Fully automatic acute ischemic lesion segmentation in DWI using convolutional neural networks

Liang, Chunnian; Bentley, Paul; Rueckert, Daniel

doi:10.1016/j.nicl.2017.06.016

Cited by 263 publications

(170 citation statements)

References 32 publications

Supporting

Mentioning

162

Contrasting

Unclassified

Order By: Relevance

“…We propose to aggregate feature maps convolved by three kernels, namely 1 × 1, 3 × 3, and 5 × 5. Inspired by the DeepLab [17], the deconvolution with a 5×5 kernel is replaced by a dilated deconvolution with a 3×3 kernel, which is more efficient in memory. To further limit the size of the parameter space, a bottleneck deconvolution is used in each branch.…”

Section: Residual Inception Blockmentioning

confidence: 99%

See 1 more Smart Citation

DRINet for Medical Image Segmentation

Chen

Bentley

Mori

et al. 2018

IEEE Trans. Med. Imaging

Self Cite

265

134

View full text Add to dashboard Cite

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Ut purus elit, vestibulum ut, placerat ac, adipiscing vitae, felis. Curabitur dictum gravida mauris. Nam arcu libero, nonummy eget, consectetuer id, vulputate a, magna. Donec vehicula augue eu neque. Pellentesque habitant morbi tristique senectus et netus et malesuada fames ac turpis egestas. Mauris ut leo. Cras viverra metus rhoncus sem. Nulla et lectus vestibulum urna fringilla ultrices. Phasellus eu tellus sit amet tortor gravida placerat. Integer sapien est, iaculis in, pretium quis, viverra ac, nunc. Praesent eget sem vel leo ultrices bibendum. Aenean faucibus. Morbi dolor nulla, malesuada eu, pulvinar at, mollis ac, nulla. Curabitur auctor semper nulla. Donec varius orci eget risus. Duis nibh mi, congue eu, accumsan eleifend, sagittis quis, diam. Duis eget orci sit amet orci dignissim rutrum.Nam dui ligula, fringilla a, euismod sodales, sollicitudin vel, wisi. Morbi auctor lorem non justo. Nam lacus libero, pretium at, lobortis vitae, ultricies et, tellus. Donec aliquet, tortor sed accumsan bibendum, erat ligula aliquet magna, vitae ornare odio metus a mi. Abstract-Convolutional neural networks (CNNs) have revolutionized medical image analysis over the past few years. The UNet architecture is one of the most well-known CNN architectures for semantic segmentation and has achieved remarkable successes in many different medical image segmentation applications. The U-Net architecture consists of standard convolution layers, pooling layers, and upsampling layers. These convolution layers learn representative features of input images and construct segmentations based on the features. However, the features learned by standard convolution layers are not distinctive when the differences among different categories are subtle in terms of intensity, location, shape, and size. In this paper, we propose a novel CNN architecture, called Dense-Res-Inception Net (DRINet), which addresses this challenging problem. The proposed DRINet consists of three blocks, namely a convolutional block with dense connections, a deconvolutional block with residual Inception modules, and an unpooling block. Our proposed architecture outperforms the U-Net in three different challenging applications, namely multi-class segmentation of cerebrospinal fluid (CSF) on brain CT images, multi-organ segmentation on abdominal CT images, multi-class brain tumour segmentation on MR images.Index Terms-Convolutional neural network, medical image segmentation, brain atrophy, abdominal organ segmentation.

show abstract

Section: Residual Inception Blockmentioning

confidence: 99%

“…This is because the thickness of the clinical CT images is large (up to 7mm) and resampling the images can introduce inaccuracies and interpolation artefacts. In terms of the image intensity normalization, we employed the similar strategy as described in [17]. We normalized CT images on a per slice basis.…”

