Anatomically Constrained Neural Networks (ACNNs): Application to Cardiac Image Enhancement and Segmentation

  title={Anatomically Constrained Neural Networks (ACNNs): Application to Cardiac Image Enhancement and Segmentation},
  author={Ozan Oktay and Enzo Ferrante and Konstantinos Kamnitsas and Mattias P. Heinrich and Wenjia Bai and Jose Caballero and Stuart A. Cook and Antonio de Marvao and Timothy J. W. Dawes and Declan P. O’Regan and Bernhard Kainz and Ben Glocker and Daniel Rueckert},
  journal={IEEE Transactions on Medical Imaging},
Incorporation of prior knowledge about organ shape and location is key to improve performance of image analysis approaches. [] Key Method State-of-the-art methods operate as pixel-wise classifiers where the training objectives do not incorporate the structure and inter-dependencies of the output. To overcome this limitation, we propose a generic training strategy that incorporates anatomical prior knowledge into CNNs through a new regularisation model, which is trained end-to-end.

Anatomical Priors for Image Segmentation via Post-Processing with Denoising Autoencoders

Post-DAE is proposed, a post-processing method based on denoising autoencoders (DAE) trained using only segmentation masks that can improve the quality of noisy and incorrect segmentation mask obtained with a variety of standard methods, by bringing them back to a feasible space, with almost no extra computational time.

3D spatial priors for semi-supervised organ segmentation with deep convolutional neural networks

STIPPLE provides a segmentation method effective with few labeled examples, which is crucial in the medical domain and offers an intuitive way to incorporate absolute position information by mimicking expert annotators.

Cardiac Segmentation With Strong Anatomical Guarantees

This paper presents a framework for producing cardiac image segmentation maps that are guaranteed to respect pre-defined anatomical criteria, while remaining within the inter-expert variability and always anatomically plausible without having to rely on a shape prior.

High-level Prior-based Loss Functions for Medical Image Segmentation: A Survey

Contour Transformer Network for One-Shot Segmentation of Anatomical Structures

This work forms anatomy segmentation as a contour evolution process and model the evolution behavior by graph convolutional networks (GCNs) and demonstrates that the one-shot learning method significantly outperforms non-learning-based methods and performs competitively to the state-of-the-art fully supervised deep learning methods.

Anatomical Priors in Convolutional Networks for Unsupervised Biomedical Segmentation

A generative probabilistic model is introduced that employs the learned prior through a convolutional neural network to compute segmentations in an unsupervised setting to facilitate CNN-based anatomical segmentation in a range of novel clinical problems, where few or no annotations are available and thus standard networks are not trainable.

Leveraging Global Binary Masks for Structure Segmentation in Medical Images

A framework leveraging recurring anatomical patterns through global binary masks for organ segmentation through U-Net was introduced to leverage the consistency of organs’ anatomical position and shape information in medical images and indicates that a surprising degree of position andshape information is encoded through globalbinary masks.

BB-UNet: U-Net With Bounding Box Prior

This paper introduces BB-UNet (Bounding Box U-Net), a deep learning model that integrates location as well as shape prior onto model training and outperforms state-of-the-art methods in fully supervised learning frameworks and registers relevant results given the weakly supervised domain.

Cardiac Segmentation from LGE MRI Using Deep Neural Network Incorporating Shape and Spatial Priors

A new method, based on deep neural networks (DNN), for fully automatic segmentation from late gadolinium enhancement MRI, which outperformed the conventional schemes and obtained a Dice score of 0.758 for myocardial segmentation, which compares with 0.757 from the inter-observer variations.



Iterative Multi-domain Regularized Deep Learning for Anatomical Structure Detection and Segmentation from Ultrasound Images

This paper proposes a multi-domain regularized deep learning method that achieves a superior detection and segmentation accuracy, outperforming other methods by a significant margin and demonstrating competitive capability even compared to human performance.

U-Net: Convolutional Networks for Biomedical Image Segmentation

It is shown that such a network can be trained end-to-end from very few images and outperforms the prior best method (a sliding-window convolutional network) on the ISBI challenge for segmentation of neuronal structures in electron microscopic stacks.

Topology Aware Fully Convolutional Networks for Histology Gland Segmentation

This work introduces the first deep network trained to encode geometric and topological priors of containment and detachment, and shows how the topology aware architecture outperforms competing methods by up to 10 % in both pixel-level accuracy and object-level Dice.

Multi-input Cardiac Image Super-Resolution Using Convolutional Neural Networks

Experimental results show that the CNN model outperforms state-of-the-art SR methods in terms of image quality while being computationally efficient and that image segmentation and motion tracking benefits more from SR-CNN when it is used as an initial upscaling method than conventional interpolation methods for the subsequent analysis.

Stratified Decision Forests for Accurate Anatomical Landmark Localization in Cardiac Images

A novel stratification based training model is proposed that achieves state-of-the-art landmark localization accuracy and outperforms standard regression and classification based approaches and is used in a multi-atlas segmentation to create a fully automatic segmentation pipeline, and the results show that it achieves state of theart segmentation accuracy.

A Fully Convolutional Neural Network for Cardiac Segmentation in Short-Axis MRI

This work proposes to tackle the problem of automated left and right ventricle segmentation through the application of a deep fully convolutional neural network architecture for pixel-wise labeling in cardiac magnetic resonance imaging.

Learning and Incorporating Shape Models for Semantic Segmentation

This work addresses the problem of incorporating shape priors within the FCN segmentation framework and demonstrates the utility of such a shape prior in robust handling of scenarios such as loss of contrast and artifacts.

Incorporating prior knowledge in medical image segmentation: a survey

This survey focuses on optimization-based methods that incorporate prior information into their frameworks and reviews and compares these methods in terms of the types of prior employed, the domain of formulation, and the optimization techniques.