Machine learning-based detection of cardiovascular disease using ECG signals: performance vs. complexity.

Introduction: Cardiovascular disease remains a significant problem in modern society. Among non-invasive techniques, the electrocardiogram (ECG) is one of the most reliable methods for detecting cardiac abnormalities. However, ECG interpretation requires expert knowledge and it is time-consuming. Developing a novel method to detect the disease early improves the quality and efficiency of medical care. Methods: The paper presents various modern approaches for classifying cardiac diseases from ECG recordings. The first approach suggests the Poincaré representation of ECG signal and deep-learning-based image classifiers. Additionally, the raw signals were processed with the one-dimensional convolutional model while the XGBoost model was facilitated to predict based on the time-series features. Results: The Poincaré-based methods showed decent performance in predicting AF (atrial fibrillation) but not other types of arrhythmia. XGBoost model gave an acceptable performance in long-term data but had a long inference time due to highly-consuming calculations within the pre-processing phase. Finally, the 1D convolutional model, specifically the 1D ResNet, showed the best results in both studied CinC 2017 and CinC 2020 datasets, reaching the F1 score of 85% and 71%, respectively, and they were superior to the first-ranking solution of each challenge. The 1D models also presented high specificity. Additionally, our paper investigated efficiency metrics including power consumption and equivalent CO2 emissions, with one-dimensional models like 1D CNN and 1D ResNet being the most energy efficient. Model interpretation analysis showed that the DenseNet detected AF using heart rate variability while the 1D ResNet assessed the AF patterns in raw ECG signals. Discussion: Despite the under-performed results, the Poincaré diagrams are still worth studying further because of the accessibility and inexpensive procedure. In the 1D convolutional models, the residual connections are useful to keep the model simple but not decrease the performance. Our approach in power measurement and model interpretation helped understand the numerical complexity and mechanism behind the model decision.

Mitigating First-Cycle Capacity Losses in NMC811 via Lithicone Layers Grown by Molecular Layer Deposition.

Nickel-rich LiNi1-x-yMnxCoyO2 (NMC, 1 - x - y ≥ 0.8) is currently considered one of the most promising cathode materials for high-energy-density automotive lithium-ion batteries. Here, we show that capacity losses occurring in balanced NMC811||graphite cells can be mitigated by lithicone layers grown by molecular layer deposition directly onto porous NMC811 particle electrodes. Lithicone layers with a stoichiometry of LiOC0.5H0.3 as determined by elastic recoil detection analysis and a nominal thickness of 20 nm determined by ellipsometry on a flat reference substrate improve the overall NMC811||graphite cell capacity by ∼5% without negatively affecting the rate capability and long-term cycling stability.

Assessing Long-Term Cycling Stability of Single-Crystal Versus Polycrystalline Nickel-Rich NCM in Pouch Cells with 6 mAh cm-2 Electrodes.

Lithium-ion batteries based on single-crystal LiNi1- x - y Cox Mny O2 (NCM, 1-x-y ≥ 0.6) cathode materials are gaining increasing attention due to their improved structural stability resulting in superior cycle life compared to batteries based on polycrystalline NCM. However, an in-depth understanding of the less pronounced degradation mechanism of single-crystal NCM is still lacking. Here, a detailed postmortem study is presented, comparing pouch cells with single-crystal versus polycrystalline LiNi0.60 Co0.20 Mn0.20 O2 (NCM622) cathodes after 1375 dis-/charge cycles against graphite anodes. The thickness of the cation-disordered layer forming in the near-surface region of the cathode particles does not differ significantly between single-crystal and polycrystalline particles, while cracking is pronounced for polycrystalline particles, but practically absent for single-crystal particles. Transition metal dissolution as quantified by time-of-flight mass spectrometry on the surface of the cycled graphite anode is much reduced for single-crystal NCM622. Similarly, CO2 gas evolution during the first two cycles as quantified by electrochemical mass spectrometry is much reduced for single-crystal NCM622. Benefitting from these advantages, graphite/single-crystal NMC622 pouch cells are demonstrated with a cathode areal capacity of 6 mAh cm-2 with an excellent capacity retention of 83% after 3000 cycles to 4.2 V, emphasizing the potential of single-crystalline NCM622 as cathode material for next-generation lithium-ion batteries.

In Situ Control of Oxygen Vacancies in TaOx Thin Films via Plasma-Enhanced Atomic Layer Deposition for Resistive Switching Memory Applications.

The plasma-enhanced atomic layer deposition (PEALD) process using Ta(OC2H5)5 as a Ta precursor and plasma-activated hydrogen as a reactant for the deposition of TaOx films with a controllable concentration of oxygen vacancies (VO) is reported herein. The VO concentration control was achieved by varying the hydrogen volume fraction of the hydrogen-argon mixture in the plasma, allowing the control of the leakage current density in the tantalum oxide films within the range of 5 orders of magnitude compared with the Ta2O5 film grown via thermal ALD using the identical Ta precursor and H2O. Temperature-dependent current-voltage measurements combined with Poole-Frenkel emission modeling demonstrated that the bulk trap depth decreases with the increasing hydrogen volume fraction, which could be attributed to the increase of the VO concentration. The possible chemical change in the PEALD TaOx films grown under different hydrogen volume fractions was confirmed by the in situ X-ray photoelectron spectroscopy (XPS) measurements of the Ta 4f core and valence band spectra. The comparison of the XPS-measured nonstoichiometry and the secondary ion mass spectrometry analysis of the hydrogen content allowed this study to conclude that the nonstoichiometry is largely related to the formation of Ta-VO sites rather than of Ta-H sites. Such oxygen-deficient TaOx layers were studied for application as an oxygen-deficient layer in a resistance switching random access memory stack (Ta2O5/TaOx) where the actual switching occurred within the stoichiometric Ta2O5 layer. The bilayer memory stack showed reliable resistance switching up to ∼106 switching cycles, whereas the single-layer Ta2O5 memory showed only several hundred switching cycles.