An Integrated Secure Efficient Computing Architecture for Embedded and Remote ECG Diagnosis

Secured and efficient real-time electrocardiogram (ECG) diagnosis of patients in a distributed network would save the lives of many patients, reduce the communication cost, and, at the same time, protect the patient’s identity from adversaries. However, the present state-of-art steganography and encryption technique covers the deployment of such technology in a centralized system rather than a distributed system which consists of multiple processing layers. In this study, our major contributions are twofold. First, we propose the novel sparse steganography that is combined with the Advanced Encryption Standard (AES) to secure the transmission between local, edge, and cloud computing. The sparse steganography is based on the fact that ECG samples are unique and represent their distinctive binary identity. Our test results have demonstrated the feasibility of the proposed sparse steganography in the state of noisy conditions during the transmission from one processing layer to another. Second, the proposed system of the Internet of Medical Things (IoMT) uses an R-peak detection algorithm and a 2D-CNN inference module in the local layer to detect and classify ECG signals, whereas edge/fog and cloud computing use the 2D-CNN modules with higher classification accuracy. A dual-mode transmission mechanism is deployed in the proposed IoMT system. The open-loop transmission ensures that healthy ECG signals are reported but not transmitted thoroughly to the medical server, while the closed-loop transmission is present to enable a non-stop transmission of ECG signals for severe patients who need extra care. Moreover, this closed-loop scheme provides an opportunity to have bidirectional communication between processing layers that is used to request extra information regarding the patient and update the status of the wearable sensor. A simulation was performed consisting of 16 patients’ data among which 6 of the patients were suffering from arrhythmia. Two modes of transmission were deployed for those 16 ECG signals, where (1) the open-loop transmission proved to be 99.99% more efficient than the regular transmission, whereas (2) the closed-loop transmission with the severity of patients was not compromised in terms of both ECG signals, and patient’s diagnosis information.