Image Processing Reference
In-Depth Information
22.4.4 Challenges for Visualization
Evo-devo data presents several challenges for visualization, the largest of which is
validation of the complex preprocessing pipeline. Assuming, however, that the data is
perfectly preprocessed, the following visualization challenges remain for supporting
data analysis:
Large-scale volumetric data
: The recorded data is typically quite large (up to sev-
eral terabytes for a single specimen). Volume visualization needs to bridge different
levels of detail to provide an overview in conjunction with detailed information
about fine-scale structures.
Multimodal and time-series data:
Combining multimodal and temporal data is an
open-research problem in visualization.
Wealth of data:
Many datasets consist of terabytes of raw data with numerous
small features in long time series. Adequate abstraction mechanisms are therefore
mandatory to help the user browse and investigate the data.
Combination of InfoVis and SciVis necessary:
Due to the versatile nature of the
input data, techniques from both information and scientific visualization need to
be combined in interactive frameworks.
Sophisticated data analysis required:
Feature selection and extraction is a crucial
part of the visualization pipeline in evo-devo and requires knowledge in many
related subjects such as computer vision, machine learning, and statistics.
22.4.5 Case Study: Visualization in Developmental Biology
As previously stated, visualization in evo-devo requires expertise in different areas
of research. In this case study, we summarize the on-going work of our interdiscipli-
nary team that studies the biological processes that occur during embryonic develop-
ment, with the goal of creating a digital embryo model. The team consists of experts
from:
developmental biology
who research the embryonic development of fish and
flies;
microscopy research
who work towards better image acquisition techniques;
multidimensional image processing
who develop algorithms for the automatic seg-
mentation and analysis of 3D spatio-temporal images; and
scientific visualization
who concentrate on data validation, feature extraction, and interactive data visual-
ization and analysis. As the scientific questions related to embryonic development
are very diverse and often change and expand over time, we decided to develop a ver-
satile visualization framework (
Scifer
—
http://scifer.info
)
with additional dedicated
algorithms to address specific biological questions.
Scifer
, shown in Fig.
22.3
,isa
multi-window environment with linked views that combines interactive information
and scientific visualization algorithms.
The primary GUI consists of three windows. The
main window
(Fig.
22.3
a) is used
to control the workflow. In this window, the user selects from a list of preprocessing
algorithms, after which an interactive dialog box opens where additional parameters
Search WWH ::
Custom Search