Healthcare Data Visualization: Geospatial and Temporal Integration

Shenhui Jiang

, Shiaofen Fang

, Sam Bloomquist

, Jeremy Keiper

, Mathew Palakal

1,2

, Yuni Xia

and Shaun Grannis

Departmet of Computer & Information Science, Indiana University Purdue University Indianapolis,

IN 46202, Indianapolis, U.S.A.

School of Informatics, Indiana University Purdue University Indianapolis, IN 46202, Indianapolis, U.S.A.

Regenstrief Institute, Indiana University School of Medicine, Indianapolis, U.S.A.

Keywords: Healthcare Data, Spatiotemporal Visualization, Geospatial Information Visualization, Data and Text

Mining, Web-based Visualization Systems.

Abstract: Healthcare data visualization is challenging due to the needs for integrating geospatial information,

temporal information, text information, and heterogenious health attributes within a common visual context.

We recently developed a web-based healthcare data visualization system, Health-Terrain, based on a

Notifiable Condition Detector (NCD) use case. In this paper, we will describe this system, with emphasis on

the visualization techniques developed specifically for healthcare data. Two new visualization techniques

will be described: (1) A spatial texture based visualization approach for multi-dimensional attributes and

time-series data; (2) A spiral theme plot technique for visualizing time-variant patient data.

1 INTRODUCTION

As electronic healthcare systems are being fully

integrated nationally, the effective visualization of

large and complex healthcare data becomes

increasingly desirable for timely decision making

(Grossman et al., 2011). The problem, however, is

very challenging for several reasons:

1) Health data is a data-rich, information-poor

domain. In Electronic Health Record (EHR)

systems, data are almost always heterogeneous,

unstructured, hierarchical, and longitudinal.

2) EHR systems are large. While it is possible to

visualize an EHR system in small scales with a

focused scope, high impact knowledge

discoveries may come from population-wide

visualization and knowledge mining.

3) Visualizing population-level health data often

involves presenting geospatial and time-series

data in a common visual context. This presents a

challenge in visual encoding of the information

space.

For heterogeneous and complex data, feature

extraction through data mining is critical. For

healthcare data, this feature space often consists of

healthcare terms (ontology) and their relationships.

Therefore, the effective integration of data

processing, data mining, and text mining is

necessary in healthcare data visualization. Although

healthcare data is very large, the visualization of

aggregated features, combined with patient level

visualization, can be very effective in revealing the

patterns and trends of population health. It is

therefore important to develop multiple visualization

tools to be integrated within a common visual

interface to allow users to visually explore the data

through an easily accessible platform such as a web

browser.

One of the unique challenges in healthcare data

visualization is how to visualize multi-attributes and

time-series data with associated geospatial

information. In our approach, we embed multiple

attributes and the time variable within a geospatial

representation to take advantage of the available

geographic space. This can be done by mapping

texture images onto the geospatial surfaces. The key

is then to properly represent the multi-attributes and

time-series information in a texture image by

constructing visually effective texture

representations. While visualizing aggregated data

for geospatial areas provides global trends and

patterns in a geospatial context, we are often

interested in visualizing individual patient records

214

Jiang, S., Fang, S., Bloomquist, S., Keiper, J., Palakal, M., Xia, Y. and Grannis, S.

Healthcare Data Visualization: Geospatial and Temporal Integration.

DOI: 10.5220/0005714002120219

In Proceedings of the 11th Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2016) - Volume 2: IVAPP, pages 214-221

ISBN: 978-989-758-175-5

and their development over time. To this end, we

also developed a spiral theme plot technique for

visualizing time-variant patient records and

attributes. These new visualization techniques have

been implemented in a web-based healthcare data

visualization system called Health-Terrain, and

tested on real healthcare databases.

2 RELATED WORK

There are several existing works and visualization

systems that deal with the secondary use of

electronic health record data in a limited scope.

