Heart Segmentation
for the
Background
Members
of the departments of Anatomy & Embryology (A&E) and Medical
Informatics (KIK) of the AMC cooperate in the development of an application for
automatically fitting 2D sections of an embryonic mouse heart into 3D computer
reconstructions of mouse hearts. A prototype of this application, written in Matlab, is already finished. The application is being
developed with two main goals in mind. The first goal is to give users the opportunity
to get information about the orientation of their tissue sections. The second
goal is to collect and make available gene expression information within its
anatomic context.
In
research in cardiac development it is unavoidable to use serially sectioned
biological material for 3D computer reconstructions, for reasons like the
required level of detail or the limited penetration of staining agents into
whole mount staining procedures (Ruijter et al. 2004). Soufan
et al. (2003, 2007) reconstructed a series of developing mouse hearts from
sections of hearts in which in situ hybridization with markers for myocardium
was used to allow for the automatic selection of the myocardium. The 3D
reconstructions show details and provide insights that are not clear from
studying the series of sections themselves. Thus, the reconstructions are a
useful tool to increase our understanding of the development of the myocardium,
which is important because of the high incidence and seriousness of congenital
heart diseases.
Although 3D reconstructions can provide additional
insights, not all researchers decide to stain the
required complete series of sections. In many cases, only a limited number of
sections are stained for specific proteins or mRNAs. In fact, in some recent research
projects high-throughput methods are employed to generate vast amounts of such
individual sections: Carson et al. (2002), Raymond et al. (2002), Visel et al. (2004). Moreover, these sections are not
always of known orientation and possibly not exactly timed. Interpretation of
such sections is then hampered by lack of information on the anatomical and
developmental context.
The automatic fitting of such individual sections
into a reference series of 3D reconstructions can be used as a tool for data collecting
for a reference atlas. Being able to display gene expression in its 3D context
would constitute a huge improvement over (typical) gene expression databases
where the location of the gene expression is only described in general terms.
Project
description
The
present version of the automatic fitting program works on images of sections
which are stained with markers for myocardium, which enables the automatic
selection of the myocardium using a relatively simple thresholding
procedure. The main goal of the proposed project is to make it possible to select
the heart in an image of a section which is not specifically stained for heart
tissue. The goal of the Amsterdam Mouse Heart Atlas Project is to automatically
fit arbitrary 2D sections into the 3D reconstructions. It is therefore
necessary that the fitting algorithm can also be applied to sections where the
myocardium is not, or only partially, labeled.
After
a literature study, a few promising segmentation methods should be selected and
tested (using available implementations) for their suitability for the task of
segmenting the heart on images of typical sections without the myocardium
markers. The best suited method should then be refined for actual
implementation in the prototype of the fitting application.
Suggestions for further reading:
See
Ruijter et al. (2004) for information on techniques for three-dimensional
visualization of gene expression patterns and Christiansen et al. (2006) for a
description of an existing gene expression database developed with similar
intentions as those underlying the Amsterdam Mouse Heart Atlas. See http://en.wikipedia.org/wiki/Segmentation_(image_processing)
for a brief overview of segmentation methods.
References
Carson
et al. (2002) J. Carson, C. Thaller and G. Eichele, A transcriptome atlas of
the mouse brain at cellular resolution, Current Opinion in Neurobiology 12
(2002) 562-565.
Christiansen
(2006) Jeffrey H. Christiansen, Yiya Yang, Shanmugasundaram Venkataraman,
Lorna Richardson, Peter Stevenson, Nicholas Burton, Richard A. Baldock and Duncan R. Davidson; EMAGE: a spatial database
of gene expression patterns during mouse embryo development, Nucleic Acids
Research, 2006, Vol. 34, Database issue D637–D641
Raymond
et al. (1989) A. Raymond, V. Marigo, M. Yaylaoglu, A. Leoni, C. Ucla, N. Scamuffa, C. Cacciopoll, E. Dermitzakis, R.
Lyle, S. Banfi, G. Eichele,
S. Antonarakis, and A. Ballabio,
Human chromosome 21 gene expression atlas in the moude,
Nature 420 (2002) 582-586.
Ruijter et al. (2004) Jan M. Ruijter, Alexandre T. Soufan, Jaco Hagoort,
and Antoon F.M. Moorman, Molecular imaging of the embryonic heart: Fables and facts on 3D imaging
of gene expression patterns, Birth Defects Research (Part C) 72:224–240
(2004)
Soufan et al. (2003) A.T. Soufan, J.
Ruijter, M. van den Hoff, P. de Boer, J. Hagoort, A.
Moorman, 3D reconstruction of gene expression patterns during cardiac
development, Physiological Genomics 13 (2003) 187-195.
Soufan
et al. (2007) A.T. Soufan, G. van den Berg, P. D. Moerland, M. M. G. Massink, M. J.
B. van den Hoff, A. F. M. Moorman, J. M. Ruijter, Three-dimensional measurement
and visualization of morphogenesis applied to cardiac embryology, Journal of
Microscopy, Vol. 225, 2007 March 3, pp. 269–274
Visel
et al. (2004) Visel A, Thaller
C, Eichele G, GenePaint.org:
an atlas of gene expression patterns in the mouse embryo. Nucleic Acids Res.
2004 Jan 1;32(Database issue):D552-6.
Contact:
Frans Voorbraak (Medical
Informatics, f.p.voorbraak@amc.uva.nl)