skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


  • NSF Public Access
  • Search Results
  • A machine-learning approach for long-term prediction of experimental cardiac action potential time series using an autoencoder and echo state networks
Title: A machine-learning approach for long-term prediction of experimental cardiac action potential time series using an autoencoder and echo state networks
Computational modeling and experimental/clinical prediction of the complex signals during cardiac arrhythmias have the potential to lead to new approaches for prevention and treatment. Machine-learning (ML) and deep-learning approaches can be used for time-series forecasting and have recently been applied to cardiac electrophysiology. While the high spatiotemporal nonlinearity of cardiac electrical dynamics has hindered application of these approaches, the fact that cardiac voltage time series are not random suggests that reliable and efficient ML methods have the potential to predict future action potentials. This work introduces and evaluates an integrated architecture in which a long short-term memory autoencoder (AE) is integrated into the echo state network (ESN) framework. In this approach, the AE learns a compressed representation of the input nonlinear time series. Then, the trained encoder serves as a feature-extraction component, feeding the learned features into the recurrent ESN reservoir. The proposed AE-ESN approach is evaluated using synthetic and experimental voltage time series from cardiac cells, which exhibit nonlinear and chaotic behavior. Compared to the baseline and physics-informed ESN approaches, the AE-ESN yields mean absolute errors in predicted voltage 6–14 times smaller when forecasting approximately 20 future action potentials for the datasets considered. The AE-ESN also demonstrates less sensitivity to algorithmic parameter settings. Furthermore, the representation provided by the feature-extraction component removes the requirement in previous work for explicitly introducing external stimulus currents, which may not be easily extracted from real-world datasets, as additional time series, thereby making the AE-ESN easier to apply to clinical data.  more » « less
Award ID(s):
1762553 2011280
PAR ID:
10415526
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
Chaos: An Interdisciplinary Journal of Nonlinear Science
Volume:
32
Issue:
6
ISSN:
1054-1500
Page Range / eLocation ID:
063117
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; ; ; (, Frontiers in Physiology)
    The electrical signals triggering the heart's contraction are governed by non-linear processes that can produce complex irregular activity, especially during or preceding the onset of cardiac arrhythmias. Forecasts of cardiac voltage time series in such conditions could allow new opportunities for intervention and control but would require efficient computation of highly accurate predictions. Although machine-learning (ML) approaches hold promise for delivering such results, non-linear time-series forecasting poses significant challenges. In this manuscript, we study the performance of two recurrent neural network (RNN) approaches along with echo state networks (ESNs) from the reservoir computing (RC) paradigm in predicting cardiac voltage data in terms of accuracy, efficiency, and robustness. We show that these ML time-series prediction methods can forecast synthetic and experimental cardiac action potentials for at least 15–20 beats with a high degree of accuracy, with ESNs typically two orders of magnitude faster than RNN approaches for the same network size. 
    more » « less
  2. ; ; (, 2022 IEEE 19th Annual Workshop on Microelectronics and Electron Devices (WMED), 2022)
    Identifying the source of integrated circuit (IC) degradation and being able to track its degradation via its electrical characteristics (e.g. the Voltage Transfer Characteristics, VTC, of an inverter) is very useful in failure analysis. This is because the electrical measurement is non-destructive, low-cost, and rapid. However, the extraction of defects from electrical characteristics requires significant domain expertise. To reduce or even obviate the need for domain expertise so that the process can be automatic for various circuits, one may use manifold learning. As a type of machine learning (ML), manifold learning also requires a large amount of accurate training data. To obtain enough defect training data, which is almost impossible from experiments, one may use SPICE simulation. Based on our previous work of using AutoEncoder (AE) to perform SPICE-augmented ML to extract the pMOS and nMOS source contact resistances from the inverter VTC, in this paper, we compare the efficacy of using another 6 types of manifold learning. They are used to predict the experimental result and it is found that most of them have reasonable performance although the AE is still the best (R2=0.9). However, when including also the variation of PMOS width (as a weak perturbation to the data), algorithms such as Locally Linear Embedding (LLE) are found to perform better than AE (R2=0.72) with LLE (R2=0.83) being the best. Therefore, multiple manifold learnings are suggested to be used in parallel in real production to enhance accuracy. 
    more » « less
  3. ; ; ; ; ; ; ; ; (, PLOS ONE)
    Tolkacheva, Elena G (Ed.)
