检索结果-内蒙古大学图书馆

Separate and Conquer: Decoupling Co-occurrence via Decomposition and Representation for Weakly Supervised Semantic Segmentation

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Yang, Zhiwei Fu, Kexue Duan, Minghong Qu, Linhao Wang, Shuo Song, Zhijian Academy for Engineering and Technology Fudan University China Digital Medical Research Center School of Basic Medical Sciences Fudan University China Shanghai Key Laboratory of Medical Image Computing and Computer Assisted Intervention China Shandong Computer Science Center National Supercomputer Center in Jinan China

Attributed to the frequent coupling of co-occurring objects and the limited supervision from image-level labels, the challenging co-occurrence problem is widely present and leads to false activation of objects in weakly supervised semantic segmentation (WSSS). In this work, we devise a’Separate and Conquer’ scheme SeCo to tackle this issue from dimensions of image space and feature space. In the image space, we propose to’separate’ the co-occurring objects with image decomposition by subdividing images into patches. Importantly, we assign each patch a category tag from Class Activation Maps (CAMs), which spatially helps remove the co-context bias and guide the subsequent representation. In the feature space, we propose to’conquer’ the false activation by enhancing semantic representation with multi-granularity knowledge contrast. To this end, a dual-teacher-single-student architecture is designed and tag-guided contrast is conducted to guarantee the correctness of knowledge and further facilitate the discrepancy among co-occurring objects. We streamline the multi-staged WSSS pipeline end-to-end and tackle co-occurrence without external supervision. Extensive experiments are conducted, validating the efficiency of our method tackling co-occurrence and the superiority over previous single-staged and even multi-staged competitors on PASCAL VOC and MS COCO. Code will be available here. Copyright © 2024, The Authors. All rights reserved.

关键词： Chemical activation

Rethinking Multiple Instance Learning for Whole Slide Image Classification: A Good Instance Classifier is All You Need

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Qu, Linhao Ma, Yingfan Luo, Xiaoyuan Guo, Qinhao Wang, Manning Song, Zhijian Digital Medical Research Center School of Basic Medical Science Fudan University Shanghai200032 China Shanghai Key Lab of Medical Image Computing and Computer Assisted Intervention Shanghai200032 China Department of Gynecologic Oncology Fudan University Shanghai Cancer Center 270 Dong-An Road Shanghai200032 China

Weakly supervised whole slide image classification is usually formulated as a multiple instance learning (MIL) problem, where each slide is treated as a bag, and the patches cut out of it are treated as instances. Existing methods either train an instance classifier through pseudo-labeling or aggregate instance features into a bag feature through attention mechanisms and then train a bag classifier, where the attention scores can be used for instance-level classification. However, the pseudo instance labels constructed by the former usually contain a lot of noise, and the attention scores constructed by the latter are not accurate enough, both of which affect their performance. In this paper, we propose an instance-level MIL framework based on contrastive learning and prototype learning to effectively accomplish both instance classification and bag classification tasks. To this end, we propose an instance-level weakly supervised contrastive learning algorithm for the first time under the MIL setting to effectively learn instance feature representation. We also propose an accurate pseudo label generation method through prototype learning. We then develop a joint training strategy for weakly supervised contrastive learning, prototype learning, and instance classifier training. Extensive experiments and visualizations on four datasets demonstrate the powerful performance of our method. Codes are available at https://***/miccaiif/INS. Copyright © 2023, The Authors. All rights reserved.

关键词： Image classification

key-Point Based Automated Diagnosis for Alveolar Dehiscence in Mandibular Incisors Using Convolutional Neural Network

学校读者我要写书评

暂无评论

SSRN

SSRN 2022年

作者： Liu, Tianyu Ye, Yingzhi Liu, Chengcheng Chen, Jing Liu, Yuehua Xing, Wenyu Ta, Dean Academy for Engineering and Technology Fudan University Shanghai200433 China Department of Orthodontics Shanghai Stomatological Hospital School of Stomatology Fudan University Shanghai200001 China Shanghai Key Laboratory of Craniomaxillofacial Development and Diseases Fudan University Shanghai200001 China Shanghai Key Laboratory of Medical Image Computing and Computer Assisted Intervention Shanghai200032 China Center for Medical Engineering School of Information Science and Technology Fudan University Shanghai200438 China

