A5103174841- Medical Image Segmentation Techniques
- AI in cancer detection
- Face and Expression Recognition
- Topic Modeling
- Image Retrieval and Classification Techniques
- Text and Document Classification Technologies
- Optical measurement and interference techniques
- Natural Language Processing Techniques
- Connexins and lens biology
- Radiomics and Machine Learning in Medical Imaging
- Advanced Image Processing Techniques
- Corneal surgery and disorders
- Advanced Image and Video Retrieval Techniques
- Bayesian Methods and Mixture Models
- Advanced Vision and Imaging
- Sentiment Analysis and Opinion Mining
- Colorectal Cancer Screening and Detection
- Advanced Clustering Algorithms Research
- Dental Radiography and Imaging
- Gene expression and cancer classification
- Generative Adversarial Networks and Image Synthesis
- Image Processing Techniques and Applications
- Brain Tumor Detection and Classification
- Digital Holography and Microscopy
- Advanced Text Analysis Techniques
Shanghai University
2019-2020
Siemens (United States)
2011-2014
Microsoft (United States)
2014
Siemens Healthcare (United States)
2013
Rensselaer Polytechnic Institute
2009-2011
Named entity recognition (NER) is an essential part of natural language processing tasks. Chinese NER task different from the many European languages due to lack delimiters. Therefore, Word Segmentation (CWS) usually regarded as first step NER. However, word-based models relying on CWS are more vulnerable incorrectly segmented boundaries and presence out-of-vocabulary (OOV) words. In this paper, we propose a novel character-based Gated Convolutional Recurrent neural network with Attention...
The automatic extraction and labeling of the rib centerlines is a useful yet challenging task in many clinical applications. In this paper, we propose new approach integrating seed point detection template matching to detect identify each chest CT scans. bottom-up learning based exploits local image cues top-down deformable imposes global shape constraints. To adapt deformation different cages whereas maintain high computational efficiency, employ Markov Random Field (MRF) articulated rigid...
The problem of learning a proper distance or similarity metric arises in many applications such as content-based image retrieval. In this work, we propose boosting algorithm, MetricBoost, to learn the that preserves proximity relationships among object triplets: i is more similar j than k. Metric-Boost constructs positive semi-definite (PSD) matrix parameterizes by combining rank-one PSD matrices. Different options weak models and combination coefficients are derived. Unlike existing...
Relation classification is an important research area in the field of natural language processing (NLP), which aims to recognize relationship between two tagged entities a sentence. The noise caused by irrelevant words and word distance may affect relation accuracy. In this paper, we present novel model multi-head attention long short term memory (LSTM) network with filter mechanism (MALNet) extract text features classify particular, combine LSTM obtain shallow local information introduce...
We present a novel method to solve sign ambiguity for phase demodulation from single interferometric image that possibly contains closed fringes. The problem is formulated in binary pairwise energy minimization framework based on gradient orientation continuity. objective function non-submodular and therefore its an NP-hard problem, which we devise multigrid hierarchy of quadratic pseudoboolean optimization problems can be improved iteratively approximate the optimal solutions. Compared with...
Automatic lesion detection is important for cancer examination and treatment, whereas it remains challenging due to the varied shape, size, contextual anatomy of diseased masses. In this paper, we present a robust effective learning based method automatic liver lesions from computed tomography data. The contributions paper are following. First, develop cascade approach comprising multiple detectors in spirit marginal space learning. Second, gradient locally adaptive segmentation proposed...
Text classification is an essential task in many natural language processing (NLP) applications; we know each sentence may have only a few words that play important role text classification, while other no significant effect on the results. Finding these keywords has impact accuracy. In this paper, propose network model, named RCNNA, recurrent convolution neural networks with attention (RCNNA), which models human conditional reflexes for classification. The model combines bidirectional LSTM...
We propose a dimension reduction technique named Resilient Subclass Discriminant Analysis (RSDA) for high dimensional classification problems. The iteratively estimates the subclass division by embedding Fisher (FDA) with Expectation-Maximization (EM) in Gaussian Mixture Models (GMM). new method maintains adaptability of SDA to wide range data distributions approximating distribution each class as mixture Gaussians, and provides superior feature selection performance modified EM clustering...
The computer aided diagnosis (CAD) problems of detecting potentially diseased structures from medical images are typically distinguished by the following challenging characteristics: extremely unbalanced data between negative and positive classes; stringent real-time requirement online execution; multiple candidates generated for same malignant structure that highly correlated spatially close to each other. To address all these problems, we propose a novel learning formulation combine...