Electrocardiogram analysis for cardiac arrhythmia classification and prediction through self attention based auto encoder

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Март 18, 2025

Sudden cardiac arrest among young people is a recent worldwide risk, and it noticed that with arrhythmia are more susceptible to various heart diseases. Manual classification can be error-prone, certainly, there need for automation classify ECG signals predict accurately. The proposed self-attention artificial intelligence auto-encoder algorithm proved an effective strategy novel modified Kalman filter pre-processing. We achieved 24.00 SNRimp, 0.055 RMSE, 22.1 PRD% -5db, 20.4 0.0245 12 whereas 14.05 0.010 7.25 PRD%, which reduces the signal noise during pre-processing improves visibility of QRS complex R-R peaks waveform. extracted features were used in network neurons execute MIT-BIH databases using newly developed autoencoder (AE) algorithm. results compared existing models, revealing system outperforms prediction precision 99.91%, recall 99.86%, accuracy 99.71%. It confirmed self-attention-AE training promising, benefits diagnosis ECGs conditions solve real-world problems.

Язык: Английский

---

LA-ResUNet: Attention-based network for longitudinal liver tumor segmentation from CT images

Computerized Medical Imaging and Graphics, Год журнала: 2025, Номер unknown, С. 102536 - 102536

Опубликована: Март 1, 2025

Язык: Английский

---

New Machine Learning Method for Medical Image and Microarray Data Analysis for Heart Disease Classification

Deleted Journal, Год журнала: 2025, Номер unknown

Опубликована: Апрель 1, 2025

Microarray technology has become a vital tool in cardiovascular research, enabling the simultaneous analysis of thousands gene expressions. This capability provides robust foundation for heart disease classification and biomarker discovery. However, high dimensionality, noise, sparsity microarray data present significant challenges effective analysis. Gene selection, which aims to identify most relevant subset genes, is crucial preprocessing step improving accuracy, reducing computational complexity, enhancing biological interpretability. Traditional selection methods often fall short capturing complex, nonlinear interactions among limiting their effectiveness tasks. In this study, we propose novel framework that leverages deep neural networks (DNNs) optimizing using data. DNNs, known ability model patterns, are integrated with feature techniques address high-dimensional The proposed method, DeepGeneNet (DGN), combines DNN-based into unified framework, ensuring performance meaningful insights underlying mechanisms. Additionally, incorporates hyperparameter optimization innovative U-Net segmentation further enhance accuracy. These optimizations enable DGN deliver scalable results, outperforming traditional both predictive accuracy Experimental results demonstrate approach significantly improves compared other methods. By focusing on interplay between learning, work advances field genomics, providing interpretable future applications.

Язык: Английский

---

Dynamic Coherent Diffractive Imaging with Modulus Enforced Probe and Low Spatial Frequency Constraints

Sensors, Год журнала: 2025, Номер 25(7), С. 2323 - 2323

Опубликована: Апрель 6, 2025

Dynamic behavior is prevalent in biological and condensed matter systems at the nano- mesoscopic scales. Typically, we capture images as "snapshots" to demonstrate evolution of a system, coherent X-ray diffraction imaging (CDI), lensless technique, provides nanoscale resolution, allowing us clearly observe these microscopic phenomena. This paper presents new dynamic CDI method based on zone-plate optics aiming overcome limitations existing techniques fast processes by integrating spatio-temporal dual constraint with probe constraint. In this method, modulus-enforced temporal correlation sample low-frequency information are exploited combined an empty static region sample. Using achieved resolution 20 Hz spatial 13.2 nm, which were verified visualized experimental results. Further comparisons showed that reconstructed consistent ptychography reconstruction results, confirming accuracy feasibility method. work expected provide tool for materials science life sciences, promoting deeper understanding complex processes.

Язык: Английский

---

A color classification system for sunglass lenses based on YCrCb-MST hyperspectral reconstruction

Measurement, Год журнала: 2025, Номер 253, С. 117527 - 117527

Опубликована: Апрель 16, 2025

Язык: Английский

---

A faster computational frame work for dye design and screening: A goal to achieve higher ionization energy

Chemical Physics Letters, Год журнала: 2025, Номер unknown, С. 142106 - 142106

Опубликована: Апрель 1, 2025

Язык: Английский

---

Micro-LED晶圆缺陷检测的宽光谱大视场显微物镜设计

ACTA PHOTONICA SINICA, Год журнала: 2025, Номер 54(3), С. 0318003 - 0318003

Опубликована: Янв. 1, 2025

---

Fast-zoom and high-resolution sparse compound-eye camera based on dual-end collaborative optimization

Opto-Electronic Advances, Год журнала: 2025, Номер 0(0), С. 240285 - 240285

Опубликована: Янв. 1, 2025

Язык: Английский

---

Deep‐Tissue, Large‐FOV 3D NIR‐II Fluorescence Confocal Microscopy With Hundred‐Nanosecond Equivalent Pixel Dwell Time

Laser & Photonics Review, Год журнала: 2025, Номер unknown

Опубликована: Май 3, 2025

Abstract Near‐infrared II (NIR‐II, 900–1880 nm) fluorescence confocal microscopy enables in vivo imaging with high spatial resolution at large depth. Nonetheless, three dimensional (3D) requires capturing substantial pixels and prolonged laser scanning, leading to phototoxicity, exogenous probe metabolic decay, loss of information on dynamic anatomical structures. Strategies diminish duration can be considered by decreasing the actual pixel dwell time without deterioration quality. In this study, a novel approach combining NIR‐II is introduced deep learning interpolation network, which substantially decreases axial sampling frequency requirements, achieving equivalent hundred‐nanosecond 3D visualization vivo. By applying cerebral vessel (CVI) network field‐of‐view (FOV) microscopic imaging, up 16‐fold increase has been achieved scanning speed, reducing from 8 µs 500 ns. This significantly reduces laser‐induced damage biological samples, lessens need for extending metabolism probes, facilitates potential rapid biomedical applications. Benchmarking tests show CVI achieves best performance compared conventional methods both lateral cross‐sectional images.

Язык: Английский

---

Integrating artificial intelligence and holographic imaging for advanced cervical cancer diagnosis

Signal Image and Video Processing, Год журнала: 2025, Номер 19(7)

Опубликована: Май 12, 2025

Язык: Английский

---