MSRF-Net: A Multi-Scale Residual Fusion Network for Biomedical Image Segmentation DOI Creative Commons
Abhishek Srivastava, Debesh Jha, Sukalpa Chanda

et al.

IEEE Journal of Biomedical and Health Informatics, Journal Year: 2021, Volume and Issue: 26(5), P. 2252 - 2263

Published: Dec. 23, 2021

Methods based on convolutional neural networks have improved the performance of biomedical image segmentation. However, most these methods cannot efficiently segment objects variable sizes and train small biased datasets, which are common for use cases. While exist that incorporate multi-scale fusion approaches to address challenges arising with sizes, they usually complex models more suitable general semantic segmentation problems. In this paper, we propose a novel architecture called Multi-Scale Residual Fusion Network (MSRF-Net), is specially designed medical The proposed MSRF-Net able exchange features varying receptive fields using Dual-Scale Dense (DSDF) block. Our DSDF block can information rigorously across two different resolution scales, our MSRF sub-network uses multiple blocks in sequence perform fusion. This allows preservation resolution, flow propagation both high- low-level obtain accurate maps. capture object variabilities provides results datasets. Extensive experiments demonstrate method outperforms cutting-edge four publicly available We achieve Dice Coefficient (DSC) 0.9217, 0.9420, 0.9224, 0.8824 Kvasir-SEG, CVC-ClinicDB, 2018 Data Science Bowl dataset, ISIC-2018 skin lesion challenge dataset respectively. further conducted generalizability tests achieved DSC 0.7921 0.7575 CVC-ClinicDB

Language: Английский

AI applications to medical images: From machine learning to deep learning DOI Open Access
Isabella Castiglioni, Leonardo Rundo, Marina Codari

et al.

Physica Medica, Journal Year: 2021, Volume and Issue: 83, P. 9 - 24

Published: March 1, 2021

Language: Английский

Citations

491
Unified Focal loss: Generalising Dice and cross entropy-based losses to handle class imbalanced medical image segmentation DOI Creative Commons
Michael Yeung, Evis Sala, Carola‐Bibiane Schönlieb

et al.

Computerized Medical Imaging and Graphics, Journal Year: 2021, Volume and Issue: 95, P. 102026 - 102026

Published: Dec. 13, 2021

Automatic segmentation methods are an important advancement in medical image analysis. Machine learning techniques, and deep neural networks particular, the state-of-the-art for most tasks. Issues with class imbalance pose a significant challenge datasets, lesions often occupying considerably smaller volume relative to background. Loss functions used training of algorithms differ their robustness imbalance, direct consequences model convergence. The commonly loss based on either cross entropy loss, Dice or combination two. We propose Unified Focal new hierarchical framework that generalises entropy-based losses handling imbalance. evaluate our proposed function five publicly available, imbalanced imaging datasets: CVC-ClinicDB, Digital Retinal Images Vessel Extraction (DRIVE), Breast Ultrasound 2017 (BUS2017), Brain Tumour Segmentation 2020 (BraTS20) Kidney 2019 (KiTS19). compare performance against six functions, across 2D binary, 3D binary multiclass tasks, demonstrating is robust consistently outperforms other functions. Source code available at: https://github.com/mlyg/unified-focal-loss.

Language: Английский

Citations

382
U-Net-Based Medical Image Segmentation DOI Creative Commons

Xiaoxia Yin,

Le Sun,

Yuhan Fu

et al.

Journal of Healthcare Engineering, Journal Year: 2022, Volume and Issue: 2022, P. 1 - 16

Published: April 15, 2022

Deep learning has been extensively applied to segmentation in medical imaging. U-Net proposed 2015 shows the advantages of accurate small targets and its scalable network architecture. With increasing requirements for performance imaging recent years, cited academically more than 2500 times. Many scholars have constantly developing This paper summarizes image technologies based on structure variants concerning their structure, innovation, efficiency, etc.; reviews categorizes related methodology; introduces loss functions, evaluation parameters, modules commonly imaging, which will provide a good reference future research.

Language: Английский

Citations

238
Segmentation of the multimodal brain tumor image used the multi-pathway architecture method based on 3D FCN DOI
Jindong Sun, Yanjun Peng, Yanfei Guo

et al.

