Article
Computer Science, Artificial Intelligence
Ahmed H. Aly, Pulkit Khandelwal, Abdullah H. Aly, Takayuki Kawashima, Kazuki Mori, Yoshiaki Saito, Judy Hung, Joseph H. Gorman, Alison M. Pouch, Robert C. Gorman, Paul A. Yushkevich
Summary: This study develops and validates a fully automated algorithm for segmentation and shape modeling of the left ventricular mitral valve complex from pre-operative 3D transesophageal echocardiography. The algorithm achieves accurate results and provides support for surgical decision making for patients with ischemic mitral regurgitation.
MEDICAL IMAGE ANALYSIS
(2022)
Review
Cardiac & Cardiovascular Systems
Valentina Mantegazza, Paola Gripari, Gloria Tamborini, Manuela Muratori, Laura Fusini, Sarah Ghulam Ali, Anna Garlasche, Mauro Pepi
Summary: Three-dimensional echocardiography (3DE) has become a routine clinical tool for evaluating mitral valve prolapse (MVP), allowing for the diagnosis of MVP, assessment of mitral valve morphology and dynamics, and quantification of mitral regurgitation. This review focuses on the role and advantages of 3DE in the comprehensive evaluation of MVP, as well as intraoperative and intraprocedural monitoring.
FRONTIERS IN CARDIOVASCULAR MEDICINE
(2023)
Article
Medicine, General & Internal
Aleksejus Zorinas, Diana Zakarkaite, Vilius Janusauskas, Donatas Austys, Lina Puodziukaite, Gitana Zuoziene, Robertas Stasys Samalavicius, Ieva Jovaisiene, Giedrius Davidavicius, Kestutis Rucinskas, Eustaquio Maria Onorato
Summary: Multimodality imaging techniques are crucial in the diagnosis and treatment of paravalvular leaks, and the application of EFF can simplify the procedure and provide accurate cardiac anatomy reconstruction. This manuscript presents basic recommendations for the application of EFF in practice, supported by challenging clinical examples.
JOURNAL OF CLINICAL MEDICINE
(2022)
Article
Cardiac & Cardiovascular Systems
Rawan K. Rumman, Subodh Verma, Vincent Chan, David Mazer, Adrian Quan, Makoto Hibino, Benoit De Varennes, Michael W. A. Chu, David Latter, Hwee Teoh, Bobby Yanagawa, Howard Leong-Poi, Kim A. Connelly
Summary: This study evaluated the effect of annuloplasty size on postoperative mitral valve hemodynamics during exercise and examined predictors of mitral valve hemodynamics. Intraoperative mean and peak mitral valve gradients by transesophageal echocardiography independently predicted mean and peak resting and exercise gradients at follow-up.
Article
Multidisciplinary Sciences
Yeshu Li, Ziming Qiu, Xingyu Fan, Xianglong Liu, Eric I-Chao Chang, Yan Xu
Summary: In this paper, a novel multi-atlas-based algorithm for 3D MRI brain structure segmentation is developed, which integrates flow, SIFT features, and classical machine learning algorithms. The experimental results show that the proposed method performs well in various settings and has potential for general applicability.
Article
Engineering, Biomedical
Jinhui Chen, Hanzhao Li, Gaowei He, Fengjuan Yao, Lixuan Lai, Jianping Yao, Longhan Xie
Summary: The aim of this study was to achieve automatic segmentation of patient-specific 3D transesophageal echocardiography (TEE) mitral valve leaflets without user interaction and to evaluate the feasibility of quantitative measurements on the automatic segmentation model. A novel pre-training strategy was proposed to improve the segmentation performance. The results showed that the automatic segmentation model had a good agreement with the clinical software in most mitral annular parameters, indicating its reliability for quantitative measurements.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2023)
Review
Cardiac & Cardiovascular Systems
Kensuke Hirasawa, Masaki Izumo
Summary: Edge-to-edge transcatheter mitral valve repair (TMVr) using MitraClip is rapidly evolving as a treatment option for patients with severe mitral regurgitation (MR) who are at high surgical risk or have contraindications for surgery. Three-dimensional (3D) echocardiography, particularly 3D transesophageal echocardiography (TEE), plays a crucial role in evaluating mitral valve geometry and quantifying the severity of MR with dedicated software. Real-time 3D TEE is commonly utilized for guidance during TMVr procedures. Further advancements in 3D echocardiography may lead to safer and more beneficial outcomes for patients undergoing TMVr.
