Article
Computer Science, Artificial Intelligence
Jiachen Yang, Yue Yang, Yang Li, Zhuo Zhang, Jiabao Wen
Summary: With the development of data-centric artificial intelligence, attention is being paid to the importance of image information quality. A two-stage image quality assessment method is proposed based on the idea that images in datasets have different intra-class information richness and inter-class information overlaps. Experiments on two public image classification datasets show that the proposed method can effectively distinguish high quality images and achieve better performances.
NEURAL PROCESSING LETTERS
(2023)
Article
Multidisciplinary Sciences
Mehmet Koc, Semih Ergin, Mehmet Bilginer Gulmezoglu, Mehmet Fidan, Omer Nezih Gerek, Atalay Barkana
Summary: This study investigates the classification performances and characteristics of two variations of the common vector approach (CVA) for images with shared backgrounds, including binary images. It finds that the discriminative CVA method carries certain risks when the dimension of the feature vector is lower than the number of training samples. Additionally, it observes that CVA outperforms its discriminative version in the classification of binary images. The study highlights the importance of considering the training data size when applying CVA methods.
ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING
(2022)
Article
Computer Science, Information Systems
Jie Shi, Zhengyu Li, Hong Zhao
Summary: In this paper, a hierarchical feature selection method is proposed to improve classification difficulty by maximizing inter-class independence and minimizing intra-class redundancy using structure and feature relations. The method utilizes the hierarchy in the class space as structural information to improve performance and transforms the feature correlations into a mathematical representation to minimize redundancy.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Lituan Wang, Lei Zhang, Xin Shu, Zhang Yi
Summary: In this article, a deep learning method is proposed to enhance the intra-class consistency and inter-class discrimination in automatic skin lesion classification. A CAM-based global-lesion localization module is introduced to optimize the distance of CAMs generated by different skin lesion tasks. Additionally, a global features guided intra-class similarity learning module is proposed to generate class centers. Experimental results demonstrate the effectiveness and generalizability of the proposed method on ISIC-2017 and ISIC-2018 datasets.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Computer Science, Information Systems
Jianping Gou, Xia Yuan, Baosheng Yu, Jiali Yu, Zhang Yi
Summary: This study proposes an intra- and inter-class induced discriminative deep dictionary learning (DDDL) method, which introduces two discriminative constraints on deep dictionary learning: intra-class compactness and inter-class separability. By optimizing layer-wise data representation and learning the deepest representation for classification, the proposed DDDL model outperforms recent shallow and deep representation learning approaches.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Multidisciplinary Sciences
Sivaramakrishnan Rajaraman, Prasanth Ganesan, Sameer Antani
Summary: In this study, the effect of model calibration on the performance of medical image classification tasks was systematically analyzed. The results show that calibration can significantly improve performance at the default classification threshold, but the differences are not significant at the PR-guided threshold. This observation holds for different image modalities and degrees of class imbalance.
Article
Computer Science, Artificial Intelligence
Pratik Mazumder, Mohammed Asad Karim, Indu Joshi, Pravendra Singh
Summary: This work shows that the change in orientation of an image affects the model prediction accuracy and proposes a data-ensemble approach to retain information about previously seen classes. It also introduces a novel training approach using a joint-incremental learning objective (JILO) that is vital to the data-ensemble method. The proposed approach significantly improves the performance of state-of-the-art class-incremental learning methods on the CIFAR-100 dataset.
Article
Computer Science, Artificial Intelligence
Benteng Ma, Yu Feng, Geng Chen, Changyang Li, Yong Xia
Summary: Medical data sharing is crucial but suffers from privacy issues. This paper proposes a novel federated learning algorithm, FedAR, which addresses data heterogeneity by employing a flexible re-weighting scheme and achieves superior performance.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Artificial Intelligence
Hassan Alhuzali, Sophia Ananiadou
Summary: Textual Emotion Recognition (TER) is a crucial task in NLP with significant real-world applications. Previous research focused on maximizing the probability of correct emotion classification, but ignored intra- and inter-class variations. To address this, we propose a triplet center loss as an auxiliary task for emotion classification, enabling TER models to learn compact and discriminative features. Our method also evaluates the impact of intra- and inter-class variations on each emotion class. Experimental results on three datasets demonstrate the effectiveness of our approach compared to previous methods, improving prediction scores and producing discriminative features.
