Article
Multidisciplinary Sciences
Debadyuti Mukherkjee, Pritam Saha, Dmitry Kaplun, Aleksandr Sinitca, Ram Sarkar
Summary: This paper discusses the application of deep learning in synthetic medical image generation, proposing an aggregation model AGGrGAN that combines different GAN models to generate synthetic MRI scans of brain tumors, and applies style transfer technique to enhance image resemblance.
SCIENTIFIC REPORTS
(2022)
Article
Computer Science, Interdisciplinary Applications
Lazaros Tsochatzidis, Panagiota Koutla, Lena Costaridou, Ioannis Pratikakis
Summary: This study proposed a method to integrate segmentation information of mammographic lesions into convolutional neural networks for improved breast cancer diagnosis. Experimental results demonstrated that the proposed method achieved better performance in diagnosis.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2021)
Article
Computer Science, Interdisciplinary Applications
Ilkay Oksuz
Summary: This study proposed a method using dense convolutional neural networks and residual U-net architecture to detect and correct brain MRI artefacts, improving both image quality and segmentation accuracy. The approach demonstrated high accuracy in stroke segmentation of brain MRI datasets.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Mariana Bustamante, Federica Viola, Jan Engvall, Carl-Johan Carlhall, Tino Ebbers
Summary: This study aimed to develop and evaluate a deep learning-based segmentation method for automatically segmenting the cardiac chambers and great thoracic vessels from 4D flow MRI. The results demonstrated that the deep learning-based method achieved good segmentation accuracy.
JOURNAL OF MAGNETIC RESONANCE IMAGING
(2023)
Article
Neurosciences
Zilong Zeng, Tengda Zhao, Lianglong Sun, Yihe Zhang, Mingrui Xia, Xuhong Liao, Jiaying Zhang, Dinggang Shen, Li Wang, Yong He
Summary: In this study, a new 3D mixed-scale asymmetric convolutional segmentation network (3D-MASNet) was proposed for tissue segmentation of 6-month-old infant brain MRI images. Compared to traditional single-scale symmetric convolutions, this approach demonstrated better accuracy and achieved the best performance in the evaluation.
HUMAN BRAIN MAPPING
(2023)
Article
Computer Science, Information Systems
Susu Kang, Muyuan Yang, X. Sharon Qi, Jun Jiang, Shan Tan
Summary: Accurate segmentation of abdominal organs on MRI is crucial for computer-aided surgery and diagnosis. Most current methods employ an encoder-decoder structure, but simply concatenating shallow and deep features is not sufficient due to the feature gap between them. To address this, we proposed a method that quantifies and bridges the feature gap using spatial and semantic losses. Experimental results on two abdominal MRI datasets show that our method improves segmentation performance without significantly increasing parameters, particularly for organs with blurred boundaries or in a small scale, outperforming state-of-the-art methods.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Wajdi Elhamzi, Wadhah Ayadi, Mohamed Atri
Summary: This article introduces a new brain tumor segmentation method based on deep learning, which uses Convolutional Neural Networks to automatically and accurately segment MRI images. The method shows good performance in the tests, with high segmentation accuracy for different tumor regions.
NEURAL COMPUTING & APPLICATIONS
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Lipeng Xie, Jayaram K. Udupa, Yubing Tong, Drew A. Torigian, Zihan Huang, Rachel M. Kogan, David Wootton, Kok R. Choy, Sanghun Sin, Mark E. Wagshul, Raanan Arens
Summary: The study develops a comprehensive deep learning-based segmentation system for upper airway segmentation on MR images. The proposed approach shows high accuracy and efficiency in segmenting static and dynamic MR images, indicating its potential for applications in dynamic MRI-related fields such as lung or heart segmentation.
Article
Chemistry, Analytical
Nuno Costa, Luis Ferreira, Augusto R. V. F. de Araujo, Bruno Oliveira, Helena R. R. Torres, Pedro Morais, Victor Alves, Joao L. Vilaca
Summary: Breast cancer is a common and deadly disease, and early screening and treatment are crucial. This study proposes a visualization system for breast biopsy using AR glasses and computer applications, which greatly improves lesion visualization and needle alignment. The system was evaluated and showed promising results, with precise needle alignment and accurate lesion targeting.
