Related references
Note: Only part of the references are listed.
Article
Computer Science, Artificial Intelligence
Eduardo Castro et al.
Summary: Breast cancer is the most common and lethal cancer in women. Efforts have been made to develop accurate neural network-based computer-aided diagnosis systems for screening and prediction of this disease. This study presents a symmetry-based regularization method to improve the model's generalization ability to unseen examples in different environments.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Engineering, Electrical & Electronic
Jun-Ding Sun et al.
Summary: In this study, a new NAS architecture named GNAS-U(2)Net is proposed for medical image segmentation, achieving significant performance improvement. Compared to common U-Net and its variants, GNAS-U(2)Net shows better performance with fewer parameters.
IEEE SIGNAL PROCESSING LETTERS
(2022)
Article
Medicine, General & Internal
Yeonwoo Jeong et al.
Summary: Automating screening and diagnosis in the medical field saves time, reduces chances of misdiagnosis, and saves on labor and costs. With the development of deep learning methods, machines are now capable of interpreting complex features in medical data, leading to rapid advancements in automation. This paper reviews recent state-of-the-art works in ophthalmology that utilize color fundus images for automated screening and diagnosis of diabetic retinopathy, age-related macular degeneration, and glaucoma. The challenges in developing these systems are also discussed.
Article
Computer Science, Interdisciplinary Applications
Junyan Lyu et al.
Summary: The article introduces a data manipulation based domain generalization method called Automated Augmentation for Domain Generalization (AADG). By sampling data augmentation policies in an appropriate search space, this method generates new domains and enriches the training set. Experimental results show that AADG has state-of-the-art generalization performance in medical image segmentation tasks.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2022)
Article
Computer Science, Artificial Intelligence
Arnaud Boutillon et al.
Summary: Clinical diagnosis of the pediatric musculoskeletal system relies on the analysis of medical imaging examinations. In this study, a novel multi-task, multi-domain learning framework is proposed for semantic segmentation of pediatric imaging datasets. The proposed approach overcomes the scarcity of pediatric data by considering multiple intensity domains and segmentation tasks, and leveraging shared features between datasets. Transfer learning, multi-scale contrastive regularization, and anatomical priors are employed to improve the model's generalization capabilities. Experimental results demonstrate the superiority of the proposed approach in bone segmentation compared to individual, transfer, and shared segmentation schemes.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Biology
Carla Sendra-Balcells et al.
Summary: This paper investigates the domain generalization problem in contrast-enhanced imaging and evaluates various deep learning techniques for achieving generalizability to unseen clinical centers. The results demonstrate that the combination of data augmentation and transfer learning can lead to single-center models that generalize well to new clinical centers, surpassing the performance of multi-center models.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Computer Science, Artificial Intelligence
Sulaiman Vesal et al.
Summary: This study introduces a novel 2.5D deep neural network for prostate segmentation on ultrasound images, addressing the challenges of reduced signal-to-noise ratio and artifacts. By combining supervised domain adaptation technique and knowledge distillation loss, the performance drop after model finetuning on new datasets is reduced. The approach achieves high segmentation accuracy and generalizes well in multi-center studies, demonstrating its potential for clinical application.
MEDICAL IMAGE ANALYSIS
(2022)
Review
Computer Science, Artificial Intelligence
Tao Li et al.
Summary: The use of deep learning in fundus image analysis is increasingly popular due to its powerful performance in various applications. This review paper introduces 143 application papers and 33 publicly available datasets, providing summaries, analyses, and solutions for common limitations in each task. The authors also promise to regularly update the state-of-the-art results and newly-released datasets on their GitHub page to keep up with the rapid development in the field.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Biochemical Research Methods
Zekuan Yu et al.
Summary: A new data augmentation method called Multiple Lesions Insertion (MLI) is proposed to generate diabetic retinopathy (DR) fundus images by inserting real lesions into healthy fundus images. The synthetic images can be generated according to clinical rules and exhibit realistic texture features and rich details. The MLI method outperforms traditional augmentation methods in DR screening tasks.
BIOMEDICAL OPTICS EXPRESS
(2021)
Article
Computer Science, Artificial Intelligence
Jose Ignacio Orlando et al.
MEDICAL IMAGE ANALYSIS
(2020)
Article
Computer Science, Interdisciplinary Applications
Ling Zhang et al.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2020)
Article
Computer Science, Interdisciplinary Applications
Shujun Wang et al.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2020)
Review
Engineering, Multidisciplinary
Marco Toldo et al.
Article
Engineering, Biomedical
Zekuan Yu et al.
BIOMEDICAL ENGINEERING ONLINE
(2019)
Review
Ophthalmology
Ahmed Almazroa et al.
JOURNAL OF OPHTHALMOLOGY
(2015)
Article
Multidisciplinary Sciences
Monika E. Danielewska et al.