IEEE TRANSACTIONS ON AFFECTIVE COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Yuefan Xu, Sen Zhang, Wendong Xiao
Summary: The variability in ECG patterns among patients and the class imbalance problem pose challenges in ECG recognition. To address these issues, a novel algorithm called ICC-WKELM is proposed for heartbeat multiclass classification. A compact and discriminative feature set is constructed, and a kernel extreme learning machine (KELM) is used for heartbeat classification. The class imbalance is measured using intra-class coherence (ICC), and a weight assignment strategy is designed for imbalanced arrhythmia classes. The proposed approach achieves high F1 scores and overall accuracy on the MIT-BIH arrhythmia dataset.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Software Engineering
Qian Jiang, Ziyu Zhang, Feipeng Da, Shaoyan Gai
Summary: This paper proposes a novel architecture for automatic facial expression recognition which improves the recognition accuracy for difficult expressions. Experimental results show that the proposed network outperforms other state-of-the-art methods on multiple datasets.
Article
Environmental Sciences
Pingping Liu, Xiaofeng Liu, Yifan Wang, Zetong Liu, Qiuzhan Zhou, Qingliang Li
Summary: With the rapid growth of remote sensing image data, there is an increased demand for remote sensing image retrieval. In this paper, a new sample generation mechanism is proposed to generate samples that meet boundary constraints and obtain quantifiable intra-class differences. An intra-class ranking loss function is designed to improve the discriminability of the generated embedding space and maintain the ranking relationship. Extensive experiments on multiple remote-sensing image datasets demonstrate the effectiveness of the proposed approach in improving the performance of remote sensing image retrieval.
Article
Engineering, Multidisciplinary
Yang Xu, Yuequan Bao, Yufeng Zhang, Hui Li
Summary: A novel nested attribute-based few-shot meta learning paradigm is proposed for structural damage identification, which utilizes external meta learning and internal attribute-based transfer learning to achieve robust classification and knowledge transfer. Validation on a real-world dataset demonstrates the approach's high accuracy and robustness across different damage types. The proposed paradigm outperforms regular supervised learning by generating higher accuracy and better equilibrium between precision and recall, showcasing its stability and reliability in training a meta learning classification model.
STRUCTURAL HEALTH MONITORING-AN INTERNATIONAL JOURNAL
(2021)
Article
Chemistry, Analytical
Huibin Li, Wei Guo, Guowen Lu, Yun Shi
Summary: This study addresses the issue of current lightweight detection models being ineffective in detecting multi-type occlusion targets during fruit picking. By introducing a multi-type occlusion apple dataset and a data balance augmentation method, the study shows significant improvement in the average detection precision of popular lightweight object detection models. The proposed augmentation method demonstrates great potential for future orchard applications in different fruit detection missions.
