[其它] HumMod Explorer: A Multi-Scale Time-Varying Human Modeling Navigator

彬彬

797 主题	1 听众	1万积分

资深设计师

Rank: 7 Rank: 7 Rank: 7

纳金币: 5568
精华: 0

电梯直达

楼主

发表于 2011-12-30 18:54:39 |只看该作者 |倒序浏览

1 Introduction

HumMod [Hester et al. 2011] models interactions between the cardiovascular,

respiratory, renal, neural, endocrine, skeletal muscle,

and metabolic physiologies. One of the difficulties which must be

overcome with HumMod is the fact that the number of parameters

is prohibitive in developing a clear and comprehensive view

of the interactions between parameters. Currently the HumMod

model contains more than 5000 variables. Therefore, good visualization

techniques are needed to discern non-obvious relationships

between variables. We present HumMod Navigator, a multiplescale

physiology data browser for exploring casual relationships of

time-varying human modeling data. The visualization makes use

of a circular layout and hierarchical relations to effectively visualize

interactions between model parameters in an attempt to obtain

both a local and comprehensive view of the physiological modeling

environment.

2 Parameter Interactions and Our Methods

In HumMod Navigator (Fig. 1, the first type of interaction

represented is the hierachical relationship. Every parameter

used in HumMod has an assigned “global” name. For

example, Co2Total includes the parameters or global names

of CO2.CO2Total.Inflow and CO2.CO2Total.Outflow.

Each global name is made up of three distinct components, namely

CO2, CO2Total, and Inflow, and CO2, CO2Total, and

Outflow. This represents the fact that within the folder describing

CO2, and the specific file describing CO2Total, there are local

elements corresponding to both Inflow and Outflow.

A second type of relationship is the functional relationships.

Each XML file for a parameter details the dependence of

its value on the value of other parameters. For example,

the XML file details the dependence of the value of

CO2.CO2Total.Inflow on the value of parameters corresponding

to the outflow of CO2 from other body organs, as well as

the dependence of the value of CO2.CO2Total.Outflow

on the values of Lung.LungCO2.Expired and

CO2.CO2Tools.LitersToMols. These relationships

constitute the functional relationships between different global

parameters.

To combine both the hierarchical and functional interactions, we

designed HumMod Navigator by employing the technique of

Holten [Holten 2006] which uses hierarchical information to induce

edge bundling. Cubic splines are used for the edges, and the possible

control points are determined by drawing the hierarchical trees

along circular rings moving inward. Possible control points along

the innermost ring (those points corresponding to level 0 names)

may be thought of as being connected by an inner cycle. The control

points for an edge between nodes i and j are chosen to be the

points corresponding to a shortest path between i and j along the

interior nodes of the circle.

e-mail: zgpan@cad.zju.edu.cn

&dagger;e-mail: rhester@umc.edu

&Dagger;e-mail: jian.chen@usm.edu

Figure 1: View of Hierarchical Tree around inner ring. The node

at the center of the figure corresponds to the level 0 name. From the

adjacent node on the left, the nodes clockwise around the cycle cor-

respond to the level 2 names given on the right. Edges from these

nodes to the root have been suppressed, except for the rightmost and

leftmost neighbor. The level 2 names can be projected downward

onto the inner circle to determine their associated global name pa-

rameters.

Each parameter is time varying and interpreting such data imposes

cognitive and perceptual load to physiologists. To address this issue,

we further show variations through a perceptually unformed

coloring space. A color space is said to be perceptually uniform if

the perceptual difference between any two colors in just noticeable

difference units is equal to the Euclidean distance between the two

colors in that color space. Figure 1 shows the result of the parameter

variances coloring mapped to the L*a*b color space at L=60 for

a half-an-hour simulation of a normal person.

Each node in the surrounding forest of hierarchical trees can be

selected. Upon selection, this highlights the adjacent nodes (and

corresponding edges) for all descendants of the selected node. At

the lowest level, this gives a local view of parameter interaction.

Selecting a node at a higher level in the hierarchical tree gives a

semi-local, semi-global view of parameter interaction. The second

important aspect of the diagrammatic view of the model was that

the overall structure of the model could be viewed at a glance.

References

HESTER, R., BROWN, A., HUSBAND, L., ILIESCU, R., PRUETT,

W., SUMMERS, R., AND COLEMAN, T. 2011. Hummod:

A modeling environment for the simulation of integrative human

physiology. Frontiers in Computational Physiology And

Medicine 2, 12.

HOLTEN, D. 2006. Hierarchical edge bundles: Visualization of

adjacency relations in hierarchical data. IEEE Transactions on

Visualization and Computer Graphics, 741–748.