Article
Chemistry, Analytical
Justin Lo, Jillian Cardinell, Alejo Costanzo, Dafna Sussman
Summary: This paper proposes a DA policy search algorithm that offers an extended set of transformations to accommodate variations in medical imaging datasets. By using the BIPOP-CMA-ES optimizer, the algorithm can return an optimal DA policy based on any input imaging dataset and DL algorithm. The proposed algorithm can be implemented by other researchers in related medical DL applications to improve model performance.
Review
Radiology, Nuclear Medicine & Medical Imaging
Akshay S. Chaudhari, Christopher M. Sandino, Elizabeth K. Cole, David B. Larson, Garry E. Gold, Shreyas S. Vasanawala, Matthew P. Lungren, Brian A. Hargreaves, Curtis P. Langlotz
Summary: Deep learning algorithms have had a significant impact on MRI data acquisition, reconstruction, and interpretation. Three major use cases of deep learning in MRI are model-free image synthesis, model-based image reconstruction, and image or pixel-level classification. Important considerations include model training, evaluation of model robustness, clinical utility, future opportunities, and reproducibility of research.
JOURNAL OF MAGNETIC RESONANCE IMAGING
(2021)
Article
Engineering, Biomedical
Abderazzak Ammar, Omar Bouattane, Mohamed Youssfi
Summary: This study proposes an automated pipeline for cardiac MRI segmentation and diagnosis using fully convolutional neural networks, achieving nearly state-of-the-art accuracy for both segmentation and disease classification challenges.
COMPUTERIZED MEDICAL IMAGING AND GRAPHICS
(2021)
Article
Engineering, Biomedical
Abderazzak Ammar, Omar Bouattane, Mohamed Youssfi
Summary: This study utilizes cardiac cine magnetic resonance imaging for cardiac structure segmentation and disease prediction, achieving impressive results in medical imaging competitions. The automated pipeline proposed in the research uses deep learning for cardiac structure segmentation and disease diagnosis.
COMPUTERIZED MEDICAL IMAGING AND GRAPHICS
(2021)
Article
Engineering, Biomedical
Abderazzak Ammar, Omar Bouattane, Mohamed Youssfi
Summary: Cardiac cine MRI, using fully convolutional neural networks, can accurately segment heart structures and predict diseases. An automated pipeline for heart segmentation and diagnosis was proposed, achieving nearly state-of-the-art accuracy in the segmentation contest and disease classification challenge of the ACDC challenge.
COMPUTERIZED MEDICAL IMAGING AND GRAPHICS
(2021)
Article
Engineering, Biomedical
Abderazzak Ammar, Omar Bouattane, Mohamed Youssfi
Summary: This study proposes an automated pipeline for cardiac segmentation and diagnosis using MRI images and deep learning techniques. By combining three classifiers, the system achieved a high accuracy for heart disease classification on unseen data.
COMPUTERIZED MEDICAL IMAGING AND GRAPHICS
(2021)
Article
Computer Science, Artificial Intelligence
Chetan L. Srinidhi, Ozan Ciga, Anne L. Martel
Summary: This paper presents a comprehensive review of state-of-the-art deep learning approaches used in histopathological image analysis. Through a survey of over 130 papers, the progress in the field based on different machine learning strategies is reviewed. Additionally, the paper discusses the application of deep learning in survival models and highlights the challenges and limitations of current deep learning methods, as well as potential directions for future research.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Engineering, Biomedical
Linde S. Hesse, Grey Kuling, Mitko Veta, Anne L. Martel
Summary: The study demonstrates that applying intensity augmentation can significantly improve segmentation performance of breast magnetic resonance images, especially when there are intensity or orientation differences between the training and test sets. Experimental results show that the proposed method can effectively enhance whole breast segmentation, which is of great significance for clinical MR scans acquired with different protocols.
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING
(2021)
Article
Computer Science, Artificial Intelligence
Ozan Ciga, Anne L. Martel
Summary: This study proposes an architecture that reduces the time spent on data curation and eases the requirements for segmentation level ground truth by utilizing image-level labels. It also helps unlock the potential of previously acquired image-level datasets for segmentation tasks.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Multidisciplinary Sciences
Ozan Ciga, Tony Xu, Sharon Nofech-Mozes, Shawna Noy, Fang-I Lu, Anne L. Martel
Summary: Training deep learning models on Whole Slide Images (WSIs) presents challenges in capturing image details and context, as well as translating dataset annotations and method performance into clinically relevant workflows effectively.
