IRUS Total

Brain volume estimation from post-mortem newborn and fetal MRI

File Description SizeFormat 
Brain volume estimation from post-mortem newborn and fetal MRI.pdfPublished version966.64 kBAdobe PDFView/Open
Title: Brain volume estimation from post-mortem newborn and fetal MRI
Authors: Orasanu, E
Melbourne, A
Cardoso, MJ
Modat, M
Taylor, AM
Thayyil, S
Ourselin, S
Item Type: Journal Article
Abstract: Objective: Minimally invasive autopsy using post-mortem magnetic resonance imaging (MRI) is a valid alternative to conventional autopsy in fetuses and infants. Estimation of brain weight is an integral part of autopsy, but manual segmentation of organ volumes on MRI is labor intensive and prone to errors, therefore unsuitable for routine clinical practice. In this paper we aim to show that volumetric measurements of the post-mortem fetal and neonatal brain can be accurately estimated using semi-automatic techniques and a high correlation can be found with the weights measured from conventional autopsy results. Methods: The brains of 17 newborn subjects, part of Magnetic Resonance Imaging Autopsy Study (MaRIAS), were segmented from post-mortem MR images into cerebrum, cerebellum and brainstem using a publicly available neonate brain atlas and semi-automatic segmentation algorithm. The results of the segmentation were averaged to create a new atlas, which was then used for the automated atlas-based segmentation of 17 MaRIAS fetus subjects. As validation, we manually segmented the MR images from 8 subjects of each cohort and compared them with the automatic ones. The semi-automatic estimation of cerebrum weight was compared with the results of the conventional autopsy. Results: The Dice overlaps between the manual and automatic segmentations are 0.991 and 0.992 for cerebrum, 0.873 and 0.888 for cerebellum and 0.819 and 0.815 for brainstem, for newborns and fetuses, respectively. Excellent agreement was obtained between the estimated MR weights and autopsy gold standard ones: mean absolute difference of 5 g and 2% maximum error for the fetus cohort and mean absolute difference of 20 g and 11% maximum error for the newborn one. Conclusions: The high correlation between the obtained segmentation and autopsy weights strengthens the idea of using post-mortem MRI as an alternative for conventional autopsy of the brain.
Issue Date: 23-Oct-2014
Date of Acceptance: 15-Oct-2014
URI: http://hdl.handle.net/10044/1/58116
DOI: https://dx.doi.org/10.1016/j.nicl.2014.10.007
ISSN: 2213-1582
Publisher: Elsevier
Start Page: 438
End Page: 444
Journal / Book Title: NeuroImage: Clinical
Volume: 6
Copyright Statement: © 2014 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/3.0/).
Keywords: Science & Technology
Life Sciences & Biomedicine
Neurosciences & Neurology
Post-mortem MRI
Brain volumes
CI, confidence interval
CSF, cerebrospinal fluid
EM, expectation maximization
GA, gestational age
GW, gestational weeks
MRI, magnetic resonance imaging
MaRIAS, Magnetic Resonance Imaging Autopsy Study
Infant Death
Infant, Newborn
Magnetic Resonance Imaging
Publication Status: Published
Appears in Collections:Department of Medicine (up to 2019)