Article
Engineering, Electrical & Electronic
Rosa-Maria Menchon-Lara, Federico Simmross-Wattenberg, Manuel Rodriguez-Cayetano, Pablo Casaseca-de-la-Higuera, Miguel A. Martin-Fernandez, Carlos Alberola-Lopez
Summary: Recently, an efficient implementation of convolution-based free form deformations (FFD) has been proposed for both groupwise 3D monomodal and 2D pairwise multimodal registrations. However, there is still a demand for groupwise L-D multimodal registration with L >= 2. In this correspondence, the authors address this need and present a solution for achieving accurate registration using two popular metrics: Renyi entropy and PCA2.
Article
Computer Science, Information Systems
Manal El Rhazi, Arsalane Zarghili, Aicha Majda, Ayat Allah Oufkir
Summary: Physical attractiveness has a significant impact on human social life, and people enhance their attractiveness through makeup and plastic surgery. This study introduces a novel system based on Bezier function for analyzing and enhancing the attractiveness of 3D faces.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Fang Bai, Adrien Bartoli
Summary: This paper introduces the problem of deformable Generalized Procrustes Analysis (GPA) and resolves fundamental ambiguities using shape constraints requiring eigenvalues of shape covariance. A closed-form and optimal solution based on eigenvalue decomposition is provided, handling regularization and favoring smooth deformation fields. This method is applicable to most common transformation models, offering a fast, globally optimal and widely applicable solution.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2022)
Article
Automation & Control Systems
Weixing Peng, Yaonan Wang, Hui Zhang, Yurong Chen, Haotian Wu, Jiawen Zhao
Summary: This article proposes a robust joint registration approach for multiview point clouds, which provides a robust initialization and resists noise by minimizing the distance between probability distributions of integrated and standard models, and utilizing Lie algebra solutions and maximum likelihood estimation.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2022)
Article
Computer Science, Artificial Intelligence
Xingxing Zhu, Mingyue Ding, Xuming Zhang
Summary: This paper presents an unsupervised image registration method using free form deformation (FFD) and symmetry constraint-based generative adversarial networks (FSGAN). The FSGAN utilizes PCA network-based structural representations for image inputs, and the generator learns FFD model parameters to produce deformation fields. Two discriminators are used to decide whether bilateral registration has been realized simultaneously. The symmetry constraint is utilized to construct the loss function and avoid deformation folding. Experimental results show that the FSGAN outperforms state-of-the-art methods in terms of visual comparisons, dice value, target registration error, and computational efficiency.
CAAI TRANSACTIONS ON INTELLIGENCE TECHNOLOGY
(2023)
Article
Environmental Sciences
Franck Thollard, Dominique Clesse, Marie-Pierre Doin, Joelle Donadieu, Philippe Durand, Raphael Grandin, Cecile Lasserre, Christophe Laurent, Emilie Deschamps-Ostanciaux, Erwan Pathier, Elisabeth Pointal, Catherine Proy, Bernard Specht
Summary: The FLATSIM service aims to process Sentinel-1 data over large areas using multi-temporal InSAR techniques, providing the ForM@ter scientific community with automatically processed products and quality indicators for research in seismology, tectonics, volcano-tectonics, and hydrological cycle.
Article
Computer Science, Artificial Intelligence
Jingyu Sun, Yadong Gong, Jibin Zhao, Huan Zhang, Liya Jin
Summary: The basis for guidance in the field of automated robot processing is modeling by visual scanning. Matching algorithms play a pivotal role by providing the exact location of design models and measurement data, and a fine registration method is proposed to meet the requirement of uniform machining allowance. The proposed method performs well in terms of solution efficiency and accuracy of results, and exhibits good robustness against Gaussian noise.