Neurocomputing, Journal Year: 2020, Volume and Issue: 423, P. 34 - 45

Published: Oct. 22, 2020

Language: Английский

Citations

147
End-to-end deep learning of lane detection and path prediction for real-time autonomous driving DOI
Der‐Hau Lee, Jinn‐Liang Liu

Signal Image and Video Processing, Journal Year: 2022, Volume and Issue: 17(1), P. 199 - 205

Published: April 22, 2022

Language: Английский

Citations

73
A review on the use of deep learning for medical images segmentation DOI

Manar Aljabri,

Manal Alghamdi

Neurocomputing, Journal Year: 2022, Volume and Issue: 506, P. 311 - 335

Published: July 28, 2022

Language: Английский

Citations

73
Review of Semantic Segmentation of Medical Images Using Modified Architectures of UNET DOI Creative Commons

M. Krithika alias AnbuDevi,

K. Suganthi

Diagnostics, Journal Year: 2022, Volume and Issue: 12(12), P. 3064 - 3064

Published: Dec. 6, 2022

In biomedical image analysis, information about the location and appearance of tumors lesions is indispensable to aid doctors in treating identifying severity diseases. Therefore, it essential segment lesions. MRI, CT, PET, ultrasound, X-ray are different imaging systems obtain this information. The well-known semantic segmentation technique used medical analysis identify label regions images. aims divide images into with comparable characteristics, including intensity, homogeneity, texture. UNET deep learning network that segments critical features. However, UNETs basic architecture cannot accurately complex MRI This review introduces modified improved models suitable for increasing accuracy.

Language: Английский

Citations

71
A review on Deep Learning approaches for low-dose Computed Tomography restoration DOI Creative Commons
K.A.S.H. Kulathilake, Nor Aniza Abdullah, Aznul Qalid Md Sabri

et al.

Complex & Intelligent Systems, Journal Year: 2021, Volume and Issue: 9(3), P. 2713 - 2745

Published: May 30, 2021

Computed Tomography (CT) is a widely use medical image modality in clinical medicine, because it produces excellent visualizations of fine structural details the human body. In procedures, desirable to acquire CT scans by minimizing X-ray flux prevent patients from being exposed high radiation. However, these Low-Dose (LDCT) scanning protocols compromise signal-to-noise ratio images noise and artifacts over space. Thus, various restoration methods have been published past 3 decades produce high-quality LDCT images. More recently, as opposed conventional methods, Deep Learning (DL)-based approaches rather common due their characteristics data-driven, high-performance, fast execution. this study aims elaborate on role DL techniques critically review applications DL-based for restoration. To achieve aim, different aspects were analyzed. These include architectures, performance gains, functional requirements, diversity objective functions. The outcome highlights existing limitations future directions best our knowledge, there no previous reviews, which specifically address topic.

Language: Английский

Citations

85
BDANet: Multiscale Convolutional Neural Network With Cross-Directional Attention for Building Damage Assessment From Satellite Images DOI
Yu Shen, Sijie Zhu, Taojiannan Yang

et al.

IEEE Transactions on Geoscience and Remote Sensing, Journal Year: 2021, Volume and Issue: 60, P. 1 - 14

Published: May 27, 2021

Fast and effective responses are required when a natural disaster (e.g., earthquake, hurricane, etc.) strikes. Building damage assessment from satellite imagery is critical before relief effort deployed. With pair of pre- post-disaster images, building aims at predicting the extent to buildings. powerful ability feature representation, deep neural networks have been successfully applied assessment. Most existing works simply concatenate images as input network without considering their correlations. In this paper, we propose novel two-stage convolutional for Damage Assessment, called BDANet. first stage, U-Net used extract locations Then weights stage shared in second two-branch multi-scale employed backbone, where fed into separately. A cross-directional attention module proposed explore correlations between images. Moreover, CutMix data augmentation exploited tackle challenge difficult classes. The method achieves state-of-the-art performance on large-scale dataset -- xBD. code available https://github.com/ShaneShen/BDANet-Building-Damage-Assessment.

Language: Английский

Citations

69
Machine learning based liver disease diagnosis: A systematic review DOI
Rayyan Azam Khan, Yigang Luo, Fang‐Xiang Wu

et al.

Neurocomputing, Journal Year: 2021, Volume and Issue: 468, P. 492 - 509

Published: Sept. 6, 2021

Language: Английский

Citations

59