Zhongrui Bai

Shanghai Jiao Tong University

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Lidong Du

Chinese Academy of Sciences

Xianxiang Chen

State Key Laboratory of Food Science and Technology

Zhen Fang

University of Chinese Academy of Sciences

Fanglin Geng

Aerospace Information Research Institute

Xianxiang Chen

Aerospace Information Research Institute

Zhen Fang

University of Technology Sydney

Zhengling He

Chinese Academy of Sciences

Zhenfeng Li

Chinese Academy of Sciences

Yicheng Yao

Chinese Academy of Sciences

Peng Wang

Chinese Academy of Sciences

Cooperative Institutions

Chinese Academy of Sciences

University of Chinese Academy of Sciences

Aerospace Information Research Institute

State Key Laboratory of Transducer Technology

Chinese Academy of Medical Sciences & Peking Union Medical College

Institute of Electronics

Zhejiang University

Fudan University

Shanghai Jiao Tong University

Beijing Institute of Technology

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Predicting Neurological Outcomes for Cardiac Arrest Patients from Longitudinal EEG Based on Short-Time Fourier Transform and 3-D Deep Residual Network

Computing in cardiology (2023)

Pan Xia Dongfang Zhao Yicheng Yao Zhongrui Bai Yizi Shao

10.22489/cinc.2023.014

Cite

Citations (0)

mmSignature: Semi-supervised human identification system based on millimeter wave radar

Engineering Applications of Artificial Intelligence (2023)

Yicheng Yao Hao Zhang Pan Xia Changyu Liu Fanglin Geng

Identification

Sensor Fusion

Data set

10.1016/j.engappai.2023.106939

Cite

Citations (6)

Advanced Deep Neural Network with Unified Feature-Aware and Label Embedding for Multi-Label Arrhythmias Classification

Tsinghua Science & Technology (2024)

Pan Xia Zhongrui Bai Yicheng Yao Lirui Xu Hao Zhang

Multi-label classification

Feature (linguistics)

Deep Neural Networks

10.26599/tst.2023.9010162

Cite

Citations (0)

LAMOST 2D pipeline

Proceedings of the International Astronomical Union (2012)

Zhongrui Bai

Abstract This paper describes the 2-dimensional data reduction of the LAMOST pilot survey.

LAMOST

10.1017/s1743921313004766

Cite

Citations (1)

A Novel Multi-Scale Convolutional Neural Network for Arrhythmia Classification on Reduced-Lead ECGs

2021 Computing in Cardiology (CinC) (2021)

Pan Xia Zhengling He Yusi Zhu Zhongrui Bai Xianya Yu

The PhysioNet/Computing in Cardiology Challenge 2021 focused on exploring the utility of reduced-lead ECGs for arrhythmia classification. Our team, AIRCAS_MEL1, proposed a novel multi-scale convolutional neural network that can classify 30 scored arrhythmias from 12-lead, 6-lead, 4-lead, 3-lead, and 2-lead ECGs. The proposed network was achieved by multiple branch networks to effectively extract the pathological information from the "detail" scale to the "approximation" scale via a series of 2-D convolution kernels (filters). The backbone of each branch network was a carefully designed 2-D convolutional neural network (CNN) with residual connection and attention mechanism, and it can adapt to multi-lead ECGs as input. The first 10 seconds of records from corresponding leads were extracted and preprocessed as inputs for end-to-end training, and the prediction probabilities of 30 categories were outputs. Finally, the proposed algorithms were evaluated on the hidden test set, and our classifiers received scores of 0.38, 0.33, 0.37, 0.43, and 0.38 (ranked 21th, 23th, 20th, 16th, and 20th out of 39 teams) for the 12-lead, 6-lead, 4-lead, 3-lead, and 2-lead versions of the hidden test set with the Challenge evaluation metric.

Lead (geology)

Cardiac arrhythmia

10.23919/cinc53138.2021.9662781

Cite

Citations (2)

Unconstrained Accurate Beat-to-beat Heart Rate Extraction Based on Piezoelectric Ceramics Sensor

电子与信息学报 (2021)

Zhen Fang Zhongrui Bai Xianxiang Chen Pan Xia Zhengling He

BallistoCardioGram (BCG) can be used for contactless detection of vital signs. In BCG’s beat-to-beat heart rate extraction, the lower mean absolute error is of great significance for accurately obtaining the user’s Heart Rate Variability (HRV) indicators. In order to solve the shortcomings in the accuracy of beat-to-beat heart rate calculation of most current methods, a BCG acquisition system based on piezoelectric ceramics sensor is designed in this paper. By adopting a suitable structure for the sensor‘s shell and a suitable sampling frequency, the sensitivity of the sensor and the time resolution of the BCG signal are increased. Through the analysis of BCG, the most suitable components in BCG is found to extract beat by beat cardiac cycle. At the same time, this paper proposes an adaptive template matching algorithm using AP clustering to extract accurately cardiac cycle information. Analysis of the data of 5741 heartbeats of 15 subjects shows that the average error of the beat-to-beat heartbeat cycle is 0.48%, the Mean Absolute Error (MAE) is 3.78 ms, and the heartbeat coverage is above 95%, which is better than other similar work.

