Article
Engineering, Industrial
Xiaobo Zhang, Zhenzhou Lu, Kai Cheng
Summary: This paper presents a cross-entropy-based directional importance sampling (CE-DIS) method for rare failure probability estimation in structural reliability analysis. By choosing a quasi-optimal sampling density, the CE-DIS method significantly improves computational efficiency and can be applied to different types of reliability problems.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2022)
Article
Multidisciplinary Sciences
Yasser Aleman-Gomez, Alessandra Griffa, Jean-Christophe Houde, Elena Najdenovska, Stefano Magon, Meritxell Bach Cuadra, Maxime Descoteaux, Patric Hagmann
Summary: In this work, a whole-brain multi-scale structural connectome atlas is presented, which can provide valuable network information for imaging studies. This tool is derived from healthy subject data, using extensively validated processing and segmentation tools, and it offers user-friendly code to extract connection-specific quantitative information from individual brain imaging data. This method contributes to analyzing the network-level consequences of regional changes.
Article
Neurosciences
Antonia Hain, Daniel Jorgens, Rodrigo Moreno
Summary: Tractography is important for brain connectivity studies, but currently faces reliability problems. A method called SIFT has been developed to remove anatomically implausible connections in tractograms, but it is not suitable for individual streamline assessment. To address this, researchers propose applying SIFT to randomly selected tractogram subsets to retrieve multiple assessments and train a classifier for distinguishing compliant and non-compliant streamlines, achieving over 80% accuracy.
Article
Computer Science, Artificial Intelligence
Jon Haitz Legarreta, Laurent Petit, Francois Rheault, Guillaume Theaud, Carl Lemaire, Maxime Descoteaux, Pierre-Marc Jodoin
Summary: The study presents a novel filtering method based on autoencoders for diffusion MRI tractography, which can achieve more reliable results than traditional methods with superior filtering performance and generalization capabilities.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Computer Science, Artificial Intelligence
Yu Yao, Peng Yang, Guangzhen Zhao, Yanyan Ge, Ying Yang
Summary: Keyphrase generation is a fundamental task in NLP, and existing methods mainly optimize the distribution to generate keyphrases. However, they neglect the manipulation of copy and generating spaces, reducing the decoder's generability. This article proposes a probabilistic keyphrase generation model that leverages a variational encoder-decoder framework and separate latent variables for the copy and generating spaces. Experimental results on social media and scientific article datasets demonstrate the superiority of the proposed model in generating accurate predictions and controllable keyphrase numbers.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Shengxi Li, Danilo Mandic
Summary: This article introduces a novel approach based on the von-Mises-Fisher (vMF) distribution to obtain an explicit and simple probability representation of skewed elliptical distributions. This method allows for the design and implementation of nonsymmetric learning systems and provides a physically meaningful and intuitive way of generalizing skewed distributions.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Neurosciences
Johannes Gruen, Samuel Groeschel, Thomas Schultz
Summary: Tractography based on diffusion Magnetic Resonance Imaging (dMRI) is the prevalent approach to the in vivo delineation of white matter tracts in the human brain. We introduce two novel approaches to make multi-fiber tractography more stable by using spatial regularization. These methods represent the fiber Orientation Distribution Function (fODF) as a symmetric fourth-order tensor and recover multiple fiber orientations via low-rank approximation.
Article
Computer Science, Artificial Intelligence
Fabrice Rossi, Florian Barbaro
Summary: The article presents a method for clustering data on the unit hypersphere using mixtures of von Mises-Fisher distributions, which is particularly suitable for high-dimensional directional data. By estimating a sparse von Mises mixture using a penalized likelihood, the clustering interpretability is improved. The approach is evaluated on simulated and real data benchmarks, showing its advantages. Additionally, a new dataset on financial reports is introduced, highlighting the benefits of the method for exploratory analysis.