PATTERN RECOGNITION
(2023)
Article
Automation & Control Systems
Zijie Wu, Yaonan Wang, He Xie, Mingtao Feng, Haotian Wu, Xuebing Liu, Jingtao Sun
Summary: In this article, a novel gravitational discriminative optimization (GDO) method based on a multiview reconstruction framework is proposed for flexible and efficient industrial manufacturing. The method includes a training phase and a reconstruction phase, and demonstrates accurate and robust reconstruction in experiments, outperforming other methods in terms of both accuracy and robustness.
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2023)
Article
Multidisciplinary Sciences
Weichen Li, Fengwen Wang, Ole Sigmund, Xiaojia Shelly Zhang
Summary: In this study, a freeform inverse design approach is used to synthesize multiple hyperelastic materials into composite structures, enabling them to achieve arbitrary prescribed responses under large deformations. The digitally synthesized structures exhibit organic shapes and motions with irregular distributions of material phases. By utilizing multi material fabrication and heteroassembly strategies, function-oriented mechanical devices with highly complex yet navigable responses can be designed.
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
(2022)
Article
Computer Science, Artificial Intelligence
Yuguo Liu, Wenyu Chen, Hanwen Liu, Yun Zhang, Malu Zhang, Hong Qu
Summary: In this study, we explore three spiking neuron models to post-process the original dense word embeddings and test the sparse temporal codes generated on several tasks involving word-level and sentence-level semantics. The experimental results demonstrate that our sparse binary word representations can capture semantic information as well as or even better than the original word embeddings, while requiring less storage. These methods provide a robust representation foundation of language in terms of neuronal activities, which could potentially be applied to more complex natural language tasks under neuromorphic computing systems.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Engineering, Multidisciplinary
Bowen Yi, Ruibin Liang, Xiaosun Wang, Shijing Wu, Nuodi Huang
Summary: This study proposes a surface reconstruction method based on surface position error and deformation error model. The method decomposes surface form error into surface position error and deformation error, and then adjusts surface location and orientation using an iterative algorithm based on point-to-surface function to achieve the best match with sampling points. The advantage of this method is that it considers the probe dimension, eliminating the need for further compensation of probe radius, while preserving the geometric features of the design surface.
PRECISION ENGINEERING-JOURNAL OF THE INTERNATIONAL SOCIETIES FOR PRECISION ENGINEERING AND NANOTECHNOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Pengfei Zheng, Qing Liu, Jingjing Lou, Chengjie Lian, Dajun Lin
Summary: This paper proposes a free-form surface flattening algorithm that combines the advantages of geometric flattening and mechanical energy flattening methods to solve the problems of many iterations, large changes in convergence, and weak visualization of deformation in existing algorithms. The algorithm meshes the point cloud surface using the triangular slice search method and wraps the 3D surface mesh around the surface using geometric mapping relationships. Iterative optimization is employed to optimize the initial flattening graph. The algorithm is shown to be general, robust, accurate, and capable of visualizing the flattening deformation.
IET IMAGE PROCESSING
(2022)
Article
Automation & Control Systems
Hossein Rastgoftar, Ilya Kolmanovsky
Summary: This paper proposes a spatio-temporal reference trajectory planner approach to safely plan continuum deformation coordination of a multi-agent system in an obstacle-laden environment. The proposed method formulates the desired n-D continuum deformation as a leader-follower problem and provides sufficient conditions for inter-agent and obstacle collision avoidance. By spatially optimizing the continuum deformation coordination and planning piece-wise constant reference inputs, the method minimizes travel time and ensures the safety of the coordination.
Article
Chemistry, Physical
Yang Cao, Jingyan Dong
Summary: This article presents a new method for fabricating soft electrothermal actuators using EHD printing for direct patterning of the resistive heater. By changing the design of the heating pattern on the actuator, complex programmable deformations can be achieved, including uniform bending, customized bending, folding, and twisting. Finite element analysis was used to validate the thermal distribution and deformation for different actuator designs, and several integrated demonstrations were presented, showcasing complex structures achieved from folding, a two-degree-of-freedom soft robotic arm, and soft walkers.
