A5100445660- AI in cancer detection
- Medical Image Segmentation Techniques
- Crystallization and Solubility Studies
- Radiomics and Machine Learning in Medical Imaging
- X-ray Diffraction in Crystallography
- Image Retrieval and Classification Techniques
- Topic Modeling
- Natural Language Processing Techniques
- Advanced Neural Network Applications
- Video Surveillance and Tracking Methods
- Digital Imaging for Blood Diseases
- Advanced Image and Video Retrieval Techniques
- Text and Document Classification Technologies
- Image and Signal Denoising Methods
- Lung Cancer Treatments and Mutations
- HER2/EGFR in Cancer Research
- Image Processing Techniques and Applications
- Spam and Phishing Detection
- Augmented Reality Applications
- Colorectal Cancer Screening and Detection
- Human Pose and Action Recognition
- Advanced Vision and Imaging
- Face recognition and analysis
- Robotics and Sensor-Based Localization
- Medical Imaging and Analysis
Nanjing Medical University
2023-2025
Southeast University
2025
ZTE (China)
2025
University of Science and Technology Beijing
2012-2025
Guangxi Academy of Sciences
2025
Southern Medical University
2023-2025
Nanfang Hospital
2023-2025
Ministry of Education of the People's Republic of China
2025
Beihang University
2015-2024
Institute of Information Engineering
2024
Histopathology image analysis is a gold standard for cancer recognition and diagnosis. Automatic of histopathology images can help pathologists diagnose tumor subtypes, alleviating the workload pathologists. There are two basic types tasks in digital analysis: classification segmentation. Typical problems with that hamper automatic include complex clinical representations, limited quantities training dataset, extremely large size singular (usually up to gigapixels). The property single also...
This paper studies the effectiveness of accomplishing high-level tasks with a minimum manual annotation and good feature representations for medical images. In image analysis, objects like cells are characterized by significant clinical features. Previously developed features SIFT HARR unable to comprehensively represent such objects. Therefore, representation is especially important. this paper, we study automatic extraction through deep learning (DNN). Furthermore, detailed often an...
3D medical image registration is of great clinical importance. However, supervised learning methods require a large amount accurately annotated corresponding control points (or morphing), which are very difficult to obtain. Unsupervised ease the burden manual annotation by exploiting unlabeled data without supervision. In this paper, we propose new unsupervised method using convolutional neural networks under an end-to-end framework, Volume Tweening Network (VTN), for registration. We three...
Feature warping is a core technique in optical flow estimation; however, the ambiguity caused by occluded areas during major problem that remains unsolved. In this paper, we propose an asymmetric occlusion-aware feature matching module, which can learn rough occlusion mask filters useless (occluded) immediately after without any explicit supervision. The proposed module be easily integrated into end-to-end network architectures and enjoys performance gains while introducing negligible...
We propose a simple, efficient and effective method using deep convolutional activation features (CNNs) to achieve stat- of-the-art classification segmentation for the MICCAI 2014 Brain Tumor Digital Pathology Challenge. Common traits of such medical image challenges are characterized by large dimensions (up gigabyte size an image), limited amount training data, significant clinical feature representations. To tackle these challenges, we transfer extracted from CNNs trained with very general...
A new image instance segmentation method is proposed to segment individual glands (instances) in colon histology images. This process challenging since the not only need be segmented from a complex background, they must also individually identified.We leverage idea of image-to-image prediction recent deep learning by designing an algorithm that automatically exploits and fuses multichannel information-regional, location, boundary cues-in gland Our algorithm, framework, alleviates heavy...
LiDAR point cloud analysis is a core task for 3D computer vision, especially autonomous driving. However, due to the severe sparsity and noise interference in single sweep cloud, accurate semantic segmentation non-trivial achieve. In this paper, we propose novel sparse framework assisted by learned contextual shape priors. practice, an initial (SS) of can be achieved any appealing network then flows into scene completion (SSC) module as input. By merging multiple frames sequence supervision,...
Numerous task-specific variants of conditional generative adversarial networks have been developed for image completion. Yet, a serious limitation remains that all existing algorithms tend to fail when handling large-scale missing regions. To overcome this challenge, we propose generic new approach bridges the gap between image-conditional and recent modulated unconditional architectures via co-modulation both stochastic style representations. Also, due lack good quantitative metrics...
