A5100379287- AI in cancer detection
- Radiomics and Machine Learning in Medical Imaging
- Cervical Cancer and HPV Research
- Speech and Audio Processing
- Speech Recognition and Synthesis
- Image Retrieval and Classification Techniques
- Digital Radiography and Breast Imaging
- Medical Image Segmentation Techniques
- Digital Imaging for Blood Diseases
- Music and Audio Processing
- Colorectal Cancer Screening and Detection
- Cryptography and Data Security
- Image Processing Techniques and Applications
- Dental Implant Techniques and Outcomes
- Medical Imaging and Analysis
- Indoor and Outdoor Localization Technologies
- Periodontal Regeneration and Treatments
- 3D Shape Modeling and Analysis
- Global Cancer Incidence and Screening
- Machine Learning and Algorithms
- Breast Lesions and Carcinomas
- Image and Object Detection Techniques
- Cryptography and Residue Arithmetic
- Elevator Systems and Control
- Industrial Vision Systems and Defect Detection
Southwest Hospital
2022-2025
Huazhong University of Science and Technology
2019-2024
Tongji Hospital
2019-2024
Zhejiang Chinese Medical University
2024
Army Medical University
2023
Hubei Cancer Hospital
2022
Nantong University
2017-2020
Meta (Israel)
2020
Northwest Normal University
2017-2019
Northwest University
2019
Although successful detection of malignant tumors from histopathological images largely depends on the long-term experience radiologists, experts sometimes disagree with their decisions. Computer-aided diagnosis provides a second option for image diagnosis, which can improve reliability experts' decision-making. Automatic and precision classification breast cancer is great importance in clinical application identifying images. Advanced convolution neural network technology has achieved...
Computer-assisted diagnosis is key for scaling up cervical cancer screening. However, current recognition algorithms perform poorly on whole slide image (WSI) analysis, fail to generalize diverse staining and imaging, show sub-optimal clinical-level verification. Here, we develop a progressive lesion cell method combining low- high-resolution WSIs recommend cells recurrent neural network-based WSI classification model evaluate the degree of WSIs. We train validate our analysis system 3,545...
Automated retinal vessel segmentation technology has become an important tool for disease screening and diagnosis in clinical medicine. However, most of the available methods still have problems such as poor accuracy low generalization ability. This is because symmetrical asymmetrical patterns between blood vessels are complicated, contrast background relatively due to illumination pathology. Robust image essential improving diseases vein occlusions diabetic retinopathy. remains a...
In view of inherent attributes breast BI-RADS 3, benign and malignant lesions are with a subtle difference the imbalanced ratio (with very small part malignancy). The objective this study is to improve detection rate 3 on ultrasound (US) images using deep convolution networks. study, 1,275 out 1,096 patients were included from Southwest Hospital (SW) Tangshan (TS). which, 629 lesions, 218 428 utilized for development dataset, internal external testing set. All biopsy-confirmed, while...
We propose to apply deep transfer learning from computer vision static malware classification. In the scheme, we borrow knowledge natural images or objects and target domain of detection. As a result, training time neural networks is accelerated while high classification performance still maintained. demonstrate effectiveness our approach on three experiments show that proposed method outperforms other classical machine methods measured in accuracy, false positive rate, true rate $F_1$ score...
Early diagnosis of breast cancer has always been a difficult clinical challenge. We developed deep-learning model EDL-BC to discriminate early with ultrasound (US) benign findings. This study aimed investigate how the could help radiologists improve detection rate while reducing misdiagnosis.In this retrospective, multicentre cohort study, we an ensemble deep learning called based on convolutional neural networks. The was trained and internally validated B-mode color Doppler US image 7955...
Diverse styles of cytopathology images have a negative effect on the generalization ability automated image analysis algorithms. This article proposes an unsupervised method to normalize styles. We design two-stage style normalization framework with removal module convert colorful into gray-scale color-encoding mask and domain adversarial reconstruction map them back user-selected style. Our enforces both hue structure consistency before after by using per-pixel regression. Intra-domain...
The morphology of the cervical cell nucleus is most important consideration for pathological identification. And a precise segmentation determines performance final classification traditional algorithms and even some deep learning-based algorithms. Many methods can accurately segment nuclei but will cost lots time, especially when dealing with whole-slide image (WSI) tens thousands cells. To address this challenge, we propose dual-supervised sampling network structure, in which...