LifeLines (Plaisant et al., 1996) uses a traditional 2D

time line visualization technique to visualize specific

patient medical and health history. It emphasizes the

visualization of temporal ordering of events with

limited aggregation effect. An extension of LifeLine,

LifeLine2 (Wang et al., 2008), enables multiple

patient comparisons and aggregation for analysis,

but the visualization design limited its scalability. A

similar system, call TimeLine (Bui et al., 2007), re-

organizes and re-groups multiple EHR content types

in a layout of Y-axis to track multiple events along

the same time line. A set of visualization tools are

described for visualizing a patient’s electronic health

record to aid physicians’ diagnosis and decision-

making. The traditional matrix view and parallel

coordinates are the main techniques applied. The

VISITORS system (Klimov and Shahar, 2005; 2009)

combines a clinical knowledge base with

visualization to enable users to explore multiple

clinical records. It relies on domain ontologies to

define clinically meaningful higher abstractions

given raw, temporal data. CLEF (Hallett, 2008) is a

system enabling visual navigation through a

patient’s medical record using semantically and

temporally organized networks to represent events

throughout the patient’s medical history. CLEF also

supports limited text processing capabilities for

generating textual summaries. Interactive techniques

have also been developed for the navigation of space

and time dimensions (Bade et al., 2004; Maciejewski

et al., 2009). None of these existing systems is

capable of visualizing large-scale integrated EHR

datasets. A review paper on visualization tools for

infectious diseases is given in (Carroll et al., 2014).

A more general survey was given in (Chittaro, 2001)

about information visualization in Medicine.

Population-level healthcare data visualization

involves both geospatial information and time-

variant attributes. The geospatial visualization of

time-series data is challenging because it is difficult

to encode the time axis in a geospatial context.

Animation based techniques (e.g. Gemmell et al.,

2005) do not provide a good space-time overview.

Other techniques, such as color-coding of time (The

New York Times, 2013), connecting time-lines

(Google, 2013), and time-curves (Eccles et al.,

2007), often introduces visual clutter and occlusion,

which are infeasible for large scale datasets. A well-

known technique in geospatial time-series

visualization is Space-Time-Cube (Kraak et al.,

2003; 2007; 2004; Kwan, 2000; Andrienko et al.,

2003). It is a 3D representation of a combination of

time axis (Z-axis) and a 2D geographic map (X-Y

plane). Time-lines or time-curves are used to depict

data evolution over time. While time and spatial

information are integrated in a 3D visual

representation in a space-time-cube, the sense of

space-time embedding diminishes as the data moves

up in the time axis. Visual clutter will also be a

problem with large datasets. Similar 3D

representation of spatio-temporal data using 3D

icons have also been presented in (Tominski et al.,

2005). Many other techniques have been developed

for the visualization of time-series data without

explicit geospatial information such as time-series

plot (Tufte, 1983) and ThemeRiver (Havre et al.,

2000). Many variations of ThemeRiver styled

techniques have been applied in different time-series

visualization applications, in particular text

visualization (Cui et al., 2011). Spiral patterns have

also been used in visualizing time-series data

(Weber et al., 2001; Tominski et al., 2008) to

provide better identification of periodic structures in

the data.

Texture-based visualization techniques have

been widely used for vector field data, in particular,

flow visualization. Typically, a grayscale texture is

smeared in the direction of the vector field by a

convolution filter, for example, the Line Integral

Convolution (LIC), such that the texture reflects the

properties of the vector field (Cabral and Leedom,

1993; Stalling and Hege, 1995; Laramee et al.,

2004). Similar techniques have also been applied to

tensor fields (McGraw and Nadar, 2007; Auer et al.,

2012).

3 THE HEALTH-TERRAIN

SYSTEM

3.1 System Overview and Use Case

Our goal is to develop a prototype system, Health-

Healthcare Data Visualization: Geospatial and Temporal Integration

215

Terrain, to support visual exploration of large

healthcare data sets on a browser based interface.

The system integrates information visualization,

web-based user interaction, and text and data mining

techniques. A concept space approach is used to

unify data representation unified data representation

through data and text mining.

To test our visualization system we used a large

public health notifiable disease reporting system.

