Recurrent Residual Convolutional Neural Network based on U-Net (R2U-Net) for Medical Image Segmentation

By Md. Zahangir Alom, Mahmudul Hasan, C. Yakopcic, Tarek M. Taha, Vijayan K. Asari. arXiv.org, 2018

1,278 citations recorded by the source corpus.

Deep learning (DL) based semantic segmentation methods have been providing state-of-the-art performance in the last few years. More specifically, these techniques have been successfully applied to medical image classification, segmentation, and detection tasks. One deep learning technique, U-Net, has become one of the most popular for these applications. In this paper, we propose a Recurrent Convolutional Neural Network (RCNN) based on U-Net as well as a Recurrent Residual Convolutional Neural Network (RRCNN) based on U-Net models, which are named RU-Net and R2U-Net respectively. The proposed models utilize the power of U-Net, Residual Network, as well as RCNN. There are several advantages of these proposed architectures for segmentation tasks. First, a residual unit helps when training deep architecture. Second, feature accumulation with recurrent residual convolutional layers ensures better feature representation for segmentation tasks. Third, it allows us to design better U-Net architecture with same number of network parameters with better performance for medical image segmentation. The proposed models are tested on three benchmark datasets such as blood vessel segmentation in retina images, skin cancer segmentation, and lung lesion segmentation. The experimental results show superior performance on segmentation tasks compared to equivalent models including U-Net and residual U-Net (ResU-Net).

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/27c761258329eddb90b64d52679ff190cb4527b5

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

Dual-Sampling Attention Network for Diagnosis of COVID-19 From Community Acquired Pneumonia (2020, methodology): “[42] utilize U-net, residual network and recurrent CNN for lung lesion segmentation.”
DUNet: A deformable network for retinal vessel segmentation (2018, neutral): “…- 0.8554 0.9862 0.9729 0.9928 Fu et al. [33] DNN 2016 - 0.7412 - 0.9585 - Roychowdhury et al. [49] STA 2017 - 0.7720 0.9730 0.9510 0.9690 Chen et al. [50] DNN 2017 - 0.7295 0.9696 0.9449 0.9557 Alom et al. [51] DNN 2018 - 0.8298 0.9862 0.9712 0.9914 DUNet DNN 2018 0.8856 0.7428 0.9920 0.9729 0.9868”
Embracing Imperfect Datasets: A Review of Deep Learning Solutions for Medical Image Segmentation (2019, background): “blocks Alom et al. (2018), dense convolution blocks Li et al.”
The History Began from AlexNet: A Comprehensive Survey on Deep Learning Approaches (2018, background): “Very recent, an improved version of U-Net[300] with Recurrent Residual Convolutional Neural Networks (RRCNN) which is named R2U-Net [301].”
U-Net and Its Variants for Medical Image Segmentation: A Review of Theory and Applications (2020, methodology): “[52], [66] devised a U-net model containing both recurrent connections and residual connections.”
TransMorph: Transformer for unsupervised medical image registration (2021, positive): “Many works in other fields (e.g., image segmentation) have addressed this limitation of U-Net (Zhou et al., 2019; Jha et al., 2019; Devalla et al., 2018; Alom et al., 2018).”
DS-TransUNet: Dual Swin Transformer U-Net for Medical Image Segmentation (2021, methodology): “Swin-Large 192 768 [2, 2, 18, 2] [6, 12, 24, 48] 7”
IterNet: Retinal Image Segmentation Utilizing Structural Redundancy in Vessel Networks (2019, result): “We also compared the results with some existing models, including the aforementioned three UNet-based models, Residual UNet [25], Recurrent UNet [25], R2UNet [25], and one iterate prediction methods, i.”
DCSAU-Net: A Deeper and More Compact Split-Attention U-Net for Medical Image Segmentation (2022, neutral): “…0.940±0.048 0 .845±0.143 0.866±0.125 0.849±0.117 0.757±0.147 ResUNet++ [24] 0.934±0.051 0 .838±0.118 0.858±0.101 0 .840±0.086 0 .736±0.121 R2U-Net [ 44] 0.933±0.056 0 .852±0.122 0.831±0.136 0.834±0.112 0.744±0.128 DoubleU-Net [22] 0.937±0.052 0 .833±0.120 0 .896±0.084 0 .858±0.089 0 .763±0.130…”
Medical Image Segmentation Using Deep Learning: A Survey (2020, methodology): “[32] proposed a medical image segmentation method that combines ResUNet with RNN.”
Nuclei Segmentation with Recurrent Residual Convolutional Neural Networks based U-Net (R2U-Net) (2018, background): “This advanced and new hybrid deep learning model generalizes U-Net, residual, and recurrent convolutional neural networks [20].”
UCTransNet: Rethinking the Skip Connections in U-Net from a Channel-wise Perspective with Transformer (2021, methodology): “Following the popularity of U-Net, many novel models have been proposed such as UNet++ (Zhou et al. 2018), Attention U-Net (Oktay et al. 2018), DenseUNet (Li et al. 2018), R2U-Net (Alom et al. 2018), and UNet 3+ (Huang et al. 2020), which are specially designed for medical image segmentation and ach”
Bi-Directional ConvLSTM U-Net with Densley Connected Convolutions (2019, methodology): “Many extensions of U-Net have been proposed so far [2, 19].”
Can You Learn an Algorithm? Generalizing from Easy to Hard Problems with Recurrent Networks (2021, methodology): “Also, the basic neural architectures we use are not new and build upon prior studies of weight sharing and recurrence [Pinheiro and Collobert, 2014, Liang and Hu, 2015, Alom et al., 2018, Bai et al., 2018, 2019, Lan et al., 2020, Jaegle et al., 2021].”