The aim of this study was to propose an automated diagnosis method for alveolar dehiscence in the anterior teeth using convolutional neural network (CNN).The Cone-beam computed tomography (CBCT) scanning was performed on 387 orthodontic patients at shanghai Stomatological Hospital. A total of 1017 mandibular incisors with the largest labiolingual sectional images were obtained from the CBCT data. We proposed two strategies of automated diagnosis methods for alveolar dehiscence. The first strategy (the binary classification method (BCM)) was to take the task directly as a classic binary classification. The second strategy (the key-point based method (KPM)) was to use the CNN to search two key points (i.e., the cement-enamel junction (CEJ) and the alveolar crest (AC)) firstly and then make a diagnosis according to the distance between the two key points. In both CNN strategies, 90% of the mandibular incisor images were assigned to the training dataset, and the rest 10% were assigned to the testing *** BCM showed limited performance in the diagnosis of alveolar dehiscence, with diagnostic accuracy below 70% for all the five classification networks. The KPM showed superior diagnostic performance with an accuracy of 90.9%, a sensitivity of 81.4% and a specificity of 94.4%, respectively, in the testing *** results proved that the diagnosis of alveolar dehiscence could be realized using CNN and the proposed KPM have the advantage of high accuracy and real-time performance. This study suggests that the KPM might have the potential for the diagnosis of alveolar dehiscence in the clinic. © 2022, The Authors. All rights reserved.

关键词： Convolutional neural networks

Intracardiac Electrogram Signals key Feature Points Recognition Using Dense Convolutional Network

学校读者我要写书评

暂无评论

Intracardiac Electrogram Signals Key Feature Points Recognit...

Pattern Recognition and Machine Learning (PRML), IEEE International Conference on

作者： Jiang Yihang Gu Kaihao Wu Xiaomei Department of Biomedical Engineering School of information Science and Technology Fudan University Shanghai China Department of Biomedical Engineering Key Laboratory of Medical Imaging Computing and Computer Assisted Intervention (MICCAI) of Shanghai School of information Science and Technology Fudan University Academy for Engineering and Technology Shanghai Engineering Research Center of Assistive Devices Shanghai China Yiwu Research Institute of Fudan University Zhejiang China

ISBN: (数字)9798350355925

ISBN: (纸本)9798350355932

During radiofrequency ablation procedures for atrial fibrillation patients, physicians can assist in locating ablation targets by analyzing the conduction pathways and low-voltage areas of the intracardiac electrogram signals. The feature points of intracardiac electrogram signals serve as criteria for assessing excitation patterns and low-voltage areas. Therefore, extracting feature points from intracardiac signals is of significant value. To identify key points in intracardiac electrogram signals through deep learning, this experiment collected intracardiac electrogram signals from 21 clinical patients, which were then filtered using wavelet transformation. Subsequently, each excitation was treated as a signal group, indicating the key feature points within. Finally, a total of 36,183 groups of intracardiac electrogram signals annotated with key points were obtained. The dataset consisted of data from 17 patients allocated for training, data from 2 patients reserved for testing, and data from the remaining 2 patients designated for validation purposes. Training was conducted using a DenseNet network incorporating both convolutional attention and feature pyramid modules. The final network successfully identifies the feature points of intracardiac electrogram signals. It outputs six feature points, including the starting point, first rising point, first falling point, peak point, valley point, and termination point, with average absolute errors of only 0.34, 0.27, 0.31, 0.01, 0.27, and 0.42 ms, respectively. The final detection results are highly satisfactory, indicating the potential utility of the findings in assisting the analysis of intracardiac electrogram signals. The research outcomes presented in this paper lay a foundation for the analysis of intracardiac electrogram signals.

