Spatially Covariant Lesion Segmentation

  title={Spatially Covariant Lesion Segmentation},
  author={Hang Zhang and Rongguang Wang and Jinwei Zhang and Dongdong Liu and Chao Li and Jiahao Li},
Compared to natural images, medical images usually show stronger visual patterns and therefore this adds flexibility and elasticity to resource-limited clinical applications by injecting proper priors into neural networks. In this paper, we propose spatially covariant pixel-aligned classifier (SCP) to improve the computational efficiency and meantime maintain or increase accuracy for lesion segmentation. SCP relaxes the spatial invariance constraint imposed by convolutional operations and… 

Figures and Tables from this paper

DeDA: Deep Directed Accumulator

A simple yet effective image processing operation, deep directed accumulator (DeDA), that provides a new perspective for injecting domain-specific inductive biases (priors) into neural networks for rim+ lesion identification.

Multiple Sclerosis Lesion Segmentation with Tiramisu and 2.5D Stacked Slices

A fully convolutional densely connected network (Tiramisu) for multiple sclerosis (MS) lesion segmentation using stacked slices from all three anatomical planes to achieve a 2.5D method that can make more accurate segmentation by combining information of different forms.

Efficient Folded Attention for 3D Medical Image Reconstruction and Segmentation

A folded attention (FA) approach to improve the computational efficiency of traditional attention methods on 3D medical images and can substantially reduce the computational complexity and GPU memory consumption.

Geometric Loss For Deep Multiple Sclerosis Lesion Segmentation

A new geometric loss formula is proposed to address the data imbalance and exploit the geometric property of MS lesions and it is shown that traditional region-based and boundary-aware loss functions can be associated with the formula.

RSANet: Recurrent Slice-Wise Attention Network for Multiple Sclerosis Lesion Segmentation

A novel recurrent slice-wise attention network (RSANet) is proposed, which models 3D MRI images as sequences of slices and captures long-range dependencies through a recurrent manner to utilize contextual information of MS lesions.

Standardized Assessment of Automatic Segmentation of White Matter Hyperintensities and Results of the WMH Segmentation Challenge

There is a cluster of four methods that rank significantly better than the other methods, with one clear winner, and the inter-scanner robustness ranking shows that not all the methods generalize to unseen scanners.

Hyper-Convolution Networks for Biomedical Image Segmentation

This paper proposes a powerful novel building block, the hyper-convolution, which implicitly represents the convolution kernel as a function of kernel coordinates, and demonstrates that replacing regular convolutions with hyper-convolutions leads to more efficient architectures that achieve improved accuracy.

Fidelity imposed network edit (FINE) for solving ill-posed image reconstruction

CTformer: convolution-free Token2Token dilated vision transformer for low-dose CT denoising

Objective. Low-dose computed tomography (LDCT) denoising is an important problem in CT research. Compared to the normal dose CT, LDCT images are subjected to severe noise and artifacts. Recently in

nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation

Biomedical imaging is a driver of scientific discovery and a core component of medical care and is being stimulated by the field of deep learning. While semantic segmentation algorithms enable image