FRONTIERS IN CARDIOVASCULAR MEDICINE
(2022)
Article
Materials Science, Multidisciplinary
A. H. Aly, E. K. Lai, N. Yushkevich, R. H. Stoffers, J. H. Gorman, A. T. Cheung, H. Gorman, R. C. Gorman, P. A. Yushkevich, A. M. Pouch
Summary: The study aims to accurately reconstruct cardiac valve morphology and motion through image analysis algorithms, providing personalized descriptions for cardiac patients and insights into disease pathophysiology. Results demonstrate that automated 4D image analysis allows for reliable modeling of mitral valve dynamics, facilitating research on pathological and normal valves.
EXPERIMENTAL MECHANICS
(2021)
Article
Computer Science, Information Systems
Wangbin Ding, Lei Li, Xiahai Zhuang, Liqin Huang
Summary: Multi-atlas segmentation (MAS) is a promising framework for medical image segmentation. Traditional MAS methods face limitations in terms of available atlases and computational burden. In this work, a novel cross-modality MAS framework using deep neural networks for image registration and label fusion is proposed, demonstrating improved efficiency and segmentation performance.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2022)
Article
Multidisciplinary Sciences
Nick Arbic, Andreea Dragulescu, Luc Mertens, Olivier Villemain
Summary: Mitral valve disease in pediatric cardiology is complex and the addition of 3D echocardiography has advantages in defining stenosis and regurgitation mechanisms. Optimized data processing and analysis make it easier for medical teams to integrate this technology, thus improving the accuracy of surgical planning.
JOVE-JOURNAL OF VISUALIZED EXPERIMENTS
(2021)
Article
Cardiac & Cardiovascular Systems
James S. Gammie, Rachael W. Quinn, Erik R. Strauss, Libin Wang, Michael N. D'Ambra, Judy Hung, Daniel A. Bernstein, Douglas Tran, MaryJoe K. Rice, Sari D. Holmes, Chetan Pasrija
Summary: Mitral valve translocation effectively corrects functional (secondary) mitral regurgitation by creating a large surface of coaptation. The procedure resulted in no postoperative mortality, stroke, or renal failure, and the majority of patients had mild or less mitral regurgitation at 1 and 6 months follow-up. Further studies are needed to evaluate the long-term durability and clinical utility of this operation.
ANNALS OF THORACIC SURGERY
(2021)
Article
Cardiac & Cardiovascular Systems
Valentina Mantegazza, Valentina Volpato, Paola Gripari, Sarah Ghulam Ali, Laura Fusini, Gianpiero Italiano, Manuela Muratori, Gianluca Pontone, Gloria Tamborini, Mauro Pepi
Summary: Mitral annular disjunction (MAD) is associated with mitral valve prolapse (MVP) and may lead to malignant ventricular arrhythmias. Different imaging techniques such as transthoracic echocardiography (TTE) and cardiac magnetic resonance (CMR) can be used for MAD identification and measurement, with an integrated approach being necessary for comprehensive assessment of patients with MVP and arrhythmia symptoms.
Article
Cardiac & Cardiovascular Systems
Francesco Melillo, Andrea Fisicaro, Stefano Stella, Francesco Ancona, Cristina Capogrosso, Giacomo Ingallina, Davide Maccagni, Vittorio Romano, Stefania Ruggeri, Cosmo Godino, Azeem Latib, Matteo Montorfano, Antonio Colombo, Eustachio Agricola
Summary: This study demonstrated the use of fluoroscopic-echocardiographic fusion imaging for intraprocedural guidance during transcatheter edge-to-edge mitral valve repair, showing a reduction in fluoroscopy time and improvement in procedural success in a population with challenging mitral anatomy for percutaneous repair.
