A5031166420- Advanced Neural Network Applications
- Medical Image Segmentation Techniques
- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- Topic Modeling
- COVID-19 diagnosis using AI
- Machine Learning in Healthcare
- Acute Ischemic Stroke Management
- Machine Learning and Algorithms
- Image Processing Techniques and Applications
- Image Retrieval and Classification Techniques
- Advanced Graph Neural Networks
- Medical Imaging Techniques and Applications
- Text and Document Classification Technologies
- Generative Adversarial Networks and Image Synthesis
- Time Series Analysis and Forecasting
- Machine Learning and Data Classification
- Remote-Sensing Image Classification
- Artificial Intelligence in Healthcare
- Remote Sensing and Land Use
- Advanced Text Analysis Techniques
- Digital Image Processing Techniques
- Recommender Systems and Techniques
- Explainable Artificial Intelligence (XAI)
- Advanced Image Processing Techniques
Xi'an Jiaotong University
2016-2023
Tencent (China)
2019-2022
Kyushu University
2020
Keio University
2020
Hiroshima University
2020
NTT (Japan)
2020
Microvascular invasion (MVI) is a valuable predictor of survival in hepatocellular carcinoma (HCC) patients. This study developed predictive models using eXtreme Gradient Boosting (XGBoost) and deep learning based on CT images to predict MVI preoperatively. In total, 405 patients were included. A total 7302 radiomic features 17 radiological extracted by radiomics feature extraction package radiologists, respectively. We XGBoost model features, clinical variables three-dimensional...
Coronavirus Disease 2019 (COVID-19) has rapidly spread worldwide since first reported. Timely diagnosis of COVID-19 is crucial both for disease control and patient care. Non-contrast thoracic computed tomography (CT) been identified as an effective tool the diagnosis, yet outbreak placed tremendous pressure on radiologists reading exams may potentially lead to fatigue-related mis-diagnosis. Reliable automatic classification algorithms can be really helpful; however, they usually require a...
We introduce a one-shot segmentation method to alleviate the burden of manual annotation for medical images. The main idea is treat as classical atlas-based problem, where voxel-wise correspondence from atlas unlabelled data learned. Subsequently, label can be transferred with learned correspondence. However, since ground truth between images usually unavailable, learning system must well-supervised avoid mode collapse and convergence failure. To overcome this difficulty, we resort...
Building a predictive model based on historical Electronic Health Records (EHRs) for personalized healthcare has become an active research area. Benefiting from the powerful ability of feature extraction, deep learning (DL) approaches have achieved promising performance in many clinical prediction tasks. However, due to lack interpretability and trustworthiness, it is difficult apply DL real cases decision making. To address this, this paper, we propose interpretable trustworthy...
Active learning aims to reduce manual labeling efforts by proactively selecting the most informative unlabeled instances query. In real-world scenarios, it's often more practical query a batch of rather than single one at each iteration. To achieve this we need keep not only informativeness but also their diversity. Many heuristic methods have been proposed tackle mode active problems, however, they suffer from two limitations which if addressed would significantly improve strategy. Firstly,...
Comparing and identifying similar patients is a fundamental task in medical domains - an efficient technique can, for example, help doctors to track patient cohorts, compare the effectiveness of treatments, or predict outcomes. The goal similarity learning derive clinically meaningful measure evaluate amongst represented by their key clinical indicators. However, it challenging learn such similarity, as data are usually high dimensional, heterogeneous, complex. In addition, desirable...
Fully convolutional neural networks have made promising progress in joint liver and tumor segmentation. Instead of following the debates over 2D versus 3D (for example, pursuing balance between large-scale pretraining context), this paper, we novelly identify wide variation ratio intra- inter-slice resolutions as a crucial obstacle to performance. To tackle mismatch information, propose slice-aware 2.5D network that emphasizes extracting discriminative features utilizing not only in-plane...
We propose a Healthcare Graph Convolutional Network (HealGCN) to offer disease self-diagnosis service for online users based on Electronic Records (EHRs). Two main challenges are focused in this paper diagnosis: (1) serving cold-start via graph convolutional networks and (2) handling scarce clinical description symptom retrieval system. To end, we first organize the EHR data into heterogeneous that is capable of modeling complex interactions among users, symptoms diseases, tailor...
