Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文

Judgement and Disposal of Academic Misconduct Article Copyright Transfer Agreement Proof of Confidentiality Requirements for Electronic Version More.... Chinese Association of Automation National ResearchCenter for Intelligent Computing System Institute of Intelligent Machines,Chinese Academy of Sciences More....

2021 Vol.34 Issue.4, Published 2021-04-25 Intelligent Medical Treatment and Medical Image Processing        Intelligent Medical Treatment and Medical Image Processing 285 2021 Vol. 34 (4): 285-286 [Abstract] ( 483 ) [HTML 1KB] [ PDF 186KB] ( 502 ) 287 Deep Learning Based Medical Image Registration: A Review YING Shihui, YANG Wan, DU Shaoyi, SHI Jun Image registration is a key technology in the field of medical image processing and intelligent analysis. The real-time registration cannot be accomplished due to the high complexity and computational cost of traditional registration methods. With the development of deep learning, learning based image registration methods achieve remarkable results. In this paper, the medical image registration methods based on deep learning are systematically summarized and divided into three categories, including supervised learning, unsupervised learning and dual supervised learning. On this basis, the advantages and disadvantages for each category are discussed. Furthermore, the regularization methods proposed in recent years are emphatically discussed, especially based on diffeomorphism and multi-scale regularization. Finally, the medical image registration methods based on deep learning are prospected according to the development trend of the current medical image registration methods. 2021 Vol. 34 (4): 287-299 [Abstract] ( 1498 ) [HTML 1KB] [ PDF 987KB] ( 775 ) 300 Pancreas Segmentation Network for Abdominal CT Based on Compressive Sampling XU Qiangqiang, ZHANG Min, REN Fenggang, LÜ Yi, FENG Jun Due to the high anatomical variability of pancreas, it is difficult for automated segmentation algorithms to achieve accurate localization of the target. To solve this problem, an encoder-decoder network embedded with compressive sampling is proposed. By training the network in different stages, the segmentation network can cascade the prior knowledge of pancreas location perceived from the label space in the pre-trained stage. Thus, the precise positioning of the pancreas is realized and the consistency between the segmentation result and the label is ensured. The experimental results of pancreas segmentation show that the performance of the proposed network is better. 2021 Vol. 34 (4): 300-310 [Abstract] ( 445 ) [HTML 1KB] [ PDF 2637KB] ( 385 ) 311 Feature Selection Method for Neuropsychiatric Disorder Based on Adaptive Sparse Structure Learning HAO Shijie, GUO Yanrong, CHEN Tao, WANG Meng, HONG Richang In the research of computer-aided diagnosis techniques for neuropsychiatric diseases, professionals are required to perform diagnostic-level semantic annotations on samples, and it is time-consuming and labor-intensive. Therefore, it is of great importance to develop unsupervised techniques for the computer-aided diagnosis on neuropsychiatric diseases. In this paper, an unsupervised feature selection method based on adaptive sparse structure learning is proposed and applied to the task of diagnosis on Schizophrenia and Alzheimer′s disease. The sparse representation and the data manifold structure are simultaneously learned in a unified framework. In this framework, the generalized norm is adopted to model the reconstruction error of sparse learning. The manifold structure of the whole dataset is iteratively updated. The lacking of robustness in the traditional feature selection methods is relieved. Experiments on two public datasets of Schizophrenia and Alzheimer′s disease demonstrate the effectiveness of the proposed method in classification of neuropsychiatric diseases. 