Review
Engineering, Biomedical
Joseph Yuan-Mou Yang, Chun-Hung Yeh, Cyril Poupon, Fernando Calamante
Summary: Diffusion magnetic resonance imaging (dMRI) tractography is currently the only imaging technique allowing for non-invasive delineation and visualization of white matter (WM) tracts in vivo, leading to rapid advances in brain MRI research. Its major clinical application lies in pre-surgical planning and intraoperative image guidance in neurosurgery, providing support for surgical resection guidance and post-surgical outcome optimization.
PHYSICS IN MEDICINE AND BIOLOGY
(2021)
Article
Computer Science, Artificial Intelligence
Viktor Wegmayr, Joachim M. Buhmann
Summary: A probabilistic model based on the Fisher-von-Mises distribution was proposed for estimating the maximum entropy posteriors of local streamline directions. The optimal precision of posteriors for streamlines was determined using information-theoretic technique to ensure stable results in retest measurements of the same subject.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2021)
Article
Anatomy & Morphology
Corentin Dauleac, Carole Frindel, Isabelle Pelissou-Guyotat, Celia Nicolas, Fang-Cheng Yeh, Juan Fernandez-Miranda, Francois Cotton, Timothee Jacquesson
Summary: Despite recent advancements in diffusion-weighted imaging, reconstructing tractograms of the spinal cord remains challenging, hindering routine clinical use. The new full tractography approach simplifies the process and offers a reliable 3D rendering of the spinal cord, aiding in adjusting neurosurgical strategies.
FRONTIERS IN NEUROANATOMY
(2022)
Article
Computer Science, Artificial Intelligence
Miguel Antonio Barbero-Alvarez, David Jimenez, Ramiro Garcia-Luna, Salvador Senent, Jose Manuel Menendez, Rafael Jimenez
Summary: This paper presents a horizontal technique for classifying the main directions of orientation of planar surfaces that make up solid bodies. By mathematically representing these surfaces using 3D point clouds and detecting their normal vectors, the technique accurately identifies the planar surfaces and retrieves their orientations. It distinguishes itself from other techniques in the field by applying specific filtering inspired by image processing techniques to mitigate the effects of mathematical noise caused by surface roughness, and by representing the data on the unit-sphere using the von Mises-Fisher mixture model. The proposed technique is evaluated using various validation cases of increasing complexity and its potential applications for a wide range of real-life bodies are analyzed.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)
Article
Computer Science, Artificial Intelligence
Alexis Reymbaut, Alex Valcourt Caron, Guillaume Gilbert, Filip Szczepankiewicz, Markus Nilsson, Simon K. Warfield, Maxime Descoteaux, Benoit Scherrer
Summary: Diffusion tensor imaging provides increased sensitivity to microstructural tissue changes compared to conventional anatomical imaging, but presents limited specificity. To address this issue, the DIAMOND model subdivides voxel content into diffusion compartments and estimates compartmental non-central matrix-variate Gamma distributions of diffusion tensors. Incorporating tensor-valued diffusion encoding, the Magic DIAMOND model demonstrates improved accuracy in estimating brain microstructural features, particularly in regions of fiber crossing.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Neurosciences
Burke Q. Rosen, Eric Halgren
Summary: The WU-Minn Human Connectome Project (HCP) provides a dataset of advanced MRI techniques for over a thousand healthy subjects, with a focus on resting-state fMRI. A full-cortex connectome derived from probabilistic diffusion tractography revealed that connection strengths are lognormally distributed and decay exponentially with tract length, among other findings. Comparisons with existing connectivity matrices suggest that the dMRI connectome is more similar to cortico-cortico-evoked potential connectivity.
Article
Radiology, Nuclear Medicine & Medical Imaging
Divya Joshi, M. Hongchul Sohn, Julius P. A. Dewald, Wendy M. Murray, Carson Ingo
Summary: The purpose of this study was to compare the sensitivity profiles of probabilistic and deterministic DTI tractography methods in estimating geometric properties in arm muscle anatomy. The findings showed that the probabilistic method provided estimates closer to conventional measurements and produced more fascicles compared to the deterministic method. A wide turning angle and lower SNR affected the results of both methods.