Article
Robotics
Matthew Young, Chris Pretty, Josh McCulloch, Richard Green
Summary: The research shows that the Mesh-GICP algorithm is more accurate, precise, and faster than the unmodified GICP. In experiments, the keyscan method is the most accurate registration method, but pairwise matching is more accurate when there is insufficient overlap.
JOURNAL OF FIELD ROBOTICS
(2021)
Article
Cardiac & Cardiovascular Systems
M. Yousuf Salmasi, Selene Pirola, Suchaya Mahuttanatan, Serena M. Fisichella, Sampad Sengupta, Omar A. Jarral, Aung Oo, Declan O'Regan, Xiao Yun Xu, Thanos Athanasiou
Summary: This study investigates the relationship between the angle of the left ventricular outflow tract and ascending thoracic aortic aneurysm disease. The findings suggest that a greater angle is associated with larger diameters in the sinus and ascending aorta. Computational fluid dynamics simulations also show that a larger angle is associated with higher wall shear stress values in the outer curve of the aorta. These results support the hypothesis of flow-mediated disease progression and propose the left ventricular outflow tract angle as a predictor of disease severity.
JOURNAL OF THORACIC AND CARDIOVASCULAR SURGERY
(2023)
Article
Computer Science, Artificial Intelligence
Lei Li, Veronika A. Zimmer, Julia A. Schnabel, Xiahai Zhuang
Summary: This study introduces a new framework named AtrialJSQnet that can conduct left atrial segmentation, scar projection onto the left atrial surface, and scar quantification simultaneously, achieving competitive performance on public datasets. Additionally, by incorporating a shape attention mechanism and spatial encoding loss, the method demonstrates significant effects in reducing noisy patches and improving task performance.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Engineering, Electrical & Electronic
Kerstin Hammernik, Thomas Kustner, Burhaneddin Yaman, Zhengnan Huang, Daniel Rueckert, Florian Knoll, Mehmet Akcakaya
Summary: Physics-driven deep learning methods have revolutionized computational MRI reconstruction by improving the performance of reconstruction. This article provides an overview of recent developments in incorporating physics information into learning-based MRI reconstruction. It discusses both linear and non-linear forward models for computational MRI, classical approaches for solving these inverse problems, as well as physics-driven deep learning approaches such as physics-driven loss functions, plug-and-play methods, generative models, and unrolled networks. Challenges specific to MRI with linear and non-linear forward models are highlighted, and common issues and open challenges are also discussed.
IEEE SIGNAL PROCESSING MAGAZINE
(2023)
Article
Computer Science, Interdisciplinary Applications
Qiang Ma, Liu Li, Emma C. Robinson, Bernhard Kainz, Daniel Rueckert, Amir Alansary
Summary: CortexODE is a deep learning framework that uses neural ordinary differential equations (ODEs) to reconstruct cortical surfaces. By modeling the trajectories of points on the surface as ODEs and parameterizing the derivatives with a learnable deformation network, CortexODE is able to prevent self-intersections. Integrated with an automatic learning-based pipeline, CortexODE can efficiently reconstruct cortical surfaces in less than 5 seconds.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Computer Science, Artificial Intelligence
Chen Qin, Shuo Wang, Chen Chen, Wenjia Bai, Daniel Rueckert
Summary: This paper proposes a novel method for myocardial motion tracking by using a generative model based on variational autoencoder to learn biomechanically plausible deformations and embed them into a neural network-parameterized transformation model. Experimental results show that the proposed method outperforms other approaches in terms of motion tracking accuracy, volume preservation, and generalizability.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Computer Science, Interdisciplinary Applications
Cheng Ouyang, Chen Chen, Surui Li, Zeju Li, Chen Qin, Wenjia Bai, Daniel Rueckert