SCIENTIFIC REPORTS
(2021)
Article
Computer Science, Artificial Intelligence
Jun Ma, Jianan Chen, Matthew Ng, Rui Huang, Yu Li, Chen Li, Xiaoping Yang, Anne L. Martel
Summary: This paper provides a comprehensive review of segmentation loss functions in deep learning-based methods and conducts a large-scale analysis, finding that compound loss functions are the most robust. The study results suggest that no single loss function consistently outperforms others across different segmentation tasks.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Computer Science, Artificial Intelligence
Chetan L. Srinidhi, Seung Wook Kim, Fu-Der Chen, Anne L. Martel
Summary: In this work, we propose novel strategies to leverage both task-agnostic and task-specific unlabeled data for improving representation learning in computational histopathology. Experimental results show that our method outperforms state-of-the-art self-supervised and supervised baselines under limited labeled data, demonstrating the effectiveness of bootstrapping self-supervised pretrained features for task-specific semi-supervised learning.
MEDICAL IMAGE ANALYSIS
(2022)
Editorial Material
Engineering, Biomedical
Anne Martel, Danail Stoyanov, Diana Mateus, Leo Joskowicz, Purang Abolmaesumi
INTERNATIONAL JOURNAL OF COMPUTER ASSISTED RADIOLOGY AND SURGERY
(2021)
Article
Computer Science, Artificial Intelligence
Fumin Guo, Matthew Ng, Grey Kuling, Graham Wright
Summary: In this study, a globally optimal label fusion (GOLF) algorithm was developed to improve the performance of deep learning for cardiac MRI segmentation using small datasets and sparse annotations. The study also proposed an uncertainty-guided coupled continuous kernel cut (ugCCKC) algorithm with shape priors to further enhance the segmentation. Experimental results showed that these methods could achieve good segmentation results on relatively small and sparsely annotated datasets, and could be adapted to different datasets.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Imaging Science & Photographic Technology
Grey Kuling, Belinda Curpen, Anne L. Martel
Summary: This study utilized a pre-trained BERT model on breast radiology reports and section segmentation to improve the accuracy of radiology reports and the extraction of clinically relevant information.
JOURNAL OF IMAGING
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Julie Bilocq-Lacoste, Romuald Ferre, Grey Kuling, Anne L. Martel, Pascal N. Tyrrell, Siying Li, Guan Wang, Belinda Curpen
Summary: This study aimed to investigate the influence of MRI features and molecular subtypes on the detectability of breast cancers on MRI in high-risk patients. The results showed no significant correlation between MRI characteristics, receptor types, and the frequency of missed cancers. The main factors for misinterpreted lesions were multiple breast lesions, prior biopsy/surgery, and long-term stability. Lesions were mostly overlooked due to their small size and high background parenchymal enhancement.
Article
Oncology
Alison Min-Yan Cheung, Dan Wang, Kela Liu, Tyna Hope, Mayan Murray, Fiona Ginty, Sharon Nofech-Mozes, Anne Louise Martel, Martin Joel Yaffe
Summary: The study revealed differences in expression levels and distributions of estrogen receptor and progesterone receptor in breast cancer, as well as distinct expression patterns in different subtypes. Protein marker multiplexing and quantitative image analysis demonstrated marked heterogeneity in protein co-expression signatures and cellular arrangement within each breast cancer subtype.
BREAST CANCER RESEARCH
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Jianan Chen, Helen M. C. Cheung, Laurent Milot, Anne L. Martel
Summary: The paper introduces an AMINN neural network model based on radiomic features, which improves survival outcome prediction for patients with multifocal colorectal cancer liver metastases. The incorporation of a two-step normalization technique enhances the training of deep neural networks in the model.
MEDICAL IMAGE COMPUTING AND COMPUTER ASSISTED INTERVENTION - MICCAI 2021, PT V
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Nicholas Petrick, Shazia Akbar, Kenny H. Cha, Sharon Nofech-Mozes, Berkman Sahiner, Marios A. Gavrielides, Jayashree Kalpathy-Cramer, Karen Drukker, Anne L. Martel
Summary: The BreastPathQ Challenge demonstrated the potential of artificial intelligence/machine learning algorithms in estimating tumor cellularity in breast cancer histology images, showing that they may approach human performance and improve efficiency and reduce reader variability in clinical practice.
JOURNAL OF MEDICAL IMAGING
(2021)