Deep-Learning-Based Estimation of the Spatial QRS-T Angle from Reduced-Lead ECGs.

The spatial QRS-T angle is a promising health indicator for risk stratification of sudden cardiac death (SCD). Thus far, the angle is estimated solely from 12-lead electrocardiogram (ECG) systems uncomfortable for ambulatory monitoring. Methods to estimate QRS-T angles from reduced-lead ECGs registered with consumer healthcare devices would, therefore, facilitate ambulatory monitoring. (1) Objective: Develop a method to estimate spatial QRS-T angles from reduced-lead ECGs. (2) Approach: We designed a deep learning model to locate the QRS and T wave vectors necessary for computing the QRS-T angle. We implemented an original loss function to guide the model in the 3D space to search for each vector's coordinates. A gradual reduction of ECG leads from the largest publicly available dataset of clinical 12-lead ECG recordings (PTB-XL) is used for training and validation. (3) Results: The spatial QRS-T angle can be estimated from leads {I, II, aVF, V2} with sufficient accuracy (absolute mean and median errors of 11.4° and 7.3°) for detecting abnormal angles without sacrificing patient comfortability. (4) Significance: Our model could enable ambulatory monitoring of spatial QRS-T angles using patch- or textile-based ECG devices. Populations at risk of SCD, like chronic cardiac and kidney disease patients, might benefit from this technology.

Noninvasive fetal QRS detection using an echo state network and dynamic programming.

We address a classical fetal QRS detection problem from abdominal ECG recordings with a data-driven statistical machine learning approach. Our goal is to have a powerful, yet conceptually clean, solution. There are two novel key components at the heart of our approach: an echo state recurrent neural network that is trained to indicate fetal QRS complexes, and several increasingly sophisticated versions of statistics-based dynamic programming algorithms, which are derived from and rooted in probability theory. We also employ a standard technique for preprocessing and removing maternal ECG complexes from the signals, but do not take this as the main focus of this work. The proposed approach is quite generic and can be extended to other types of signals and annotations. Open-source code is provided.

Optimization and applications of echo state networks with leaky-integrator neurons.

Standard echo state networks (ESNs) are built from simple additive units with a sigmoid activation function. Here we investigate ESNs whose reservoir units are leaky integrator units. Units of this type have individual state dynamics, which can be exploited in various ways to accommodate the network to the temporal characteristics of a learning task. We present stability conditions, introduce and investigate a stochastic gradient descent method for the optimization of the global learning parameters (input and output feedback scalings, leaking rate, spectral radius) and demonstrate the usefulness of leaky-integrator ESNs for (i) learning very slow dynamic systems and replaying the learnt system at different speeds, (ii) classifying relatively slow and noisy time series (the Japanese Vowel dataset--here we obtain a zero test error rate), and (iii) recognizing strongly time-warped dynamic patterns.

Development of teleconsultations systems for e-health.

Two prototype telemedicine systems have been developed: 1) a wireless system for status assessment of cardiology patients (WSCP), 2) a system for medical image management and teleconsultations (IMTS). The former system enables the patient to record an ECG on a personal digital assistant (PDA), view it and send it via a wireless connection. The doctor on duty is then able to view the received ECG and make appropriate decisions, also to apply for consultation by sending the received ECG to the PDA of a cardiology expert. The system logs all performed operations. The hardware used in the system consists of personal computers (PCs), PDAs, analog-digital converters, ECG sensors and GPRS modems. Software consists of programs for patients, doctors on duty, cardiology experts and administration, along with a central database. The second system is intended to be used by professional doctors for management of collected images and for teleconsultations via videoconferencing in order to obtain a second opinion. The system provides an integrated environment eliminating the need to jump between many applications. By using the system, doctors are able to acquire images from analog and digital cameras, process and enhance them, as well as upload them to local or remote databases. Doctors are also able to design custom database forms. The teleconsultation part of the system supports video and audio over ISDN and TCP-IP, using both a hardware codec (Zydacron Z360) and a software codec (based on MS Netmeeting). Images are sent from one client to another using the standard protocol T.120. Images become synchronized immediately upon reception by another client.