Beat (acoustics)

Heart beat

Cardiac cycle

Heart beat

10.11999/jeit200045

Cite

Citations (2)

Binaries with possible compact components discovered from the LAMOST Time-Domain Survey of four $K$2 plates

arXiv (Cornell University) (2022)

Li Xue Wang Song Zhao Xinlin Zhongrui Bai Yuan Hai-long

Time-domain (TD) spectroscopic data from the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) can provide accurate and high-cadence radial velocities (RVs). In this work, we search for binaries with compact components with RV monitoring method by using the LAMOST TD survey of four $K$2 plates. Three binary systems including an unseen white dwarf or neutron star are found. For each binary system, we estimate the stellar parameters of the visible star and orbital parameters, and finally calculate the binary mass function and the minimum mass of the unseen star. No obvious double-lined feature is seen from the LAMOST medium-resolution spectra of the three sources. In addition, we found no X-ray counterpart for all these sources but UV companions for two of them. Spectral disentangling also shows no additional component with optical absorption spectra, supporting that these systems contain compact objects.

LAMOST

Mass ratio

10.48550/arxiv.2209.00765

Cite

Citations (0)

Early Sepsis Prediction Using Ensemble Learning with Features Extracted from LSTM Recurrent Neural Network

Zhengling He Yichen Pan Xianxiang Chen Zhen Fang Weidong Yi

Early prediction of sepsis can help to identify potential risks in time and help take necessary measures to prevent more dangerous situations from occurring. In PhysioNet/Computing in Cardiology Challenge 2019, we integrate Long Short Term Memory (LSTM) recurrent neural network and ensemble learning to achieve early sepsis prediction. Specifically, we tackle the problem of class imbalance and data missing firstly, and then we manually extract features according to the prior knowledge from the medical field. In addition, we regard the prediction of sepsis as a time series prediction problem and pre-train LSTM-based models as feature extractors to obtain the deep features on time series that might be related to the onset of sepsis. Manual features and deep features are then used to train prediction models under the framework of ensemble learning, including Extreme Gradient Boosting (XGBoost) and Gradient Boosting Decision Tree (GBDT) regressor. The final normalized utility score our team (UCAS_DataMiner) have obtained was 0.313 on full hidden test set.

Boosting

Gradient boosting

Ensemble Learning

Ensemble forecasting

Feature (linguistics)

10.23919/cinc49843.2019.9005929

Cite

Citations (1)

A Practical Deconvolution Computation Algorithm to Extract 1D Spectra from 2D Images of Optical Fiber Spectroscopy

Publications of the Astronomical Society of the Pacific (2015)

LI Guang-wei Haotong Zhang Zhongrui Bai

Bolton and Schlegel presented a promising deconvolution method to extract 1D spectra from a 2D optical fiber spectral CCD image. The method could eliminate the PSF difference between fibers, extract spectra to the photo noise level, as well as improve the resolution. But the method is limited by its huge computation requirement and thus cannot be implemented in actual data reduction. In this article, we develop a practical computation method to solve the computation problem. The new computation method can deconvolve a 2D fiber spectral image of any size with actual PSFs, which may vary with positions. Our method does not require large amounts of memory and can extract a 4k multi 4k noise-free CCD image with 250 fibers in 2 hr. To make our method more practical, we further consider the influence of noise, which is thought to be an intrinsic illposed problem in deconvolution algorithms. We modify our method with a Tikhonov regularization item to depress the method induced noise. Compared with the results of traditional extraction methods, our method has the least residual and influence by cross talk and noise, even for extreme situation. Our method can convergent in 2 to 4 iterations, and the computation times are about 3.5 hr for the extreme fiber distance and about 2 hr for nonextreme cases. Finally, we apply our method to real LAMOST (Large sky Area Multi-Object fiber Spectroscopic Telescope, a.k.a. Guo Shou Jing Telescope) data. We find that the 1D spectra extracted by our method have both higher signal-to-noise ratio and resolution than the traditional methods.

10.1086/682052

Cite

Citations (6)

Early Sepsis Prediction Using Ensemble Learning with Features Extracted from LSTM Recurrent Neural Network

Computing in cardiology (2019)

Zhengling He Xianxiang Chen Zhen Fang Weidong Yi Chenshuo Wang

Early prediction of sepsis can help to identify potential risks in time and help take necessary measures to prevent more dangerous situations from occurring.In PhysioNet/Computing in Cardiology Challenge 2019, we integrate Long Short Term Memory (LSTM) recurrent neural network and ensemble learning to achieve early sepsis prediction.Specifically, we tackle the problem of class imbalance and data missing firstly, and then we manually extract features according to the prior knowledge from the medical field.In addition, we regard the prediction of sepsis as a time series prediction problem and pre-train LSTM-based models as feature extractors to obtain the "deep" features on time series that might be related to the onset of sepsis.Manual features and "deep" features are then used to train prediction models under the framework of ensemble learning, including Extreme Gradient Boosting (XGBoost) and Gradient Boosting Decision Tree (GBDT) regressor.The final normalized utility score our team (UCAS_DataMiner) have obtained was 0.313 on full hidden test set.

Ensemble Learning

10.22489/cinc.2019.269

Cite

Citations (7)