The Regenstrief Institute implemented and maintains

an unparalleled HIE-based, automated electronic lab

reporting (ELR) and case-notification system for

over ten years in the State of Indiana. The Notifiable

Condition Detector (NCD) System uses a standards-

based messaging and vocabulary infrastructure that

includes Health Level Seven (HL7) and Logical

Observation Identifiers Names and Codes (LOINC)

(Overhage, et al., 2008). The NCD receives real-

time HL7 version 2 clinical transactions daily,

including diagnoses, laboratory studies, and

transcriptions from hospitals, national labs and local

ancillary service organizations. The NCD dataset

contains 833,710 public health notifiable cases

spanning more than 10 years from among 439,547

unique patients. An additional dataset containing

325,791 unstructured clinical discharge summaries,

laboratory reports, and patient histories were

extracted. In order to comply with the patient

privacy policies and protocols of the institutes where

the datasets came from, the actual data visualized in

this paper has been altered or perturbed.

3.2 Concept Space

The “concept space” represents a uniform layer of

clinical observations and their associations, and

enables users to explore data using various

visualization and analysis methods. Concept terms

are derived from data mining and text-mining

processes applied to the use case datasets. Disease

concepts were extracted from the NCD dataset. Text

mining algorithms were then applied to additional

linked text dataset (unstructured clinical summaries)

to construct ontologies for different concept types,

including disease, symptom, mental behaviour, and

risky behaviour.

The concept space uses a controlled vocabulary

that can be pre-defined based on application needs,

and enhanced by data/text mining algorithms. These

terms and their relationships are represented in an

association map, as a space of extracted partial

knowledge. This association map is often the

starting point of a visual exploration process. Figure

1 shows an example of the association map of

diseases. Association map is a graph visualization of

the association relationships among the diseases and

other terms in the concept space. It can serve as a

platform supporting interactive selection of concepts

to dynamically visualize data using a variety of tools

in the visualization system. To draw an association

graph, a spring-embedder algorithm (Kobourov,

2012) is used to layout the graph nodes. Nodes

picked on the association map are then be visualized

with geospatial information, possibly with time

varying variables.

Figure 1: A Disease association map.

In text mining, we processed 325,791

unstructured clinical notes containing patient

discharge summaries, laboratory reports, and

medical histories. Advanced NLP was applied in the

form of named entity recognition (NER) for

extracting diseases and other terms, with the help of

the Unified Medical Language System (UMLS)

(Humphreys, et al., 1998). Stemming and concept

clustering algorithms (Osinski and Weiss, 2005)

were applied to normalize the lexical variants and

duplications of the terms. Term correlations were

computed using the tf-idf (term frequency – inverse

document frequency) vector space model to identify

the significantly co-occurring diseases. An

association-mining algorithm was applied to the

combined terms to generate an association graph

among all the concepts terms. The resulting concept

space, along with the processed NCD data, is

represented in a data model designed to support our

specific ontology.

4 SPATIAL TEXTURE BASED

APPROACH

Population-level healthcare data and information are

IVAPP 2016 - International Conference on Information Visualization Theory and Applications

216

often tightly coupled with geospatial regions. The

visualization of this type data requires the

integration of geo-visualization and multi-

dimensional and time-variant information

visualization. For this purpose, we propose a Spatial

Texture based approach. In this approach, we encode

multi-dimensional attributes or time-variant

attributes for a geospatial region into a texture

image, and then map the texture image to the surface

of the geospatial region to provide an integrated

visual representation. The key is the visual encoding

of multiple attributes or a time-variant attribute in a

texture image.

4.1 Noise Texture

We aim to represent multiple attributes for each

geospatial region using color coded texture patterns

so that the users can visually perceive the

representations of different attributes, not only

within one region, but also its overall geospatial

distributions across many regions in a geographic

area (e.g. a state).