Article
Biology
Yaxiong Chen, Yibo Tang, Jinghao Huang, Shengwu Xiong
Summary: In this paper, a new Multi-scale Triplet Hashing (MTH) algorithm is proposed, which leverages multi-scale information, convolutional self-attention, and hierarchical similarity to learn effective hash codes simultaneously. The MTH algorithm addresses the problems of feature learning, the neglect of discriminate area, and hierarchical similarity in medical image retrieval.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Computer Science, Artificial Intelligence
Yutong Xie, Jianpeng Zhang, Zhibin Liao, Johan Verjans, Chunhua Shen, Yong Xia
Summary: In this paper, a semi-supervised model based on intra- and inter-pair consistency is proposed for gland segmentation in histology tissue images. By utilizing the relationships between different images in the feature space and imposing consistency constraints, the model achieves improved accuracy.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Computer Science, Interdisciplinary Applications
Zehui Liao, Yutong Xie, Shishuai Hu, Yong Xia
Summary: In this paper, a multi-view 'divide-and-rule' model is proposed to learn from reliable and ambiguous annotations for lung nodule malignancy prediction. The nodules are divided into three sets based on the consistency and reliability of the annotations. The proposed model consists of three DAR models and is trained following a two-stage procedure. Experimental results show the effectiveness and superiority of the model in learning from ambiguous labels and predicting lung nodule malignancy.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2022)
Article
Computer Science, Information Systems
Mengyang Sun, Wei Suo, Peng Wang, Yanning Zhang, Qi Wu
Summary: This paper presents a proposal-free one-stage (PFOS) framework that can directly regress the region-of-interest from the image or generate unambiguous descriptions in an end-to-end manner. By taking the dense-grid of images as input and using a cross-attention transformer, the model learns multi-modal correspondences and eliminates the need for additional annotations or off-the-shelf detectors in the mainstream two-stage methods. Furthermore, the traditional two-stage listener-speaker framework is expanded to be jointly trained by a one-stage learning paradigm, resulting in state-of-the-art performance on comprehension and competitive results for generation.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Computer Science, Artificial Intelligence
Chaorui Deng, Qi Wu, Qingyao Wu, Fuyuan Hu, Fan Lyu, Mingkui Tan
Summary: Visual grounding aims to locate the most relevant object or region in an image based on natural language queries. This paper proposes an attention module to reduce internal redundancies and an accumulated attention mechanism to capture the relationship among different kinds of information. Additionally, noise is introduced to bridge the distribution gap between human-labeled training data and real-world poor quality data, improving the performance and robustness of the VG models. Experimental results demonstrate the superiority of the proposed methods on various datasets in terms of accuracy.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Wei Suo, Mengyang Sun, Peng Wang, Yanning Zhang, Qi Wu
Summary: Referring Expression Comprehension (REC) is a crucial task in the vision-and-language community, and it plays a vital role in various cross-modal tasks. Existing research focuses on a one-stage paradigm, treating REC as a language-conditioned object detection task to achieve a balance between speed and accuracy. However, previous frameworks overlook the importance of integrating multi-level features and often rely on single-scale features for target localization.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Yanyuan Qiao, Yuankai Qi, Yicong Hong, Zheng Yu, Peng Wang, Qi Wu
Summary: This paper proposes an enhanced and history-aware pre-training method for Vision-and-Language Navigation (VLN), which introduces three novel VLN-specific proxy tasks and a memory network to improve historical knowledge learning and action prediction. The proposed method achieves new state-of-the-art performance on four downstream VLN tasks.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Engineering, Electrical & Electronic
Mengge He, Wenjing Du, Zhiquan Wen, Qing Du, Yutong Xie, Qi Wu
Summary: In this paper, a Multi-Granularity Aggregation Transformer (MGAT) is proposed for joint video-audio-text representation learning. The method overcomes the limitations of existing methods by designing a multi-granularity transformer module and an attention-guided aggregation module. The aggregated information is aligned with text information at different hierarchical levels using consistency loss and contrastive loss. Experimental results demonstrate the superiority of the proposed method on tasks such as video-paragraph retrieval and video captioning.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Hao Li, Jinfa Huang, Peng Jin, Guoli Song, Qi Wu, Jie Chen
Summary: TextVQA aims to produce correct answers for questions about images with multiple scene texts. This paper introduces 3D geometric information into the spatial reasoning process to capture contextual knowledge. Experimental results show that the proposed method achieves state-of-the-art performance on TextVQA and ST-VQA datasets.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Cybernetics
Zihan Wang, Olivia Byrnes, Hu Wang, Ruoxi Sun, Congbo Ma, Huaming Chen, Qi Wu, Minhui Xue
Summary: The use of deep learning techniques in data hiding has greatly advanced secure communication and identity verification fields. Digital watermarking and steganography techniques, by embedding information into noise-tolerant signals like audio, video, or images, can protect sensitive intellectual property (IP) and enable confidential communication for authorized parties. This survey provides a systematic overview of recent developments in deep learning techniques for data hiding, based on model architectures and noise injection methods. It also suggests and discusses potential future research directions that combine digital watermarking and steganography in software engineering to enhance security and mitigate risks. This contribution promotes the creation of a more trustworthy digital world and advances responsible artificial intelligence (AI).
IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Yutong Xie, Jianpeng Zhang, Lingqiao Liu, Hu Wang, Yiwen Ye, Johan Verjans, Yong Xia
Summary: This paper proposes a hybrid pre-training paradigm that combines self-supervised learning and supervised learning to improve the representation quality for medical image segmentation tasks. It introduces a reference task in self-supervised learning and optimizes the model using a gradient matching method. The experimental results demonstrate the effectiveness of this approach on multiple medical image segmentation benchmarks.
MEDICAL IMAGE ANALYSIS
(2024)
Proceedings Paper
Computer Science, Artificial Intelligence
Yutong Xie, Jianpeng Zhang, Yong Xia, Qi Wu
Summary: This paper introduces the application of self-supervised learning in medical image analysis and proposes a universal medical self-supervised representation learning framework called UniMiSS, which utilizes 2D images to compensate for the lack of 3D data. To enable self-supervised learning with both 2D and 3D images, the paper designs a medical Transformer (MiT) and trains it using self-distillation. Experiments demonstrate that UniMiSS achieves promising performance on various medical image analysis tasks.
COMPUTER VISION, ECCV 2022, PT XXI
(2022)
Article
Computer Science, Information Systems
Amin Parvaneh, Ehsan Abbasnejad, Qi Wu, Javen Qinfeng Shi, Anton van den Hengel
Summary: This study proposes a modular deep neural network called Price Negotiator to improve negotiation in online shopping. It addresses the challenges by considering item images, finding similar items, predicting price actions, and adjusting prices based on predicted actions.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Computer Science, Information Systems
Zeren Sun, Huafeng Liu, Qiong Wang, Tianfei Zhou, Qi Wu, Zhenmin Tang
Summary: This paper proposes an end-to-end framework named Co-LDL for addressing the performance degradation of deep neural networks caused by label noise. The framework incorporates the low-loss sample selection strategy with label distribution learning and trains two deep neural networks simultaneously to communicate useful knowledge. Additionally, a self-supervised module is introduced to enhance the learned representations.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Computer Science, Information Systems
Chuanyi Zhang, Qiong Wang, Guosen Xie, Qi Wu, Fumin Shen, Zhenmin Tang
Summary: This article introduces a method for learning fine-grained tasks from web data, which purifies noisy training sets by identifying and distinguishing noisy images, and trains models to alleviate the effects of noise.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Computer Science, Artificial Intelligence
Hu Wang, Hao Chen, Qi Wu, Congbo Ma, Yidong Li
Summary: The control of traffic signals is crucial in relieving traffic congestion in urban areas. However, it is difficult due to the complexity of real-world traffic dynamics. To address this, the researchers propose a new dataset and a novel model based on deep reinforcement learning for optimizing multi-intersection traffic control. The experimental results show that the proposed model outperforms other methods.
