparse embedding is validated by applying AWGN to the channel. The embedded patient data and original ECG signal are recovered even at low SNR conditions. The proposed diagnosis algorithms for the wearable device include an R-peak detection algorithm and a 2D-CNN inference module. To demonstrate the feasibility of these algorithms we applied them to the abnormal ECG signals using MATLAB. We obtained satisfactory results from multiple diagnosis experiments that used different sets of ECG datasets. The accuracy of 2D-CNN classification ranged from 91.11 % to 99.3%.
Security Mechanisms and Communication Strategies for the Adaptive Partition of Remote Electrocardiogram (ECG) Diagnosis Between Wearable Sensor Net and Cloud
CEC RM #127
Smart ECG healthcare system consists of multiple processing layers: wearable sensors, edge devices, and cloud servers that form a distributed network. In this network, wearable ECG sensors are mostly dependent on the cloud or other powerful devices for remote diagnosis and other computations because of their limited resources. All-time dependency on the cloud may slow down the advancement of wearable devices in many areas and would directly affect the versatility of wearable sensors. These areas include special diagnosis algorithms and efficient communication strategies for resource-constrained devices. In this work, we propose a novel approach to partitioning the ECG diagnosis algorithm among wearable devices, edge devices, and medical servers so that even wearable devices can play a vital role to diagnose the patient. In this way, wearable ECG sensors can present their pre-diagnosis result to medical servers, and experts and even warn the patient themselves. In addition, we propose a novel communication strategy: open-loop and closed-loop switch mode algorithms for wearable devices and edge devices. Open-loop switch mode is initiated in less severe conditions that conserve battery power whereas closed-loop switch mode is initiated in severe conditions with a feedback channel. Along with the partition of ECG diagnosis and communication strategies we have addressed the need for a secure platform to document ECG medical sessions and other information by proposing a novel Medical Virtual Chain (MVC). The proposed MVC contributes to the authorization and authentication of the incoming processing layer in the network via entrance application and layer signature. For data confidentiality and patients’ anonymity, the combination of multilayer encryption and sparse embedding is proposed. The robustness of the sparse embedding is validated by applying AWGN to the channel. The embedded patient data and original ECG signal are recovered even at low SNR conditions. The proposed diagnosis algorithms for the wearable device include an R-peak detection algorithm and a 2D-CNN inference module. To demonstrate the feasibility of these algorithms we applied them to the abnormal ECG signals using MATLAB. We obtained satisfactory results from multiple diagnosis experiments that used different sets of ECG datasets. The accuracy of 2D-CNN classification ranged from 91.11 % to 99.3%.