MAGNETIC RESONANCE IN MEDICINE
(2023)
Article
Computer Science, Artificial Intelligence
Lu Bai, Lixin Cui, Yuhang Jiao, Luca Rossi, Edwin R. Hancock
Summary: In this paper, a novel backtrackless aligned-spatial graph convolutional network (BASGCN) model is proposed for learning effective features for graph classification. This model addresses the issues of information loss and imprecise information representation in existing spatially-based graph convolutional network (GCN) models, and bridges the theoretical gap between traditional convolutional neural network (CNN) models and spatially-based GCN models. Experimental results demonstrate the effectiveness of the proposed model.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Xing Ai, Chengyu Sun, Zhihong Zhang, Edwin R. Hancock
Summary: The study proposes a novel GNN framework called TL-GNN, which combines subgraph-level information with node-level information to enrich the features captured by GNNs. The study also provides a mathematical analysis of the LPI problem and proposes a subgraph counting method based on the dynamic programming algorithm.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Chen Wang, Xiang Wang, Jiawei Zhang, Liang Zhang, Xiao Bai, Xin Ning, Jun Zhou, Edwin Hancock
Summary: This paper proposes a novel approach to estimate uncertainties in stereo matching end-to-end, using the NIG distribution to calculate uncertainties and additional loss functions to enhance sensitivity and smoothness. Experimental results show that this method improves stereo matching results, particularly performing well on out-of-distribution data.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Artificial Intelligence
Lei Zhou, Yang Liu, Pengcheng Zhang, Xiao Bai, Lin Gu, Jun Zhou, Yazhou Yao, Tatsuya Harada, Jin Zheng, Edwin Hancock
Summary: Zero-shot learning aims to recognize novel classes by transferring semantic knowledge. The proposed bidirectional embedding based generative model introduces an information bottleneck constraint to preserve attribute information. Experimental results show that the method outperforms state-of-the-art methods on benchmark datasets.
Article
Computer Science, Artificial Intelligence
Xingchen Guo, Xuexin Xu, Xunquan Chen, Jinhui Chen, Rong Jia, Zhihong Zhang, Tetsuya Takiguchi, Edwin R. Hancock
Summary: This paper presents an effective method for multi-talker localization using only a single microphone in a room, which can successfully and accurately process the localization task. Experiments demonstrate the effectiveness of the proposed method.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Artificial Intelligence
Zhihong Zhang, Dongdong Chen, Lu Bai, Jianjia Wang, Edwin R. Hancock
Summary: This article introduces the efficient representation of network structure using motifs and studies the distribution of subgraphs using statistical mechanics to understand the motif structure of a network. By mapping network motifs to clusters in a gas model, the partition function for a network is derived to calculate global thermodynamic quantities. Analytical expressions for the number of specific types of motifs and their associated entropy are presented. Numerical experiments on synthetic and real-world data sets evaluate the qualitative and quantitative characteristics of motif entropy derived from the partition function. The motif entropy for real-world networks, such as financial stock market networks, is found to be sensitive to the variance in network structure, indicating well-defined information-processing functions of network motifs.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Lu Bai, Lixin Cui, Zhihong Zhang, Lixiang Xu, Yue Wang, Edwin R. Hancock
Summary: This paper presents a novel framework for computing kernel-based similarity measures between dynamic time-varying financial networks, which is used to analyze financial time series. The commute time (CT) matrix is computed to identify a reliable set of correlated time series and their associated probability distributions. The dominant probability distributions are then used to construct a Shannon entropy time series, which is further used to develop an entropic dynamic time warping kernel for financial time series analysis. Experimental results demonstrate the effectiveness of the proposed method.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Dongdong Chen, Yuxing Dai, Lichi Zhang, Zhihong Zhang, Edwin R. Hancock
Summary: This paper presents a novel neural framework that converts the graph matching problem into a linear assignment problem in a high-dimensional space. By leveraging relative position information at the node level and high-order structural arrangement information at the subgraph level, the method improves the performance of graph matching tasks and establishes reliable node-to-node correspondences.
