On the Compactness, Efficiency, and Representation of 3D Convolutional Networks: Brain Parcellation as a Pretext Task

By Wenqi Li, Guotai Wang, Lucas Fidon, Sébastien Ourselin, M. Jorge Cardoso, Tom Kamiel Magda Vercauteren. 2017

364 citations recorded by the source corpus.

Deep convolutional neural networks are powerful tools for learning visual representations from images. However, designing efficient deep architectures to analyse volumetric medical images remains challenging. This work investigates efficient and flexible elements of modern convolutional networks such as dilated convolution and residual connection. With these essential building blocks, we propose a high-resolution, compact convolutional network for volumetric image segmentation. To illustrate its efficiency of learning 3D representation from large-scale image data, the proposed network is validated with the challenging task of parcellating 155 neuroanatomical structures from brain MR images. Our experiments show that the proposed network architecture compares favourably with state-of-the-art volumetric segmentation networks while being an order of magnitude more compact. We consider the brain parcellation task as a pretext task for volumetric image segmentation; our trained network potentially provides a good starting point for transfer learning. Additionally, we show the feasibility of voxel-level uncertainty estimation using a sampling approximation through dropout.

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

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

An overview of deep learning in medical imaging focusing on MRI (2018, background): “for medical image analysis and image-guided therapy [320,321] http://niftynet.”
Aleatoric uncertainty estimation with test-time augmentation for medical image segmentation with convolutional neural networks (2018, neutral): “Zhu and Zabaras [13] used Stochastic Variational Gradient Descent (SVGD) to perform approximate Bayesian inference on uncertain CNN parameters.”
Short Acquisition Time PET/MR Pharmacokinetic Modelling Using CNNs (2018, background): “the ‘highresnet’ convolutional neural network with 20 convolutional layers [5], which uses a stack of residual dilated convolutions with increasingly large dilation factors.”
DeepIGeoS: A Deep Interactive Geodesic Framework for Medical Image Segmentation (2017, methodology): “After generating X (cid:48) and Y (cid:48) , we We implemented our 2D networks by Caffe 1 [56] and 3D networks by Tensorﬂow 2 [57] using NiftyNet 3 [58].”
Interactive Medical Image Segmentation Using Deep Learning With Image-Specific Fine Tuning (2017, methodology): “Compared with large isotropic 3D receptive fields [8], it has less memory consumption [26].”
UNETR: Transformers for 3D Medical Image Segmentation (2021, methodology): “Several efforts have used atrous convolutional layers [9, 27, 18] to enlarge the”
3D Whole Brain Segmentation using Spatially Localized Atlas Network Tiles (2019, methodology): “The source code of patch-based network [44] and HC-Net [52] were obtained from the link provided by the publications (https://github.com/adbrebs/brain_segmentation and https://github.com/gift-surg/HighRes3DNet), whose hyperparameters had been kept the same as the publications.”
Standardized Assessment of Automatic Segmentation of White Matter Hyperintensities; Results of the WMH Segmentation Challenge (2019, methodology): “After that, candidate WMH is selected and possible false positives are pruned based on their location [59]. tignet a neural network with the HighResNet architecture [20].”
Distributed Weight Consolidation: A Brain Segmentation Case Study (2018, methodology): “Several deep neural network architectures have been proposed for brain segmentation, such as U-net [26], QuickNAT [27], HighResNet [18] and MeshNet [7, 6].”
Recurrent residual U-Net for medical image segmentation (2019, methodology): “In both cases, the proposed models outperformed the existing state-of-the-art methods, including SegNet, 11 U-Net, 13 ResU-Net, 31 and FCRN-38, 63 in terms of AUC and AC on all three datasets.”
Recurrent Residual Convolutional Neural Network based on U-Net (R2U-Net) for Medical Image Segmentation (2018, methodology): “High-Res3DNet was proposed using residual networks for 3D segmentation tasks in 2016 [37].”
Stochastic Filter Groups for Multi-Task CNNs: Learning Specialist and Generalist Convolution Kernels (2019, methodology): “Medical imaging network: We used the HighResNet architecture [18] for CT synthesis and organ segmentation.”
Automatic Brain Tumor Segmentation Using Cascaded Anisotropic Convolutional Neural Networks (2017, methodology): “…by employing a mixture of convolution and downsampling operations [12,9], maintaining high resolution multi-scale features with dilated convolution and residual connection [17,14,5], and handling training data imbalance issue by designing new loss functions [7,16] and sampling strategies [16].”