    Recent advances in human induced pluripotent stem cell (hiPSC)-derived cardiac microtissues provide a unique opportunity for cardiotoxic assessment of pharmaceutical and environmental compounds. Here, we developed a series of automated data processing algorithms to assess changes in action potential (AP) properties for cardiotoxicity testing in 3D engineered cardiac microtissues generated from hiPSC-derived cardiomyocytes (hiPSC-CMs). Purified hiPSC-CMs were mixed with 5–25% human cardiac fibroblasts (hCFs) under scaffold-free conditions and allowed to self-assemble into 3D spherical microtissues in 35-microwell agarose gels. Optical mapping was performed to quantify electrophysiological changes. To increase throughput, AP traces from 4x4 cardiac microtissues were simultaneously acquired with a voltage sensitive dye and a CMOS camera. Individual microtissues showing APs were identified using automated thresholding after Fourier transforming traces. An asymmetric least squares method was used to correct non-uniform background and baseline drift, and the fluorescence was normalized (ΔF/F0). Bilateral filtering was applied to preserve the sharpness of the AP upstroke. AP shape changes under selective ion channel block were characterized using AP metrics including stimulation delay, rise time of AP upstroke, APD30, APD50, APD80, APDmxr(maximum rate change of repolarization), and AP triangulation (APDtri= APDmxr−APD50). We also characterized changes in AP metrics under various ion channel block conditions with multi-class logistic regression and feature extraction using principal component analysis of human AP computer simulations. Simulation results were validated experimentally with selective pharmacological ion channel blockers. In conclusion, this simple and robust automated data analysis pipeline for evaluating key AP metrics provides an excellentin vitrocardiotoxicity testing platform for a wide range of environmental and pharmaceutical compounds. 
    more » « less
  4. ; ; ; ; ; ; ; ; ; ; et al (, Frontiers in Physiology)
    Optical techniques for recording and manipulating cellular electrophysiology have advanced rapidly in just a few decades. These developments allow for the analysis of cardiac cellular dynamics at multiple scales while largely overcoming the drawbacks associated with the use of electrodes. The recent advent of optogenetics opens up new possibilities for regional and tissue-level electrophysiological control and hold promise for future novel clinical applications. This article, which emerged from the international NOTICE workshop in 2018 1 , reviews the state-of-the-art optical techniques used for cardiac electrophysiological research and the underlying biophysics. The design and performance of optical reporters and optogenetic actuators are reviewed along with limitations of current probes. The physics of light interaction with cardiac tissue is detailed and associated challenges with the use of optical sensors and actuators are presented. Case studies include the use of fluorescence recovery after photobleaching and super-resolution microscopy to explore the micro-structure of cardiac cells and a review of two photon and light sheet technologies applied to cardiac tissue. The emergence of cardiac optogenetics is reviewed and the current work exploring the potential clinical use of optogenetics is also described. Approaches which combine optogenetic manipulation and optical voltage measurement are discussed, in terms of platforms that allow real-time manipulation of whole heart electrophysiology in open and closed-loop systems to study optimal ways to terminate spiral arrhythmias. The design and operation of optics-based approaches that allow high-throughput cardiac electrophysiological assays is presented. Finally, emerging techniques of photo-acoustic imaging and stress sensors are described along with strategies for future development and establishment of these techniques in mainstream electrophysiological research. 
    more » « less
  5. ; ; ; ; (, American Society of Mechanical Engineers)
    Abstract Real-time thermal monitoring and regulation are critical to the mitigation of thermal runaways and device failures in two-phase cooling systems. Compared to conventional approaches that rely on the Joule effect, thermal gradient or transverse thermoelectric effect, acoustic emission (AE)-based remote sensing is more promising for robust and non-intrusive thermal monitoring. Nevertheless, due to the high stochasticity and noise of acoustic signals, existing implementations of AE in thermal systems have been limited to qualitative state monitoring. In this paper, we present a technology for real-time heat flux quantification during two-phase cooling by coupling acoustic sensing using hydrophones and condenser microphones and regression-based machine learning frameworks. These frameworks integrate a fast Fourier transform feature extraction algorithm with regressors, i.e., Gaussian process regressor and multilayer perceptron regressor for heat flux predictions. The acoustic signals and heat fluxes are collected from pool boiling tests under transient heat loads. It is shown that both hydrophone and condenser microphone signals are successful in predicting heat flux. Multiple models are trained and compared some using only one form of acoustic data while others combine both acoustic types (i.e., hydrophone and microphone) in fusion ML models (i.e., early, joint, late). The models using only hydrophone data are shown to perform better than the models using only microphone data. Also, some forms of fusion are shown to have better performance than either of the single input data type models. This AE-ML technology is demonstrated for accurate heatflux quantification. As such, this work will not only lead to a light, low-cost, and non-contact thermal measurement technology but also a new perspective for the physical explanation of bubble dynamics during boiling. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email