PATTERN RECOGNITION
(2023)
Article
Automation & Control Systems
Yuqing Ma, Xianglong Liu, Shihao Bai, Lei Wang, Aishan Liu, Dacheng Tao, Edwin R. Hancock
Summary: In this study, a generic inpainting framework is proposed to handle incomplete images with both contiguous and discontiguous large missing areas. By employing an adversarial modeling and regionwise operations, the framework is able to generate semantically reasonable and visually realistic images, outperforming existing methods on large contiguous and discontiguous missing areas, as demonstrated by qualitative and quantitative experiments.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Computer Science, Artificial Intelligence
Lu Bai, Yuhang Jiao, Lixin Cui, Luca Rossi, Yue Wang, Philip S. Yu, Edwin R. Hancock
Summary: This paper proposes a new Quantum Spatial Graph Convolutional Neural Network (QSGCNN) model that can directly learn a classification function for graphs of arbitrary sizes. Unlike state-of-the-art Graph Convolutional Neural Network (GCNN) models, the proposed QSGCNN model incorporates the process of identifying transitive aligned vertices between graphs and transforms arbitrary sized graphs into fixed-sized aligned vertex grid structures. The effectiveness of the proposed QSGCNN model is demonstrated through experiments on benchmark graph classification datasets.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Lixin Cui, Ming Li, Lu Bai, Yue Wang, Jing Li, Yanchao Wang, Zhao Li, Yunwen Chen, Edwin R. Hancock
Summary: This paper proposes a novel framework for computing Quantum-based Entropic Representations (QBER) for un-attributed graphs using Continuous-time Quantum Walk (CTQW). By transforming each original graph into a family of k-level neighborhood graphs, the framework captures multi-level topological information of the original global graph. The structure of each neighborhood graph is characterized using the Average Mixing Matrix (AMM) of CTQW, enabling the computation of Quantum Shannon Entropy and entropic signature. Experimental results demonstrate the effectiveness of the proposed approach in classification accuracies, outperforming other entropic complexity measuring methods, graph kernel methods, and graph deep learning methods.
PATTERN RECOGNITION
(2024)
Proceedings Paper
Computer Science, Artificial Intelligence
Lu Bai, Lixin Cui, Edwin R. Hancock
Summary: In this paper, a new graph kernel called Hierarchical Transitive-Aligned Kernel is proposed, which transitsively aligns vertices between graphs and incorporates the locational correspondence information and all graph information into the kernel computation process. It has shown to be effective in experimental evaluations.
INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 162
(2022)
Article
Computer Science, Artificial Intelligence
Silvia Tozza, Dizhong Zhu, William A. P. Smith, Ravi Ramamoorthi, Edwin R. Hancock
Summary: In this paper, we present a method for estimating shape from polarisation and shading information under unknown illumination conditions. We propose alternative photo-polarimetric constraints and demonstrate how to express them using a unified system of partial differential equations, which allows for linear least squares solutions. We also introduce new methods for estimating polarisation images, albedo, and refractive index, and evaluate their performance on both synthetic and real-world data, showing improvements over existing state-of-the-art methods.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Geochemistry & Geophysics
W. A. P. Smith, P. Lewinska, M. A. Cooper, E. R. Hancock, J. A. Dowdeswell, D. M. Rippin
Summary: This paper studies the problem of structure-from-motion for images with varying principal point. Initialization and pose estimation methods specific to this scenario are proposed and the performance is demonstrated on challenging real-world examples.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geography, Physical
Michael A. Cooper, Paulina Lewinska, William A. P. Smith, Edwin R. Hancock, Julian A. Dowdeswell, David M. Rippin
Summary: This study presents an approach to extract quantifiable information from archival aerial photographs to extend the record of change in central eastern Greenland Ice Sheet. The insights gained from a longer record of ice margin change are crucial for understanding glacier response to climate change. The study also focuses on relatively small and understudied outlet glaciers from the eastern margin of the ice sheet, revealing significant heterogeneity in their response with non-climatic controls playing a key role.
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)