Image Processing Reference
In-Depth Information
The Neuroterrain 3D mouse brain atlas [
5
] consists of segmented 3D structures rep-
resented as geometry and references a large collection of normalized 3D confocal
images.
Visual exploration and analysis.
3D microscopy data is often visualized using
Maximum Intensity Projection (MIP), which displays the maximum values along
viewing rays. Direct Volume Rendering (DVR) enables better perception of spatial
relationships, but has the disadvantage of added complexity, as an additional transfer
function is required. It can lead to problems with occlusions, particularly when mul-
tiple channels need to be visualized simultaneously. Maximum Intensity Difference
Accumulation (MIDA) [
9
] improves this situation by combining the simplicity of
MIP with additional spatial cues provided by DVR. Wan et al.[
105
] presented a tool
for the visualization of multi-channel data tailored to the needs of neurobiologists. As
acquired volumetric data is typically visualized together with segmented structures,
it is important to avoid occlusions as well as visual clutter. Kuß et al. [
56
] proposed
and evaluated several techniques to make spatial relationships more apparent.
However, to enable the exploration of large-scale collections of neuroanatomi-
cal data, massive sets of data must be presented in a way that enables them to be
browsed, analyzed, queried and compared. An overview of a processing and visual-
ization pipeline for large collections of 3D microscopy images is provided in a study
by de Leeuw et al. [
59
]. NeuARt II [
13
] provides a general 2D visual interface to 3D
neuroanatomical atlases including interactive visual browsing by stereotactic coor-
dinate navigation. Brain Explorer [
58
], an interface to the Allen Brain Atlas, allows
the visualization of mouse brain gene expression data in 3D. The CoCoMac-3D
Viewer developed by Bezgin et al. [
6
] implements a visual interface to two databases
containing morphology and connectivity data of the macaque brain for analysis and
quantification of connectivity data. An example of an interface to neuroanatomical
image collections and databases that features basic visual query functionalities is
the European Computerized Human Brain Database (ECHBD) [
31
]. It connects a
conventional database with an infrastructure for direct queries on raster data. Visual
queries on image contents can be performed by interactive definition of a volume of
interest in a 3D reference image. Press et al. [
75
] focused on the graphical search
within neuroanatomical atlases. Their system, called XANAT, allows for the study,
analysis, and storage of neuroanatomical connections. Users perform searches by
graphically defining a region of interest to display the connectivity information for
this region. Furthermore, their system also supports textual search using keywords
describing a particular region. Kuß et al. [
55
] proposed ontology-based high-level
queries in a database of bee brain images based on pre-generated 3D representa-
tions of atlas information. In the BrainGazer system [
9
] anatomical structures can be
visually mined based on their spatial location, neighborhood, and overlap with other
structures. By delineating staining patterns in a volume rendered image, for exam-
ple, the database can be searched for known anatomical objects in nearby locations
(see Fig.
21.5
). Lin et al. [
61
] presented an approach to explore neuronal structures
forming pathways and circuits using connectivity queries. In order to explore the
similarity and differences of a large population of anatomical variations, Joshi et
Search WWH ::
Custom Search