Section: A Csf Segmentation In Ct Imagesmentioning

confidence: 99%

DRINet for Medical Image Segmentation

Chen

Bentley

Mori

et al. 2018

IEEE Trans. Med. Imaging

Self Cite

265

134

View full text Add to dashboard Cite

show abstract

“…However, this model also has problems objectively: 1) most medical images have weak edges, which make the network perform better classification difficultly and cause partial loss of details; 2) structurally, simply superimposing the convolution layer can improve the expression ability of network, which will increase a mass of parameters and make training network difficult. Up to now, many scholars have proposed many improved methods for the U-Net [3,4,10,12,13]. Chen L et al, "proposed DRINet [3] and a fully automatic acute ischemic lesion segmentation model (EDD+MUSCLE Net) [4].…”

Section: Introductionmentioning

confidence: 99%

A Triple Residual Multiscale Fully Convolutional Network Model for Multimodal Infant Brain MRI Segmentation

2020

KSII TIIS

View full text Add to dashboard Cite

The accurate segmentation of infant brain MR image into white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) is very important for early studying of brain growing patterns and morphological changes in neurodevelopmental disorders. Because of inherent myelination and maturation process, the WM and GM of babies (between 6 and 9 months of age) exhibit similar intensity levels in both T1-weighted (T1w) and T2-weighted (T2w) MR images in the isointense phase, which makes brain tissue segmentation very difficult. We propose a deep network architecture based on U-Net, called Triple Residual Multiscale Fully Convolutional Network (TRMFCN), whose structure exists three gates of input and inserts two blocks: residual multiscale block and concatenate block. We solved some difficulties and completed the segmentation task with the model. Our model outperforms the U-Net and some cutting-edge deep networks based on U-Net in evaluation of WM, GM and CSF. The data set we used for training and testing comes from iSeg-2017 challenge (http://iseg2017.web.unc.edu).

show abstract

“…Zhiyang Liu et al [4] proposed a residual-structured fully convolutional network (Res FCN) that does 2D slice-based segmentation using dice coefficient as the loss function. Chen et al [5] made use of EDD Net (an ensemble of DeconvNets) and MUSCLE Net (Multiscale Convolutional Label Evaluation Net) to segment and refine the lesion by removing False positives. By embedding a residual unit into a U shaped network, Liangliang Liu et al [6] achieved an average dice score of 88.43 in SPES.…”

Section: Introductionmentioning

confidence: 99%

Parallel Capsule Net for Ischemic Stroke Segmentation

Sharique

Pundarikaksha

Sridar

et al. 2019

Preprint

View full text Add to dashboard Cite

Stroke is one of the leading causes of disability. Segmentation of ischemic stroke could help in planning an optimal treatment. Currently, radiologists use manual segmentation, which can often be time-consuming, laborious and error-prone. Automatic segmentation of ischemic stroke in MRI brain images is a challenging problem due to its small size, multiple occurrences and the need to use multiple image modalities. In this paper, we propose a new architecture for image segmentation, called Parallel Capsule Net, which uses max pooling in every parallel pathways along with dense connections between the parallel layers. We hypothesise that the spatial information lost due to max pooling in these layers can be retrieved by the use of such dense connections. In order to combine the information encoded by the parallel layers, outputs of the layers are concatenated before upsampling. We also propose the use of a modified loss function which consists of a regional term (Generalized Dice loss + Focal Loss) and a boundary term (Boundary loss) to address the problem of class imbalance which is prevalent in medical images. We achieved a competitive Dice score of 0.754, on ISLES SISS data set, compared to a score of 0.67 reported in earlier studies. We also obtained a Dice score of 0.902 with another popular data set, ATLAS. The proposed parallel capsule net can be extended to other similar medical image segmentation problems.

show abstract

Fully automatic acute ischemic lesion segmentation in DWI using convolutional neural networks

Cited by 263 publications

References 32 publications

DRINet for Medical Image Segmentation

DRINet for Medical Image Segmentation

A Triple Residual Multiscale Fully Convolutional Network Model for Multimodal Infant Brain MRI Segmentation

Parallel Capsule Net for Ischemic Stroke Segmentation

Contact Info

Product

Resources

About