JOURNAL OF THE AMERICAN SOCIETY OF ECHOCARDIOGRAPHY
(2021)
Article
Chemistry, Multidisciplinary
Patrick Carnahan, John Moore, Daniel Bainbridge, Elvis C. S. Chen, Terry M. Peters
Summary: Three-dimensional ultrasound mosaicing technique can improve image quality and expand the field of view for cardiac procedures, particularly for mitral valve operations. This study proposes a new compounding technique that utilizes multiple ultrasound volumes from different positions to reconstruct high-detail images of the mitral valve and sub-valvular structures. The technique has been validated using excised porcine mitral valve units and patient data, demonstrating its accuracy in capturing the physical structures present.
APPLIED SCIENCES-BASEL
(2022)
Article
Computer Science, Interdisciplinary Applications
Iman Aganj, Bruce Fischl
Summary: A new approach for medical image segmentation is proposed, which calculates the probability of all possible atlas-to-image transformations and the expected label value (ELV), avoiding the issue of local optima. This method does not require actually performing deformable registration, hence saving computational costs.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2021)
Article
Computer Science, Interdisciplinary Applications
Paul A. Yushkevich, Artem Pashchinskiy, Ipek Oguz, Suyash Mohan, J. Eric Schmitt, Joel M. Stein, Dzenan Zukic, Jared Vicory, Matthew McCormick, Natalie Yushkevich, Nadav Schwartz, Yang Gao, Guido Gerig
Article
Neurosciences
Long Xie, Sandhitsu R. Das, Arun Pilania, Molly Daffner, Grace E. Stockbower, Sudipto Dolui, Paul A. Yushkevich, John A. Detre, David A. Wolk
Article
Materials Science, Multidisciplinary
A. H. Aly, E. K. Lai, N. Yushkevich, R. H. Stoffers, J. H. Gorman, A. T. Cheung, H. Gorman, R. C. Gorman, P. A. Yushkevich, A. M. Pouch
Summary: The study aims to accurately reconstruct cardiac valve morphology and motion through image analysis algorithms, providing personalized descriptions for cardiac patients and insights into disease pathophysiology. Results demonstrate that automated 4D image analysis allows for reliable modeling of mitral valve dynamics, facilitating research on pathological and normal valves.
EXPERIMENTAL MECHANICS
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Carrie E. Zimmerman, Pulkit Khandelwal, Long Xie, Hyunyeol Lee, Hee Kwon Song, Paul A. Yushkevich, Arastoo Vossough, Scott P. Bartlett, Felix W. Wehrli
Summary: This study evaluated an automatic multi-atlas segmentation pipeline for cranial vault images, eliminating the need for manual intervention. The results showed good agreement between CT and automated MRI-based 3D cranial vault renderings, effectively eliminating the labor-intensive manual segmentation process.
ACADEMIC RADIOLOGY
(2022)
Article
Clinical Neurology
Paul A. Yushkevich, Monica Munoz Lopez, Maria Mercedes Iniguez de Onzono Martin, Ranjit Ittyerah, Sydney Lim, Sadhana Ravikumar, Madigan L. Bedard, Stephen Pickup, Weixia Liu, Jiancong Wang, Ling Yu Hung, Jade Lasserve, Nicolas Vergnet, Long Xie, Mengjin Dong, Salena Cui, Lauren McCollum, John L. Robinson, Theresa Schuck, Robin de Flores, Murray Grossman, M. Dylan Tisdall, Karthik Prabhakaran, Gabor Mizsei, Sandhitsu R. Das, Emilio Artacho-Perula, Mari'a Del Mar Arroyo Jimenez, Mari'a Pilar Marcos Raba, Francisco Javier Molina Romero, Sandra Cebada Sanchez, Jose Carlos Delgado Gonzalez, Carlos De la Rosa-Prieto, Marta Corcoles Parada, Edward B. Lee, John Q. Trojanowski, Daniel T. Ohm, Laura E. M. Wisse, David A. Wolk, David J. Irwin, Ricardo Insausti
Summary: This study utilized ex vivo MRI and dense serial histological imaging to construct three-dimensional quantitative maps of neurofibrillary tangle burden in the medial temporal lobe, revealing significant variation along different anatomical regions. The findings provide valuable insights into the distribution of this neurodegenerative pathology and may support the development and validation of neuroimaging biomarkers.
