Fully automatic, multi-organ segmentation in normal whole body magnetic resonance imaging (MRI), using classification forests (CFs), convolutional neural networks (CNNs) and a multi-atlas (MA) approach.

File Description SizeFormat 
lavdas2017fully.pdfAccepted version2.35 MBAdobe PDFView/Open
Title: Fully automatic, multi-organ segmentation in normal whole body magnetic resonance imaging (MRI), using classification forests (CFs), convolutional neural networks (CNNs) and a multi-atlas (MA) approach.
Authors: Lavdas, I
Glocker, B
Kamnitsas, K
Rueckert, D
Mair, H
Sandhu, A
Taylor, SA
Aboagye, EO
Rockall, AG
Item Type: Journal Article
Abstract: PURPOSE: As part of a programme to implement automatic lesion detection methods for whole body magnetic resonance imaging (MRI) in oncology, we have developed, evaluated and compared three algorithms for fully automatic, multi-organ segmentation in healthy volunteers. METHODS: The first algorithm is based on classification forests (CFs), the second is based on 3D convolutional neural networks (CNNs) and the third algorithm is based on a multi-atlas (MA) approach. We examined data from 51 healthy volunteers, scanned prospectively with a standardised, multi-parametric whole body MRI protocol at 1.5T. The study was approved by the local ethics committee and written consent was obtained from the participants. MRI data were used as input data to the algorithms, while training was based on manual annotation of the anatomies of interest by clinical MRI experts. Five-fold cross-validation experiments were run on 34 artefact-free subjects. We report three overlap and three surface distance metrics to evaluate the agreement between the automatic and manual segmentations, namely the Dice similarity coefficient (DSC), recall (RE), precision (PR), average surface distance (ASD), root mean square surface distance (RMSSD) and Hausdorff distance (HD). Analysis of variances was used to compare pooled label metrics between the three algorithms and the DSC on a 'per-organ' basis. A Mann-Whitney U test was used to compare the pooled metrics between CFs and CNNs and the DSC on a 'per-organ' basis, when using different imaging combinations as input for training. RESULTS: All three algorithms resulted in robust segmenters that were effectively trained using a relatively small number of data sets, an important consideration in the clinical setting. Mean overlap metrics for all the segmented structures were: CFs: DSC=0.70±0.18, RE=0.73±0.18, PR=0.71±0.14, CNNs: DSC=0.81±0.13, RE=0.83±0.14, PR=0.82±0.10, MA: DSC=0.71±0.22, RE=0.70±0.34, PR=0.77±0.15. Mean surface distance metrics for all the segmented structures were: CFs: ASD=13.5±11.3 mm, RMSSD=34.6±37.6 mm and HD=185.7±194.0 mm, CNNs; ASD=5.48±4.84 mm, RMSSD=17.0±13.3 mm and HD=199.0±101.2 mm, MA: ASD=4.22±2.42 mm, RMSSD=6.13±2.55 mm and HD=38.9±28.9 mm. The pooled performance of CFs improved when all imaging combinations (T2w+T1w+DWI) were used as input, while the performance of CNNs deteriorated, but in neither case, significantly. CNNs with T2w images as input, performed significantly better than CFs with all imaging combinations as input for all anatomical labels, except for the bladder. CONCLUSIONS: Three state-of-the-art algorithms were developed and used to automatically segment major organs and bones in whole body MRI; good agreement to manual segmentations performed by clinical MRI experts was observed. CNNs perform favourably, when using T2w volumes as input. Using multi-modal MRI data as input to CNNs did not improve the segmentation performance. This article is protected by copyright. All rights reserved.
Issue Date: 30-Jul-2017
Date of Acceptance: 10-Jul-2017
URI: http://hdl.handle.net/10044/1/50095
DOI: https://dx.doi.org/10.1002/mp.12492
ISSN: 0094-2405
Publisher: American Association of Physicists in Medicine
Start Page: 5210
End Page: 5220
Journal / Book Title: Medical Physics
Volume: 44
Issue: 10
Copyright Statement: This is the peer reviewed version of the following article: Lavdas, I., Glocker, B., Kamnitsas, K., Rueckert, D., Mair, H., Sandhu, A., Taylor, S. A., Aboagye, E. O. and Rockall, A. G. (2017), Fully automatic, multiorgan segmentation in normal whole body magnetic resonance imaging (MRI), using classification forests (CFs), convolutional neural networks (CNNs), and a multi-atlas (MA) approach. Med. Phys.. doi:10.1002/mp.12492, which has been published in final form at https://dx.doi.org/10.1002/mp.12492. This article may be used for non-commercial purposes in accordance With Wiley Terms and Conditions for self-archiving.
Sponsor/Funder: Cancer Research UK
Imperial College Healthcare NHS Trust- BRC Funding
National Institute for Health Research
Medical Research Council (MRC)
Funder's Grant Number: C2536/A16584
RDC04 79560
EME/13/122/01
MR/N020782/1
Keywords: classification forests
convolutional neural networks
fully automatic segmentation
multi-atlas segmentation
whole body MRI
0299 Other Physical Sciences
0903 Biomedical Engineering
Nuclear Medicine & Medical Imaging
Publication Status: Published online
Appears in Collections:Faculty of Engineering
Division of Surgery
Computing
Division of Cancer
Faculty of Medicine



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Creative Commonsx