Cargando…

Attention Gated Networks: Learning to Leverage Salient Regions in Medical Images

We propose a novel attention gate (AG) model for medical image analysis that automatically learns to focus on target structures of varying shapes and sizes. Models trained with AGs implicitly learn to suppress irrelevant regions in an input image while highlighting salient features useful for a spec...

Descripción completa

Detalles Bibliográficos
Autores principales: Schlemper, Jo, Oktay, Ozan, Schaap, Michiel, Heinrich, Mattias, Kainz, Bernhard, Glocker, Ben, Rueckert, Daniel
Formato: Online Artículo Texto
Lenguaje:English
Publicado: 2019
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7610718/
https://www.ncbi.nlm.nih.gov/pubmed/30802813
http://dx.doi.org/10.1016/j.media.2019.01.012
_version_ 1783605222972588032
author Schlemper, Jo
Oktay, Ozan
Schaap, Michiel
Heinrich, Mattias
Kainz, Bernhard
Glocker, Ben
Rueckert, Daniel
author_facet Schlemper, Jo
Oktay, Ozan
Schaap, Michiel
Heinrich, Mattias
Kainz, Bernhard
Glocker, Ben
Rueckert, Daniel
author_sort Schlemper, Jo
collection PubMed
description We propose a novel attention gate (AG) model for medical image analysis that automatically learns to focus on target structures of varying shapes and sizes. Models trained with AGs implicitly learn to suppress irrelevant regions in an input image while highlighting salient features useful for a specific task. This enables us to eliminate the necessity of using explicit external tissue/organ localisation modules when using convolutional neural networks (CNNs). AGs can be easily integrated into standard CNN models such as VGG or U-Net architectures with minimal computational overhead while increasing the model sensitivity and prediction accuracy. The proposed AG models are evaluated on a variety of tasks, including medical image classification and segmentation. For classification, we demonstrate the use case of AGs in scan plane detection for fetal ultrasound screening. We show that the proposed attention mechanism can provide efficient object localisation while improving the overall prediction performance by reducing false positives. For segmentation, the proposed architecture is evaluated on two large 3D CT abdominal datasets with manual annotations for multiple organs. Experimental results show that AG models consistently improve the prediction performance of the base architectures across different datasets and training sizes while preserving computational efficiency. Moreover, AGs guide the model activations to be focused around salient regions, which provides better insights into how model predictions are made. The source code for the proposed AG models is publicly available.
format Online
Article
Text
id pubmed-7610718
institution National Center for Biotechnology Information
language English
publishDate 2019
record_format MEDLINE/PubMed
spelling pubmed-76107182021-05-04 Attention Gated Networks: Learning to Leverage Salient Regions in Medical Images Schlemper, Jo Oktay, Ozan Schaap, Michiel Heinrich, Mattias Kainz, Bernhard Glocker, Ben Rueckert, Daniel Med Image Anal Article We propose a novel attention gate (AG) model for medical image analysis that automatically learns to focus on target structures of varying shapes and sizes. Models trained with AGs implicitly learn to suppress irrelevant regions in an input image while highlighting salient features useful for a specific task. This enables us to eliminate the necessity of using explicit external tissue/organ localisation modules when using convolutional neural networks (CNNs). AGs can be easily integrated into standard CNN models such as VGG or U-Net architectures with minimal computational overhead while increasing the model sensitivity and prediction accuracy. The proposed AG models are evaluated on a variety of tasks, including medical image classification and segmentation. For classification, we demonstrate the use case of AGs in scan plane detection for fetal ultrasound screening. We show that the proposed attention mechanism can provide efficient object localisation while improving the overall prediction performance by reducing false positives. For segmentation, the proposed architecture is evaluated on two large 3D CT abdominal datasets with manual annotations for multiple organs. Experimental results show that AG models consistently improve the prediction performance of the base architectures across different datasets and training sizes while preserving computational efficiency. Moreover, AGs guide the model activations to be focused around salient regions, which provides better insights into how model predictions are made. The source code for the proposed AG models is publicly available. 2019-04-01 2019-02-05 /pmc/articles/PMC7610718/ /pubmed/30802813 http://dx.doi.org/10.1016/j.media.2019.01.012 Text en https://creativecommons.org/licenses/by-nc-nd/4.0/This manuscript version is made available under the CC-BY-NC-ND 4.0 license (http://creativecommons.org/licenses/by-nc-nd/4.0/ (https://creativecommons.org/licenses/by-nc-nd/4.0/) )
spellingShingle Article
Schlemper, Jo
Oktay, Ozan
Schaap, Michiel
Heinrich, Mattias
Kainz, Bernhard
Glocker, Ben
Rueckert, Daniel
Attention Gated Networks: Learning to Leverage Salient Regions in Medical Images
title Attention Gated Networks: Learning to Leverage Salient Regions in Medical Images
title_full Attention Gated Networks: Learning to Leverage Salient Regions in Medical Images
title_fullStr Attention Gated Networks: Learning to Leverage Salient Regions in Medical Images
title_full_unstemmed Attention Gated Networks: Learning to Leverage Salient Regions in Medical Images
title_short Attention Gated Networks: Learning to Leverage Salient Regions in Medical Images
title_sort attention gated networks: learning to leverage salient regions in medical images
topic Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7610718/
https://www.ncbi.nlm.nih.gov/pubmed/30802813
http://dx.doi.org/10.1016/j.media.2019.01.012
work_keys_str_mv AT schlemperjo attentiongatednetworkslearningtoleveragesalientregionsinmedicalimages
AT oktayozan attentiongatednetworkslearningtoleveragesalientregionsinmedicalimages
AT schaapmichiel attentiongatednetworkslearningtoleveragesalientregionsinmedicalimages
AT heinrichmattias attentiongatednetworkslearningtoleveragesalientregionsinmedicalimages
AT kainzbernhard attentiongatednetworkslearningtoleveragesalientregionsinmedicalimages
AT glockerben attentiongatednetworkslearningtoleveragesalientregionsinmedicalimages
AT rueckertdaniel attentiongatednetworkslearningtoleveragesalientregionsinmedicalimages