Lei Zhang

Shenyang Aerospace University

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Fei Yang

Children's Hospital of Zhejiang University

Kuanquan Wang

Harbin Institute of Technology

Wangmeng Zuo

Harbin Institute of Technology

Weigang Lu

Ocean University of China

Shandong Wu

University of Pittsburgh

Henggui Zhang

University of Manchester

David Dreizin

University of Maryland, Baltimore

Aly A. Mohamed

University of Pittsburgh

Ruimei Chai

Affiliated Hospital of Qingdao University

Yue Zhang

Harbin Institute of Technology

Cooperative Institutions

University of Maryland, Baltimore

Chinese Academy of Sciences

Harbin Institute of Technology

Zhejiang University

University of Maryland, College Park

Shandong University

University of Baltimore

Memorial Sloan Kettering Cancer Center

Johns Hopkins University

University of Chinese Academy of Sciences

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Deep learning for identifying breast cancer malignancy and false recalls: a robustness study on training strategy

Medical Imaging 2018: Computer-Aided Diagnosis (2019)

Kadie Clancy Shandong Wu Lei Zhang Aly A. Mohamed Sarah Aboutalib

Identification of malignancy and false recalls (women who are recalled in screening for additional workup, but later proven benign) in screening mammography has significant clinical value for accurate diagnosis of breast cancer. Deep learning methods have recently shown success in the area of medical imaging classification. However, there are a multitude of different training strategies that can significantly impact the overall model performance for a specific classification task. In this study, we aimed to investigate the impact of training strategy on classification of digital mammograms by performing a robustness analysis of deep learning models to distinguish malignancy and false-recall from normal (benign) findings. Specifically, we employed several pre-training strategies including transfer learning with medical and non-medical datasets, layer freezing, and varied network structure on both binary and three-class classification tasks of digital mammography images. We found that, overall, deep learning models appear to be robust to some modifications of network structure and pre-training strategy that we tested for mammogram-specific classification tasks. However, for specific classification tasks, some training strategies offer performance gains. The most notable performance gains in our experiments involved residual network models.

Robustness

Binary classification

Breast Cancer Screening

10.1117/12.2512942

Cite

Citations (6)

Expression of sialyltransferase activity on intact human neutrophils

Journal of Leukocyte Biology (2008)

S Rifat Tae Jin Kang Dean L. Mann Lei Zhang Adam C. Puché

Abstract Endogenous polymorphonuclear leukocyte (PMN)-associated sialidase activity enhances PMN adhesion to and migration across the endothelium through the removal of sialylated cell-surface residues. We tested the hypothesis that PMNs also express sialyltransferase (ST) activity that restores sialyl residues to the PMN surface. We developed a highly sensitive fluorometric assay to demonstrate that intact human PMNs can mediate and accept sialyl residue transfer. This ST activity is inhibited by a ST inhibitor, CMP, which also inhibits the transendothelial migration of PMNs in response to IL-8 in vitro and in vivo. We conclude that intact PMNs express sialidase and ST activities that permit rapid modulation of their surface sialylation and their ability to adhere to and migrate across the endothelium.

Sialyltransferase

Sialidase

10.1189/jlb.0706462

Cite

Citations (28)

Fully automated tumor localization and segmentation in breast DCEMRI using deep learning and kinetic prior

Lei Zhang Dooman Arefan Yuan Guo Shandong Wu

Breast magnetic resonance imaging (MRI) plays an important role in high-risk breast cancer screening, clinical problemsolving, and imaging-based outcome prediction. Breast tumor segmentation in MRI is an essential step for quantitative radiomics analysis, where automated and accurate tumor segmentation is needed but very challenging. Automated breast tumor segmentation methods have been proposed and can achieve promising results. However, these methods still need a pre-defined a region of interest (ROI) before performing segmentation, which makes them hard to run fully automatically. In this paper, we investigated automated localization and segmentation method for breast tumor in breast Dynamic Contrast-Enhanced MRI (DCE-MRI) scans. The proposed method takes advantage of kinetic prior and deep learning for automatic tumor localization and segmentation. We implemented our method and evaluated its performance on a dataset consisting of 74 breast MR images. We quantitatively evaluated the proposed method by comparing the segmentation with the manual annotation from an expert radiologist. Experimental results showed that the automated breast tumor segmentation method exhibits promising performance with an average Dice Coefficient of 0.89±0.06.

Breast MRI

Sørensen–Dice coefficient

Breast tumor

Region growing

10.1117/12.2551372

Cite

Citations (6)

Open-Set Recognition of Wood Species Based on Deep Learning Feature Extraction Using Leaves

Journal of Imaging (2023)

Tianyu Fang Zhenyu Li Jialin Zhang Dawei Qi Lei Zhang

An open-set recognition scheme for tree leaves based on deep learning feature extraction is presented in this study. Deep learning algorithms are used to extract leaf features for different wood species, and the leaf set of a wood species is divided into two datasets: the leaf set of a known wood species and the leaf set of an unknown species. The deep learning network (CNN) is trained on the leaves of selected known wood species, and the features of the remaining known wood species and all unknown wood species are extracted using the trained CNN. Then, the single-class classification is performed using the weighted SVDD algorithm to recognize the leaves of known and unknown wood species. The features of leaves recognized as known wood species are fed back to the trained CNN to recognize the leaves of known wood species. The recognition results of a single-class classifier for known and unknown wood species are combined with the recognition results of a multi-class CNN to finally complete the open recognition of wood species. We tested the proposed method on the publicly available Swedish Leaf Dataset, which includes 15 wood species (5 species used as known and 10 species used as unknown). The test results showed that, with F1 scores of 0.7797 and 0.8644, mixed recognition rates of 95.15% and 93.14%, and Kappa coefficients of 0.7674 and 0.8644 under two different data distributions, the proposed method outperformed the state-of-the-art open-set recognition algorithms in all three aspects. And, the more wood species that are known, the better the recognition. This approach can extract effective features from tree leaf images for open-set recognition and achieve wood species recognition without compromising tree material.