IEEE OPEN JOURNAL OF INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Hong Liu, Dong Wei, Donghuan Lu, Xiaoying Tang, Liansheng Wang, Yefeng Zheng
Summary: This study proposes a framework based on hybrid 2D-3D convolutional neural networks for obtaining continuous 3D retinal layer surfaces from OCT volumes. The framework works well with both full and sparse annotations and utilizes alignment displacement vectors and layer segmentation to align the B-scans and segment the layers. Experimental results show that the framework outperforms state-of-the-art 2D deep learning methods in terms of layer segmentation accuracy and cross-B-scan 3D continuity.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Simon Oxenford, Ana Sofia Rios, Barbara Hollunder, Clemens Neudorfer, Alexandre Boutet, Gavin J. B. Elias, Jurgen Germann, Aaron Loh, Wissam Deeb, Bryan Salvato, Leonardo Almeida, Kelly D. Foote, Robert Amaral, Paul B. Rosenberg, David F. Tang-Wai, David A. Wolk, Anna D. Burke, Marwan N. Sabbagh, Stephen Salloway, M. Mallar Chakravarty, Gwenn S. Smith, Constantine G. Lyketsos, Michael S. Okun, William S., Zoltan Mari, Francisco A. Ponce, Andres Lozano, Wolf-Julian Neumann, Bassam Al-Fatly, Andreas Horn
Summary: Spatial normalization is a method to map subject brain images to an average template brain, allowing comparison of brain imaging results. We introduce a novel tool called WarpDrive, which enables manual refinements of image alignment after automated registration. The tool improves accuracy of data representation and aids in understanding patient outcomes.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Ricards Marcinkevics, Patricia Reis Wolfertstetter, Ugne Klimiene, Kieran Chin-Cheong, Alyssia Paschke, Julia Zerres, Markus Denzinger, David Niederberger, Sven Wellmann, Ece Ozkan, Christian Knorr, Julia E. Vogt
Summary: This study presents interpretable machine learning models for predicting the diagnosis, management, and severity of suspected appendicitis using ultrasound images. The proposed models utilize concept bottleneck models (CBM) that facilitate interpretation and intervention by clinicians, without compromising performance or requiring time-consuming image annotation.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Jian-Qing Zheng, Ziyang Wang, Baoru Huang, Ngee Han Lim, Bartlomiej W. Papiez
Summary: This article introduces a new method for medical image registration, which utilizes a separable motion backbone and a residual aligner module to better handle the discontinuous motion of multiple neighboring objects. The proposed method achieves excellent registration results on abdominal CT scans and lung CT scans.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Xiangqiong Wu, Guanghua Tan, Hongxia Luo, Zhilun Chen, Bin Pu, Shengli Li, Kenli Li
Summary: This study develops a user-friendly framework for the automated diagnosis of thyroid nodules in ultrasound videos, simulating the diagnostic workflow of radiologists. By interpreting image characteristics and modeling temporal contextual information, the efficiency and generalizability of the diagnosis can be improved.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Riddhish Bhalodia, Shireen Elhabian, Jadie Adams, Wenzheng Tao, Ladislav Kavan, Ross Whitaker
Summary: This paper introduces DeepSSM, a deep learning-based framework for image-to-shape modeling. By learning the functional mapping from images to low-dimensional shape descriptors, DeepSSM can directly infer statistical representation of anatomy from 3D images. Compared to traditional methods, DeepSSM eliminates the need for heavy manual preprocessing and segmentation, and significantly improves computational time.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Florentin Liebmann, Marco von Atzigen, Dominik Stutz, Julian Wolf, Lukas Zingg, Daniel Suter, Nicola A. Cavalcanti, Laura Leoty, Hooman Esfandiari, Jess G. Snedeker, Martin R. Oswald, Marc Pollefeys, Mazda Farshad, Philipp Furnstahl
Summary: This study presents a marker-less approach for automatic registration and real-time navigation of lumbar spinal fusion surgery using a deep neural network, avoiding radiation exposure and surgical errors. The method was validated on an ex-vivo surgery and a public dataset.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Piyush Tiwary, Kinjawl Bhattacharyya, A. P. Prathosh