In this paper, we develop a new weakly supervised learning algorithm to learn segment cancerous regions in histopathology images. This paper is under multiple instance (MIL) framework with formulation, deep weak supervision (DWS); also propose an effective way introduce constraints our neural networks assist the process. The contributions of are threefold: 1) build end-to-end system that segments fully convolutional (FCNs) which image-to-image weakly-supervised performed; 2) DWS formulation...
In this paper, we tackle the problem of common object (multiple classes) discovery from a set input images, where assume presence one class in each image. This is, loosely speaking, unsupervised since do not know priori about type, location, and scale We observe that general task fully manner is intrinsically ambiguous; here adopt saliency detection to propose candidate image windows/patches turn an learning into weakly-supervised problem. algorithm for simultaneously localizing objects...
Automatic Non-rigid Histological Image Registration (ANHIR) challenge was organized to compare the performance of image registration algorithms on several kinds microscopy histology images in a fair and independent manner. We have assembled 8 datasets, containing 355 with 18 different stains, resulting 481 pairs be registered. accuracy evaluated using manually placed landmarks. In total, 256 teams registered for challenge, 10 submitted results, 6 participated workshop. Here, we present...
We present a cross-modality generation framework that learns to generate translated modalities from given in MR images. Our proposed method performs Image Modality Translation (abbreviated as IMT) by means of deep learning model leverages conditional generative adversarial networks (cGANs). jointly exploits the low-level features (pixel-wise information) and high-level representations (e.g. brain tumors, structure like gray matter, etc.) between cross which are important for resolving...
We present recursive cascaded networks, a general architecture that enables learning deep cascades, for deformable image registration. The proposed is simple in design and can be built on any base network. moving warped successively by each cascade finally aligned to the fixed image; this procedure way every learns perform progressive deformation current image. entire system end-to-end jointly trained an unsupervised manner. In addition, enabled architecture, one iteratively applied multiple...
A system that translates narrative text in the medical domain into structured representation is great demand. The performs three sub-tasks: concept extraction, assertion classification, and relation identification.The overall consists of five steps: (1) pre-processing sentences, (2) marking noun phrases (NPs) adjective (APs), (3) extracting concepts use a dosage-unit dictionary to dynamically switch two models based on Conditional Random Fields (CRF), (4) classifying assertions voting...
Trabecular bone (TB) is a complex quasi-random network of interconnected plates and rods. TB constantly remodels to adapt the stresses which it subjected (Wolff's Law). In osteoporosis, this dynamic equilibrium between formation resorption perturbed, leading loss structural deterioration. Both deterioration increase fracture risk. Bone's mechanical behavior can only be partially explained by variations in mineral density, led notion quality. Previously, we developed digital topological...
Cancer tissues in histopathology images exhibit abnormal patterns; it is of great clinical importance to label a image as having cancerous regions or not and perform the corresponding segmentation. However, detailed annotation cancer cells often an ambiguous challenging task. In this paper, we propose new learning method, multiple clustered instance (MCIL), classify, segment cluster colon images. The proposed MCIL method simultaneously performs image-level classification (cancer vs....
Learning from synthetic faces, though perhaps appealing for high data efficiency, may not bring satisfactory performance due to the distribution discrepancy of and real face images. To mitigate this gap, we propose a 3D-Aided Deep Pose-Invariant Face Recognition Model (3D-PIM), which automatically recovers realistic frontal faces arbitrary poses through 3D model in novel way. Specifically, 3D-PIM incorporates simulator with aid Morphable (3D MM) obtain shape appearance prior accelerating...
The visual attributes of cells, such as the nuclear morphology and chromatin openness, are critical for histopathology image analysis. By learning cell-level representation, we can obtain a rich mix features that highly reusable various tasks, celllevel classification, nuclei segmentation, cell counting. In this paper, propose unified generative adversarial networks architecture with new formulation loss to perform robust representation in an unsupervised setting. Our model is not only...
Background Consisting of dictated free-text documents such as discharge summaries, medical narratives are widely used in natural language processing. Relationships between anatomical entities and human body parts crucial for building text mining applications. To achieve this, we establish a mapping system consisting Wikipedia-based scoring algorithm named entity normalization method (NEN). The makes full use information available on Wikipedia, which is comprehensive Internet knowledge base....