2021 Vol. 34 (4): 311-321 [Abstract] ( 506 ) [HTML 1KB] [ PDF 1376KB] ( 491 ) 322 A Scalable Local Analysis and Integration Approach to Intrinsic Image Decomposition SHI Xue, XU Haiping, LI Chunming A unified mathematical model and an algorithm are proposed to solve the problems of the estimation of illumination and reflectance images of a natural image and the segmentation and bias field estimation of a magnetic resonance image(MRI). The proposed model only requires a basic assumption that the observed image can be approximated by the product of two intrinsic images with different properties. One of the two intrinsic images is a smooth image, S-image, and the other is a piece-wise approximately constant image, L-image. To fully exploit the properties of the intrinsic images, a scalable local analysis and integration(SLAI) approach is proposed for the problem of intrinsic image estimation. Due to the smoothness of the S-image, a low order Taylor expansion or a linear combination of general smooth basis functions is utilized to locally approximate the S-image. The obtained local smooth approximation of the S-image can be extended to a smooth image on the entire region of interest(ROI) using partition of unity subordinate to a cover of ROI. Meanwhile, the segmentation result and the estimation of the L-image are obtained. The proposed method is based on a weaker assumption than the methods in the literature, and therefore it is applicable to more images. The proposed method produces satisfactory results on MR images and natural images. 2021 Vol. 34 (4): 322-332 [Abstract] ( 349 ) [HTML 1KB] [ PDF 2252KB] ( 552 ) 333 A Patient-Specific Method for Epileptic Seizure Prediction During Sleep Based on Deep Neural Network CHENG Chenchen, YOU Bo, LIU Yan, DAI Yakang The existing epileptic seizure prediction methods present the problems of low accuracy, high false alarm rate, sleep electroencephalogram(EEG) specificity of epileptic patients and differences in EEG signals caused by differences in the location and type of epileptic foci . In this paper, a patient-specific method for epileptic seizure prediction during sleep based on deep neural network is proposed to help doctors and patients to take timely and effective treatment measures. Consequently, the probability of patients suffering from complications and sudden death is reduced. The original EEG signals are filtered and segmented to remove noise and trigger the alarm in a short time. Discrete wavelet transform is utilized to decompose the EEG, and statistical features are extracted to reveal the time-frequency characteristics of EEG signals. Then, the bi-direction long-short term memory(Bi-LSTM) is employed to mine the most discriminative features combined with the leave-one-out method for classification. The prediction results are obtained after the optimization of the decision-making process. Experiments with different frequency band restrictions show that the δ band signal related to sleep epilepsy affects the prediction performance and the performance of the proposed method is better than the existing sleep epileptic seizure prediction methods. 2021 Vol. 34 (4): 333-342 [Abstract] ( 616 ) [HTML 1KB] [ PDF 886KB] ( 533 ) 343 Traditional Chinese Medicine Aided Diagnosis and Treatment System for Rheumatoid Arthritis Based on Artificial Intelligence SUN Mingjun, ZHANG Dan, ZHENG Mingzhi, MEI Shuhuan Rheumatoid arthritis(RA) is a widespread, chronic and refractory systemic immune rheumatism. Traditional Chinese medicine(TCM) presents the advantages of less side effects and low price. However, the spread of RA TCM diagnosis and treatment scheme with curative effect advantages is limited due to the lack of experienced TCM practitioners, especially in primary medical institutions. In this paper, a traditional Chinese medicine aided diagnosis and treatment system for RA based on artificial intelligence is proposed. RA and pattern of syndrome in RA can be determined after learning patient medical records and medical imaging of joints, and then TCM prescription is recommended intelligently according to the pattern. Next, the information is exploited to assist doctors in diagnosis. Based on RA TCM knowledge, the knowledge graph is built. It provides doctors with knowledge guidance in the process of diagnosis and treatment. The system can assist less experienced doctors in making treatment decisions, improving the treatment level of RA, and studying and promoting RA treatment. 2021 Vol. 34 (4): 343-352 [Abstract] ( 1173 ) [HTML 1KB] [ PDF 1382KB] ( 653 ) 353 Multimodal Corpus Construction Based on Medical Image Segmentation Algorithm LIN Yuping, ZHENG Yaoyue, ZHENG Haojie, ZHANG Dong, WANG Cong, LI Xiaomian, LI Yingyu, TIAN Zhiqiang Electronic medical records(EMRs) corpus provides qualitative diagnosis results of related medical images. However, the good management of medical data may be affected due to the lacking of labeled images and texts and it is hard for medical students to acquire the related medical knowledge independently. To solve this problem, a medical image segmentation method based on the deep level set algorithm is proposed to segment medical images automatically and output contour results of the interested area and related