Summary: Domain shift refers to the change of distributional characteristics between training and testing datasets, leading to performance drop. For medical image tasks, domain shift can be caused by changes in imaging modalities, devices, and staining mechanisms. Existing approaches based on generative models suffer from training difficulties and lack of diversity. In this paper, the authors propose the use of energy-based models (EBMs) for unpaired image-to-image translation in medical images. The proposed method, called Cycle Consistent Twin EBMs (CCT-EBM), employs a pair of EBMs in the latent space of an Auto-Encoder to ensure translation symmetry and coupling between domains.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Yutong Xie, Jianpeng Zhang, Lingqiao Liu, Hu Wang, Yiwen Ye, Johan Verjans, Yong Xia
Summary: This paper proposes a hybrid pre-training paradigm that combines self-supervised learning and supervised learning to improve the representation quality for medical image segmentation tasks. It introduces a reference task in self-supervised learning and optimizes the model using a gradient matching method. The experimental results demonstrate the effectiveness of this approach on multiple medical image segmentation benchmarks.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Youyi Song, Jing Zou, Kup-Sze Choi, Baiying Lei, Jing Qin
Summary: Cell classification is crucial for intelligent cervical cancer screening, but the variation in cells' appearance and shape poses challenges. A new learning algorithm, worse-case boosting, is proposed to improve classification accuracy for under-represented data. Experimental results demonstrate the effectiveness of this algorithm in two publicly available datasets, achieving a 4% improvement in accuracy.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Sangjoon Park, Eun Sun Lee, Kyung Sook Shin, Jeong Eun Lee, Jong Chul Ye
Summary: The increasing demand for AI systems to monitor human errors and abnormalities in healthcare presents challenges. This study presents a model called Medical X-VL, which is tailored for the medical domain and outperformed current state-of-the-art models in two medical image datasets. The model enables various zero-shot tasks for monitoring AI in the medical domain.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Anna Klimovskaia Susmelj, Berkan Lafci, Firat Ozdemir, Neda Davoudi, Xose Luis Dean-Ben, Fernando Perez-Cruz, Daniel Razansky
Summary: Optoacoustic imaging is a technique that uses optical excitation and ultrasound detection for biological tissue imaging. The quality of the images depends on the extent of tomographic coverage provided by the ultrasound detector arrays. However, full coverage is not always possible due to experimental constraints. The proposed signal domain adaptation network aims to reduce limited-view artifacts in the images.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Srijay Deshpande, Muhammad Dawood, Fayyaz Minhas, Nasir Rajpoot
Summary: In this work, a novel framework called SynCLay is proposed for automated synthesis of histology images based on user-defined cellular layouts. The framework can generate realistic and high-quality histology images with different cellular arrangements, which is helpful for studying the role of cells in the tumor microenvironment. The framework integrates a nuclear segmentation and classification model to refine nuclear structures and generate nuclear masks. Evaluation using quantitative metrics and feedback from pathologists shows that the synthetic images generated by SynCLay have high realism scores and can accurately differentiate between benign and malignant tumors.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Ahmed H. Shahin, An Zhao, Alexander C. Whitehead, Daniel C. Alexander, Joseph Jacob, David Barber
Summary: Survival analysis is a valuable tool in healthcare for predicting the time to specific events. This paper introduces CenTime, a novel approach that directly estimates the time to event. The method performs well with censored data and can be easily integrated with deep learning models. Compared to standard methods, CenTime offers superior performance in predicting event time while maintaining comparable ranking performance.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Computer Science, Artificial Intelligence
Bingyuan Liu, Jose Dolz, Adrian Galdran, Riadh Kobbi, Ismail Ben Ayed
Summary: Most segmentation losses, such as CE and Dice, are variants of the Cross-Entropy or Dice losses. This work provides a theoretical analysis that shows a deeper connection between CE and Dice than previously thought. From a constrained-optimization perspective, both CE and Dice decompose into similar ground-truth matching terms and region-size penalty terms. The analysis uncovers hidden region-size biases: Dice has an intrinsic bias towards extremely imbalanced solutions, while CE implicitly encourages the ground-truth region proportions. Based on this analysis, a principled and simple solution is proposed to explicitly control the region-size bias.
MEDICAL IMAGE ANALYSIS
(2024)