The need for short-text classification arises in many text mining applications particularly health care applications. In such shorter texts mean linguistic ambiguity limits the semantic expression, which turns would make typical methods fail to capture exact semantics of scarce words. This is true domains when contains domain-specific or infrequently appearing words, whose embedding can not be easily learned due lack training data. Deep neural network has shown great potentials boost...
Universal lesion detection in computed tomography (CT) images is an important yet challenging task due to the large variations type, size, shape, and appearance. Considering that data clinical routine (such as DeepLesion dataset) are usually annotated with a long short diameter according standard of Response Evaluation Criteria Solid Tumors (RECIST) diameters, we propose RECIST-Net, new approach which four extreme points center point RECIST diameters detected. By detecting keypoints, provide...
Active learning is usually used in scenarios where few labels are available and manual labeling expensive. To improve model performance, it necessary to find the most valuable instance among all instances label maximize benefits of labeling. In practical scenarios, often more efficient query a group instead individual during each iteration. achieve this goal, we need explore similarities between ensure informativeness diversity. Many ad-hoc algorithms proposed for batch mode active learning,...
Thanks to the huge accumulation of Electronic Health Records (EHRs), numerous deep learning based predictive models were proposed for this task. Among them, most existing state-of-the-art (SOTA) built with recurrent neural networks (RNNs). Regardless their success, RNN-based mainly suffer from three limitations. (i) Accuracy: prediction accuracy drops quickly as length EHR sequences increases. (ii) Efficiency: recurrence property makes computation parallelization impossible, and accordingly...
Vast amounts of remote sensing (RS) data provide Earth observations across multiple dimensions, encompassing critical spatial, temporal, and spectral information which is essential for addressing global-scale challenges such as land use monitoring, disaster prevention, environmental change mitigation. Despite various pre-training methods tailored to the characteristics RS data, a key limitation persists: inability effectively integrate within single unified model. To unlock potential we...
Graph neural networks have been widely used in recent recommender systems, where negative sampling plays an important role. Existing methods restrict the relationship between nodes as either hard positive pairs or pairs. This leads to loss of structural information, and lacks mechanism generate for with few neighbors. To overcome limitations, we propose a novel soft link-based method, namely MixDec Sampling, which consists Mixup Sampling module Decay module. The augments node features by...
We introduce a one-shot segmentation method to alleviate the burden of manual annotation for medical images. The main idea is treat as classical atlas-based problem, where voxel-wise correspondence from atlas unlabelled data learned. Subsequently, label can be transferred with learned correspondence. However, since ground truth between images usually unavailable, learning system must well-supervised avoid mode collapse and convergence failure. To overcome this difficulty, we resort...
Semantic segmentation is important in medical image analysis. Inspired by the strong ability of traditional analysis techniques capturing shape priors and inter-subject similarity, many deep learning (DL) models have been recently proposed to exploit such prior information achieved robust performance. However, these two types are usually studied separately existing models. In this paper, we propose a novel DL model both type within single framework. Specifically, introduce an extra encoder...
Low-shot (one/few-shot) segmentation has attracted increasing attention as it works well with limited annotation. State-of-the-art low-shot methods on natural images usually focus implicit representation learning for each novel class, such prototypes, deriving guidance features via masked average pooling, and segmenting using cosine similarity in feature space. We argue that medical should step further to explicitly learn dense correspondences between utilize the anatomical similarity. The...
Deep neural network-based medical image classifications often use "hard" labels for training, where the probability of correct category is 1 and those others are 0. However, these hard targets can drive networks over-confident about their predictions prone to overfit training data, affecting model generalization adaption. Studies have shown that label smoothing softening improve classification performance. Nevertheless, existing approaches either non-data-driven or limited in applicability....
Segmentation of objects interest is one the central tasks in medical image analysis, which indispensable for quantitative analysis. When developing machine-learning based methods automated segmentation, manual annotations are usually used as ground truth toward models learn to mimic. While bulky parts segmentation targets relatively easy label, peripheral areas often difficult handle due ambiguous boundaries and partial volume effect, etc., likely be labeled with uncertainty. This...