quantitative indicators. Electronic medical record text is annotated grounded on natural language processing methods. The information representation of medical record texts and images of multimodal corpus is enhanced. Experimental results on the glaucoma image dataset show that the proposed method segments the optic disc and the optic cup in the fundus image accurately and a multimodal corpus with self-evident labeled images and EMRs is constructed effectively as well. 2021 Vol. 34 (4): 353-360 [Abstract] ( 601 ) [HTML 1KB] [ PDF 1577KB] ( 314 ) 361 Self-supervised Edge-Fusion Network for MRI Reconstruction LI Zhongnian, ZHANG Tao, ZHANG Daoqiang The research on compressed sensing magnetic resonance imaging(CS-MRI) suggests that the edge information is the hardest part of medical image reconstruction. In most deep-learning based methods, the explicit consideration for edge information is not taken into account. To tackle this problem, a self-supervised edge-fusion network(SEN) is proposed to explore beneficial edge properties to reconstruct MRI. Firstly, edge annotations are generated by utilizing canny edge detector without involving any time-consuming and expensive human labeling. Secondly, a self-supervised auxiliary network is introduced to incorporate edge annotations into a feature learning to capture fusible representations. A top-down fusion strategy is proposed to fuse the learned representations into reconstruction network for CS-MRI restoring. Experimental results show that SEN catches the edge information effectively and achieves better performance in CS-MRI reconstruction. 2021 Vol. 34 (4): 361-366 [Abstract] ( 397 ) [HTML 1KB] [ PDF 2564KB] ( 415 ) 367 Optimizing k-Space Subsampling Pattern toward MRI Reconstruction XUAN Kai, WANG Qian Imaging velocity is a major factor affecting clinical applications of magnetic resonance(MR) imaging. And an effective solution of reducing scanning time is to under-sample in k-space and reconstruct the image from under-sampled MR signals. In this paper, the impact of under-sampling pattern on reconstruction quality is analyzed and a joint optimization strategy is proposed to update the under-sampling pattern with image reconstruction model in the context of deep-learning. To optimize the non-continuous under-sampling pattern, it is firstly initialized with full-sampling pattern. Then, relatively less important phase-encodings are gradually pruned until the sparsity requirement in k-space is satisfied. And the optimization of k-space under-sampling pattern is conducted alternatively with that of the reconstruction model. Moreover, the relative importance is estimated with the weight by assigning weight to each phase-coding. Experiments demonstrate that the proposed method improves the quality of the reconstructed MR image compared with the proposed method. 2021 Vol. 34 (4): 367-374 [Abstract] ( 816 ) [HTML 1KB] [ PDF 1883KB] ( 845 ) 375 Semantic Segmentation Network of Pathological Images of Liver Tissue Based on Multi-scale Feature and Attention Mechanism ZHANG Aoqi, KANG Yuxin, WU Zhuoyue, CUI Lei, BU Qirong To address the problem of difficult segmentation and many voids in the transition regions of normal and abnormal tissues in liver histopathology images segmentation, a semantic segmentation network of pathological images of liver tissue based on multi-scale feature and attention mechanism is proposed. The fused multi-scale features are extracted in the encoder to improve the segmentation of the transition regions between normal and abnormal tissues. The attention mechanism is employed to model the correlation between spatial dimension and channel dimension to obtain the response of each pixel within each class as well as the dependency between channels, and the impact of many voids in liver histopathology images on the network learning is alleviated. Experiments demonstrate that the proposed network can segment the damaged regions of liver histopathology images more quickly and accurately. 2021 Vol. 34 (4): 375-384 [Abstract] ( 631 ) [HTML 1KB] [ PDF 4471KB] ( 437 )

Supervised by China Association for Science and Technology Sponsored by Chinese Association of Automation NationalResearchCenter for Intelligent Computing System Institute of Intelligent Machines, Chinese Academy of Sciences Published by Science Press

Copyright © 2010 Editorial Office of Pattern Recognition and Artificial Intelligence Address: No.350 Shushanhu Road, Hefei, Anhui Province, P.R. China Tel: 0551-65591176 Fax:0551-65591176 Email: bjb@iim.ac.cn Supported by Beijing Magtech 　Email:support@magtech.com.cn