10.3390/jimaging9080154

Cite

Citations (0)

Joint Representation and Truncated Inference Learning for Correlation Filter Based Tracking

Lecture notes in computer science (2018)

Yingjie Yao Xiaohe Wu Lei Zhang Shiguang Shan Wangmeng Zuo

BitTorrent tracker

Representation

Feature Learning

Feature (linguistics)

Tracking (education)

10.1007/978-3-030-01240-3_34

Cite

Citations (48)

Robust object tracking using semi-supervised appearance dictionary learning

Pattern Recognition Letters (2015)

Lei Zhang Wen Wu Terrence Chen Norbert Strobel Dorin Comaniciu

Active appearance model

Tracking (education)

Representation

10.1016/j.patrec.2015.04.010

Cite

Citations (34)

Visualization for Non-linear Enhanced Volume Data

Fei Yang Weigang Lu Lei Zhang Kuanquan Wang Yemin Li

There is a strong need for acquisition of high quality volume data sets, with the vision of visualizing organs, engines or other materials for variant fields. Visualized computer model provides interactive investigation and improves understanding of details of substances. Since the quality of volume data which is acquired by CT and MRI devices is impacted by noisy, partial effect and bias seriously, in this paper we first apply non-linear filters in pre-processing stage to improve the quality of 3D volume data, followed by construction of 2D histogram based on external voxels to design the transfer function. Finally, we use ray-casting algorithm to render volume data. We applied the method presented above to four datasets. The noise is suppressed effectively after denoising with nonlinear anisotropic filtering. Boundary between materials can be following differentiated clearly by designing transfer function with 2D histogram constructed based on the pre-processed data.

Ray casting

10.1109/iih-msp.2014.93

Cite

Citations (1)

FPGA Embedded System Design of Multi-Channel Guided Waves Ultrasonic Testing Instrument for Pipeline Inspection

Lei Zhang Peter Mudge P.J. Jackson Ian L. Daniel

Long range ultrasonic testing (LRUT) is an emerging non-destructive testing (NDT) ultrasound testing method. The LRUT technique uses low frequency ultrasonic guided waves to inspect plates and pipes efficiently for corrosion and other degradation. Increasing numbers of requirements for quick long range pipeline inspection have led to urgent need for improvement of testing methods and the development of new testing equipment to help the researchers in laboratory and help technicians in field inspection. The market for multi-channel ultrasonic instruments is characterized by small volumes of sales and a high ratio of development costs to sale price. The consequence is that the investment required to develop an instrument is significant and upgrades and other changes to instrument specification or configuration are difficult for many manufacturers to justify. These factors are particularly relevant for long range guided wave technology, where the technology is still relatively new in the market place and the instrumentation has different characteristics from other type of ultrasonic systems. In order to support the design and manufacture of the Teletest® FOCUS+ system, Plant Integrity Ltd has supported this study into the use of a fully programmable gate array (FPGA) based control circuits for the electronics, which allows the component count to be decreased and also permits the of the system to support future functions to be modified via flexible reconfiguration of the FPGA. This has benefits of reduction of size, weight and power consumption of the electronics and provides a means of upgrading the product without expensive re-design of hardware.

Instrumentation

Control reconfiguration

10.1061/9780784412619.101

Cite

Citations (1)

Fusion visualization for cardiac anatomical and ischemic models with depth weighted optic radiation function

2019 Computing in Cardiology Conference (CinC) (2015)

Fei Yang Weigang Lu Lei Zhang Wangmeng Zuo Kuanquan Wang

Despite the emerging of various innovations of visualization methods, highlighting and exploring the important function of interest of computational cardiac model still remain a challenging problem. Due to the complex and fine geometries of heart, the silhouette of the whole volume structure may be destroyed when there is a necessity to demonstrate cardiac electrophysiological behaviors. In this paper, we proposed a fusion visualization framework, which combines the electrophysiology pattern with the anatomy pattern through a novel multi-dimensional fusion transfer function. A depth weighted optic radiation model is thus presented in the framework to inspect the occluded cardiac ischemia information. According to the depth to the tissue boundary, the hidden ischemia region of pathological tissue can be revealed from complex overlapping features. Experiment results on the cardiac anatomy and electrophysiological simulation data verifies the effectiveness of the proposed method for intuitively exploring and inspecting electrophysiological activities with the preserved authentic context of anatomical structure, which is found to be useful in analyzing and explaining cardiac functioning phenomena.

Cardiac Electrophysiology

10.1109/cic.2015.7411066

Cite

Citations (3)

GPU-based fusion method for 3D electrophysiological data visualization

Lei Zhang Kuanquan Wang Wangmeng Zuo Yunfeng Wu Dongchen Han

Transfer function design is a challenging issue of volume visualization. Some simulation and medical applications which require visualization often produce at least two values per sampling point in the so called multi-modal data, which makes transfer function design even more complex. In this paper, we present a novel visualization method for assisting the assessment of electrophysiological simulation results. The main idea of the proposed method is that the anatomical data of the heart is added as background information to enhance the illustration results of the tissue with electrophysiological behaviour. A improved context-preserving model is proposed to enhance the boundary information of different tissue. The two modalities data of both tissue with electrophysiological simulation and the other tissues are rendering by the proposed fusion rendering method with user friendly interactive facilities. Compared to conventional visualization method, our method can not only provide rendering result of tissue with electrophysiological simulation, but also can provide the spatial information of the given tissue in the heart.

10.1109/icch.2012.6724470

Cite

Citations (8)