关键词： Training Wavelet transforms Radio frequency Low voltage Training data medical services Myocardium Feature extraction Pattern recognition Testing

Recognition of temporal lobe epilepsy based on high throughput analysis of MRI images

学校读者我要写书评

暂无评论

Yi Qi Yi Biao Xue Bao/Chinese Journal of Scientific Instrument 2016年 37卷 78-82页

作者： Lian, Yuxi Yu, Jinhua Wang, Yuanyuan Shi, Zhifeng Chen, Liang Department of Electronic Engineering Fudan University Shanghai200433 China Key Laboratory of Medical Imaging Computing and Computer Assisted Intervention of Shanghai Shanghai200032 China Department of Neurosurgery Huashan Hospital Fudan University Shanghai200400 China

Temporal lobe epilepsy is a usual kind of epileptic *** recognition of temporal lobe epilepsyfrom MRI (magnetic resonance imaging) images can avoid the harm of radiation generated by PET (position emission computed tomography). An algorithm is proposed to recognize the temporal lobe epilepsybased onhigh throughput analysis of MRI ***, the brain MRI images are registered to the standardbrain atlas with SPM12 platform and the hippocampusregion is extracted as the ROI (Region of Interest) according to the AAL (anatomical automatic labeling) atlas in the same coordinate *** high throughput analysis is conducted to extract 540 image features from the ROI region. Finally, the Bagging method is used to distinguish normal persons and *** for 20 normal persons and 10 patients withtemporal lobe epilepsy are made to validate the proposed algorithm. The results indicate that the algorithm can obtain good classification performance with 100% accuracy. Hence, the proposed algorithm based on high throughout analysis and the Bagging classification method can be adopted to recognize temporal lobe epilepsy effectively. © 2016, Science Press. All right reserved.

关键词： Magnetic resonance imaging

DEEP RECURSIVE EMBEDDING FOR HIGH-DIMENSIONAL DATA

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zhou, Zixia Zu, Xinrui Wang, Yuanyuan Lelieveldt, Boudewijn P.F. Tao, Qian The Department of Electronic Engineering Fudan University Shanghai200433 China University of Twente Drienerlolaan 5 Enschede7522 NB Netherlands Key Laboratory of Medical Imaging Computing and Computer Assisted Intervention of Shanghai Shanghai200032 China The Division of Image Processing Department of Radiology Leiden University Medical Center Albinusdreef 2 Leiden2333 ZA Netherlands The Department of Imaging Physics Delft University of Technology Lorentzweg 1 Delft2628 CJ Netherlands

Embedding high-dimensional data onto a low-dimensional manifold is of both theoretical and practical value. In this paper, we propose to combine deep neural networks (DNN) with mathematics-guided embedding rules for high-dimensional data embedding. We introduce a generic deep embedding network (DEN) framework, which is able to learn a parametric mapping from high-dimensional space to low-dimensional space, guided by well-established objectives such as Kullback-Leibler (KL) divergence minimization. We further propose a recursive strategy, called deep recursive embedding (DRE), to make use of the latent data representations for boosted embedding performance. We exemplify the flexibility of DRE by different architectures and loss functions, and benchmarked our method against the two most popular embedding methods, namely, t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP). The proposed DRE method can map out-of-sample data and scale to extremely large datasets. Experiments on a range of public datasets demonstrated improved embedding performance in terms of local and global structure preservation, compared with other state-of-the-art embedding methods. © 2021, CC BY-NC-SA.

关键词： Deep neural networks

Deep recursive embedding for high-dimensional data

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zhou, Zixia Wang, Yuanyuan Lelieveldt, Boudewijn P.F. Tao, Qian The Department of Electronic Engineering Fudan University Shanghai200433 China The Department of Electronic Engineering Fudan University Shanghai200433 China Key Laboratory of Medical Imaging Computing and Computer Assisted Intervention of Shanghai Shanghai200032 China The Division of Image Processing Department of Radiology Leiden University Medical Center Albinusdreef 2 Leiden2333 ZA Netherlands The Division of Image Processing Department of Radiology Leiden University Medical Center Albinusdreef 2 Leiden2333 ZA Netherlands