Summary: In this work, the authors investigate the problem of training a deep network that is robust to unseen domains using only data from one source domain. They propose a causality-inspired data augmentation approach to expose the model to synthesized domain-shifted training examples. The approach is validated on three cross-domain segmentation scenarios and shows consistent performance improvements compared to competitive methods.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Computer Science, Artificial Intelligence
Tamara T. Mueller, Johannes C. Paetzold, Chinmay Prabhakar, Dmitrii Usynin, Daniel Rueckert, Georgios Kaissis
Summary: Graph Neural Networks (GNNs) have become the state-of-the-art for many machine learning applications, but differentially private training of GNNs has remained under-explored. In this work, we propose a framework for differentially private graph-level classification using DP-SGD, which is applicable to multi-graph datasets.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Robert Wright, Alberto Gomez, Veronika A. Zimmer, Nicolas Toussaint, Bishesh Khanal, Jacqueline Matthew, Emily Skelton, Bernhard Kainz, Daniel Rueckert, Joseph V. Hajnal, Julia A. Schnabel
Summary: This paper introduces a novel method to fuse partially imaged fetal head anatomy from multiple views into a single coherent 3D volume. The method aligns and fuses ultrasound images to improve image detail and minimize artifacts, achieving state-of-the-art performance in terms of image quality and robustness.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Cardiac & Cardiovascular Systems
Valentina Quintero Santofimio, Adam Clement, Declan P. O'Regan, James S. Ware, Kathryn A. McGurk
Summary: Study finds that spinal curvature in patients with scoliosis affects cardiac function and increases the risk of cardiovascular events, such as heart failure and atrial fibrillation. These findings have important clinical implications for the management of scoliosis patients.
Article
Multidisciplinary Sciences
Mit Shah, Marco de A. H. Inacio, Chang Lu, Pierre-Raphael Schiratti, Sean L. L. Zheng, Adam Clement, Antonio de Marvao, Wenjia Bai, Andrew P. King, James S. Ware, Martin R. Wilkins, Johanna Mielke, Eren Elci, Ivan Kryukov, Kathryn A. McGurk, Christian Bender, Daniel F. Freitag, Declan P. O'Regan
Summary: Cardiovascular ageing is a progressive process that leads to changes in structure and decline in function. Machine learning can predict biological age and identify key genetic risk factors.
NATURE COMMUNICATIONS
(2023)
Article
Computer Science, Interdisciplinary Applications
Adam Marcus, Paul Bentley, Daniel Rueckert
Summary: The proposed study introduces a novel end-to-end multi-task transformer-based model for concurrent segmentation and age estimation of cerebral ischemic lesions. The method captures long-range dependencies using gated positional self-attention and CT-specific data augmentation, and can be effectively trained with low-data regimes in medical imaging. Experimental results demonstrate promising performance in lesion age classification, outperforming existing task-specific algorithms.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Computer Science, Artificial Intelligence
Jiazhen Pan, Manal Hamdi, Wenqi Huang, Kerstin Hammernik, Thomas Kuestner, Daniel Rueckert
Summary: This article introduces a learning-based and unrolled MCMR framework that can achieve accurate and rapid CMR reconstruction, delivering artifacts-free motion estimation and high-quality reconstruction even at imaging acceleration rates up to 20x.
MEDICAL IMAGE ANALYSIS
(2024)
Proceedings Paper
Cardiac & Cardiovascular Systems
Michael Tanzer, Sea Hee Yook, Pedro Ferreira, Guang Yang, Daniel Rueckert, Sonia Nielles-Vallespin
Summary: This study compares the effects of different input types, dimensionalities, and input types on the performance of a deep learning-based model in accelerating cardiac DTI. The results show that simpler 2D real-valued models outperform 3D or complex models, and the best performance is achieved by a real-valued model trained using both magnitude and phase components.
STATISTICAL ATLASES AND COMPUTATIONAL MODELS OF THE HEART: REGULAR AND CMRXMOTION CHALLENGE PAPERS, STACOM 2022
(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)