Article
Geriatrics & Gerontology
Laura Em Wisse, Long Xie, Sandhitsu R. Das, Robin de Flores, Oskar Hansson, Mohamad Habes, Jimit Doshi, Christos Davatzikos, Paul A. Yushkevich, David A. Wolk
Summary: The study found that CSF p-tau levels partially mediated the effect of age on hippocampal atrophy rates, while no significant associations were observed for WMHs with temporal lobe structural changes. These results suggest a potential role of tau pathology in age-related MTL structural changes.
NEUROBIOLOGY OF AGING
(2022)
Article
Biology
Danni Tu, Manu S. Goyal, Jordan D. Dworkin, Samuel Kampondeni, Lorenna Vidal, Eric Biondo-Savin, Sandeep Juvvadi, Prashant Raghavan, Jennifer Nicholas, Karen Chetcuti, Kelly Clark, Timothy Robert-Fitzgerald, Theodore D. Satterthwaite, Paul Yushkevich, Christos Davatzikos, Guray Erus, Nicholas J. Tustison, Douglas G. Postels, Terrie E. Taylor, Dylan S. Small, Russell T. Shinohara
Summary: A central challenge in medical imaging studies is to extract biomarkers that can characterize disease pathology or outcomes. This paper presents a fully automated framework for translating radiological diagnostic criteria into image-based biomarkers, with excellent classification performance.
Article
Clinical Neurology
Claire Andre, Elizabeth Kuhn, Stephane Rehel, Valentin Ourry, Solene Demeilliez-Servouin, Cassandre Palix, Francesca Felisatti, Pierre Champetier, Sophie Dautricourt, Paul Yushkevich, Denis Vivien, Vincent de la Sayette, Gael Chetelat, Robin de Flores, Geraldine Rauchs, Medit Ageing Res Grp
Summary: This study aimed to investigate the association between sleep disordered breathing (SDB) and medial temporal lobe neurodegeneration, as well as subsequent episodic memory impairment. The study found that SDB was associated with reduced volumes of medial temporal lobe subregions in amyloid-positive individuals, but not in amyloid-negative individuals. Additionally, lower baseline volumes of the whole hippocampus and CA1 were associated with worse episodic memory performance at follow-up.
Article
Neuroimaging
Alessandra M. Valcarcel, John Muschelli, Dzung L. Pham, Melissa Lynne Martin, Paul Yushkevich, Rachel Brandstadter, Kristina R. Patterson, Matthew K. Schindler, Peter A. Calabresi, Rohit Bakshi, Russell T. Shinohara
NEUROIMAGE-CLINICAL
(2020)
Meeting Abstract
Clinical Neurology
Lauren McCollum, Laura Wisse, Salena Cui, Robin de Flores, Sandhitsu Das, Long Xie, Paul Yushkevich, David Wolk
Meeting Abstract
Clinical Neurology
Lauren McCollum, Laura Wisse, Sandhitsu Das, Robin de Flores, Paul Yushkevich, David Wolk
Meeting Abstract
Clinical Neurology
Simon Miller, Jeffrey Phillips, Ranjit Ittyerah, Claire Peterson, Edward Lee, John Trojanowski, David Wolk, Paul Yushkevich, Murray Grossman, David Irwin
Meeting Abstract
Clinical Neurology
Jeffrey Phillips, Laura Wisse, Paul Yushkevich, James Gee, Murray Grossman, David Irwin
Meeting Abstract
Clinical Neurology
Sandhitsu Das, Long Xie, Laura Wisse, Ranjit Ittyerah, Paul Yushkevich, David Wolk
Meeting Abstract
Clinical Neurology
Ranjit Ittyerah, David G. Coughlin, Jeffrey Phillips, Simon Miller, Edward B. Lee, John Q. Trojanowski, Daniel Weintraub, Andrew Siderowf, John E. Duda, Howard I. Hurtig, David A. Wolk, Paul Yushkevich, Murray Grossman, David J. Irwin
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)