t-distributed stochastic neighbor embedding (t-SNE) is a well-established visualization method for complex high-dimensional data. However, the original t-SNE method is nonparametric, stochastic, and often cannot well prevserve the global structure of data as it emphasizes local neighborhood. With t-SNE as a reference, we propose to combine the deep neural network (DNN) with the mathematical-grounded embedding rules for high-dimensional data embedding. We first introduce a deep embedding network (DEN) framework, which can learn a parametric mapping from high-dimensional space to low-dimensional embedding. DEN has a flexible architecture that can accommodate different input data (vector, image, or tensor) and loss functions. To improve the embedding performance, a recursive training strategy is proposed to make use of the latent representations extracted by DEN. Finally, we propose a two-stage loss function combining the advantages of two popular embedding methods, namely, t-SNE and uniform manifold approximation and projection (UMAP), for optimal visualization effect. We name the proposed method Deep Recursive Embedding (DRE), which optimizes DEN with a recursive training strategy and two-stage losse. Our experiments demonstrated the excellent performance of the proposed DRE method on high-dimensional data embedding, across a variety of public databases. Remarkably, our comparative results suggested that our proposed DRE could lead to improved global structure preservation. Copyright © 2021, The Authors. All rights reserved.

关键词： Clustering algorithms

EndoBoost: a plug-and-play module for false positive suppression during computer-aided polyp detection in real-world colonoscopy (with dataset)

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Wang, Haoran Zhu, Yan Qin, Wenzheng Zhang, Yizhe Zhou, Pinghong Li, Quanlin Wang, Shuo Song, Zhijian Digital Medical Research Center School of Basic Medical Sciences Fudan University Shanghai200032 China Shanghai Key Laboratory of Medical Image Computing and Computer Assisted Intervention Shanghai200032 China Endoscopy Center Endoscopy Research Institute Zhongshan Hospital Fudan University Shanghai200032 China Shanghai Collaborative Innovation Center of Endoscopy Shanghai200032 China School of Computer Science and Engineering Nanjing University of Science and Technology Jiangsu210014 China

The advance of computer-aided detection systems using deep learning opened a new scope in endoscopic image analysis. However, the learning-based models developed on closed datasets are susceptible to unknown anomalies in complex clinical environments. In particular, the high false positive rate of polyp detection remains a major challenge in clinical practice. In this work, we release the FPPD-13 dataset, which provides a taxonomy and real-world cases of typical false positives during computer-aided polyp detection in real-world colonoscopy. We further propose a post-hoc module EndoBoost, which can be plugged into generic polyp detection models to filter out false positive predictions. This is realized by generative learning of the polyp manifold with normalizing flows and rejecting false positives through density estimation. Compared to supervised classification, this anomaly detection paradigm achieves better data efficiency and robustness in open-world settings. Extensive experiments demonstrate a promising false positive suppression in both retrospective and prospective validation. In addition, the released dataset can be used to perform ‘stress’ tests on established detection systems and encourages further research toward robust and reliable computer-aided endoscopic image analysis. The dataset and code will be publicly available at http://***. © 2022, CC BY.

关键词： Anomaly detection

Predicting Ejection Fraction from Electrocardiogram Signals using a Multi-task Learning Model

学校读者我要写书评

暂无评论

Predicting Ejection Fraction from Electrocardiogram Signals ...

International Workshop on Wearable and Implantable Body Sensor Networks (BSN)

作者： Gaoyan Zhong Yueyi Wang Sen Liu Xintao Deng Aiguo Wang Cuiwei Yang Center for Biomedical Enginering School of Information Science and Technology Fudan Shanghai China Department of Cardiology Xinghua City People’s Hospital Jiangsu China Center for Biomedical Enginering School of Information Science and Technology Fudan The Key Laboratory of Medical Imaging Computing and Computer Assisted Intercention of Shanghai Shanghai China

The aim of this study was to investigate the feasibility and effectiveness of using electrocardiogram (ECG) signals to predict ejection fraction (EF) in an extremely imbalanced dataset. We collected ECG signals from 9365 patients and conducted a correlation analysis with EF. After collecting and preprocessing the ECG signal, we developed a deep learning-based multi-task learning model designed to extract features from ECG signals and perform predictions. Our study employed a model based on Transformer, multi-scale convolutional neural networks (CNN), channel attention, and pre-trained ResNet to predict EF (EF > 50 or EF ≤ 50) and EF value. The experimental results demonstrate that the proposed model exhibits excellent predictive performance, with an AUC of 0.825 and MAE of 4.855 for predicting EF. It outperforms other models and shows better results in comparison. Our study validates the feasibility and effectiveness of using ECG signals to predict EF and provides strong support for early diagnosis and treatment of cardiovascular diseases.

关键词：