Security is a significant issue for ad-hoc networks. Heterogeneous signcryption schemes proposed in recent years satisfy privacy and authentication of messages reduce cost computation communication logical step. In this article, we present new efficient heterogeneous scheme that shifts between the public key infrastructure (PKI) identity-based cryptosystem (IBC). The suitable broadcast conditions which senders can signcrypt multiple receivers simultaneously This proved to confidentiality...
Multi-center cervical cytology images have various image styles due to the differences in staining and imaging techniques, which pose a significant challenge performance of automated cancer diagnosis tools. We propose dual-head network architecture that explicitly disentangles features into content style features, applies contrastive self-supervised learning large number unlabeled images, achieving enhanced generalization across styles. pretrain our model on 1,024,855 cropped from 3,561...
I-vector algorithm was previously adopted to improve the performance of ASR (Automatic Speaker Recognition) system which is degraded by emotion variability. The variability compensation technique LDA (Linear Discriminant Analysis) assumes speaker-independent. However, this assumption not suitable for because we discover that pattern speaker-dependent. Therefore, a novel synthesis AASR (Atom Aligned Sparse Representation) proposed characterize speaker-dependent and compensate within...
In this paper, we present a new efficient authenticated key agreement protocol. Compared with other cross-domain protocols which are all in the same cryptosystem, our protocol is innovatively designed to shift between Public Key Infrastructure (PKI) and Identity-based Cryptosystem (IBC). This combined heterogeneous signcryption method proved satisfy confidentiality unforgeability on basis of assumptions Computational Diffie-Hellman problem (CDHP) Bilinear (BDHP) authentication phases, it...
Abstract Computer-assisted diagnosis is key for popularizing cervical cancer screening. However, current recognition algorithms are insufficient in accuracy and generalization lesion cells, especially when facing diversity data clinical applications. Inspired by manual reading slide under microscopes, we develop a progressive cell method combing low high resolutions WSIs to recommend cells recurrent neural network-based WSI classification model evaluate the degree of WSIs. After validating...
With the popularity of deep neural network, speech synthesis task has achieved significant improvements based on end-to-end encoder-decoder framework in recent days. More and more applications relying technology have been widely used our daily life. Robust model depends high quality customized data which needs lots collecting efforts. It is worth investigating how to take advantage low-quality low resource voice can be easily obtained from Internet for usage synthesizing personalized voice....
This study is part of the research improving early detection breast cancer in screening mammograms by focusing on computerized analysis and cancers missed radiologists. It directed to density cases effect tissue detection. A total 100 were collected which used generate three different datasets including with cancer, screening-detected normal mammograms. statistical-based method was applied segment tissue. The percentage segmented area out whole calculated as index density. set tests examine...
Abstract Automatic classification of H&E breast cancer histopathology images is a challenging task. Computer-aided diagnostic systems help reduce costs and increase the efficiency process. Although existing research on image higher than 90% accurate in binary classifications (non-carcinoma/carcinoma), accuracy four (normal, benign, situ, invasive) less 80%. This paper proposes framework for stained histopathological images, which includes two methods based convolutional neural network....
Beamforming weights prediction via deep neural networks has been one of the main methods in multi-channel speech enhancement tasks. The spectral-spatial cues are crucial beamforming estimation, however, many existing works fail to optimally predict with an absence adequate information learning. To tackle this challenge, we propose a Fourier convolutional attention encoder (FCAE) provide global receptive field over frequency axis and boost learning spectral contexts cross-channel features....
Histological assessment of glands is one the major concerns in colon cancer grading. Considering that poorly differentiated colorectal cannot be accurately segmented, we propose an approach for segmentation images, based on characteristics lumens and rough gland boundaries. First, use a U-net stain separation to obtain H-stain, E-stain, background intensity maps. Subsequently, epithelial nucleus identified histopathology lumen performed map. Then, axis least inertia-based similar triangles...
Abstract Background The distal aspect of the second molar (d-M2) often exhibits infrabony defects due to adjacent third molar. Although can be treated by guided tissue regeneration (GTR) after removing molar, optimal timing remains uncertain following removal in clinical decision-making. This study aimed compare delayed and immediate GTR treatments assist Methods D-M2 with a minimum 1-year follow-up were collected divided into three groups: Immediate group, which underwent extraction...
Cervical cancer is a major global health issue, particularly in developing countries where access to healthcare limited. Early detection of pre-cancerous lesions crucial for successful treatment and reducing mortality rates. However, traditional screening diagnostic processes require cytopathology doctors manually interpret huge number cells, which time-consuming, costly, prone human experiences. In this paper, we propose Multi-scale Window Transformer (MWT) cervical image recognition. We...