We first construct noise patterns to create a

random variation in color intensity, similar to the

approach in (Gossett and Chen, 2004). Different

color hues will be used to represent different types

of attributes, for example the occurrences of

different diseases. A turbulence function (Perlin,

1985) will be used to generate the noise patterns of

different frequencies (sizes of the sub-regions of the

noise pattern). These multi-scale patterns may be

applied to different scales of geographic areas (e.g.

counties vs zip-codes). Since the noise pattern

involves the mixing and blending of different color

hues, we choose to use an RYB color model instead

of RGB model, as proposed in (Gossett and Chen,

2004), since RYB color model provides more

intuitive representation of the weights of different

colors after blending. Figure 2 shows two examples

of the texture mapped views of three diseases,

Diabetes. Hepatitis B, and Chlamydia, over the

Indiana state map. For example, more reddish areas

exhibits higher rate of Diabetes and bluish areas

show higher occurrence of Chlamydia.

4.2 Offset Contours

Offset contouring is designed to represent attribute

changes over time within a geographic region. It can

also be used to represent multiple attributes by

assigning each attribute to each contour. Similar to

the Noise Pattern texture, we first construct a texture

image using offset contour curves to form shape-

preserving sub-regions. We will then use varying

color shades or hues to fill the sequence of sub-

regions to represent the change of attribute values

over time, or to simply fill the sub-regions with

different color values to represent multiple

attributes.

The offset contours are generated by offsetting

the boundary curve toward the interior of the region,

creating multiple offset boundary curves (Figure 3).

There are several offset curve algorithms available

in curve/surface modeling. But since in our

application, the offset curves do not need to be very

accurate, we opt to use a simple image erosion

algorithm (Rosenfeld and Kak, 1982) directly on the

2D image of the map to generate the offset contours.

(a) (b)

Figure 2: Noise textures mapped over the Indiana State

map: (a) county based; (b) zip-code based.

Figure 3: Offset contours with different colors or different

shades of the same color.

In time-series data visualization, the time line

can be divided into multiple time intervals and

represented by the offset contours. Varying shades

of a color hue can be used to represent the attribute

changes (e.g. occurrence of a disease) over time.

This approach, however, has two limitations. First,

when the boundary shape of a region is highly

Healthcare Data Visualization: Geospatial and Temporal Integration

217

concave, the image erosion technique sometimes

does not generate clean offset contours. This usually

can be corrected using a geometric offset curve

algorithm such as the one in (Hoschek, 1988). A

second limitation of this approach is that it requires a

certain amount of spatial area to layout the contours

and color patterns. In public health data, however,

these attributes are typically defined on geographic

areas, which provides a perfect platform for texture

based visual encoding. Figure 4 shows a few

examples of the texture mapped views of offset

contours over the Indiana state map. Figure 4 (a-b)

show the time-series views of Influenza, from 2004

to 2012. The time interval is divided into 8

subintervals. Figure 4 (c-d) show three diseases,

Influenza, Typhoid Fever, and Hepatitis B.

(a) (b)

Figure 4: Texture mapped views of offset contours over

the Indiana state map: (a) County based time-series data;

(b) Zip-code based time-series data; (c) County based

multi-diseases data; (d) Zip-code based multi-diseases

data.

5 SPIRAL THEME PLOT

Spatial texture provides overviews of health care

data associated with geographic regions. It is

however often desirable for health administrators

and physicians to also see the details of individual

patients and theirs medical history (over time).

When this is done with a large population, the

collective view of patient medical histories often

exhibit identifiable patterns and trends that may not

be easily detected from the visualization of statistical

data over geographical regions.

We developed a new time-series visualization

method called Spiral Theme Plot by integrating

ThemeRiver (Havre et al., 2000) and spiral pattern

(Weber et al., 2001) to plot patients as points in

stacked spiral rings. Time is represented as a spiral

base curve. Diseases (or any other term) are

represented as stacked themes along a spiral base

curve. Patients are plotted within the regions of the

themes as points with proper visual attributes. One

significant attribute, for example “age”, will be

represented as radius. Other attributes of the

patients, such as race and gender, are represented as

color and shape of the dots. Spiral Theme Plot

allows multiple years of patients data be plotted

periodically such that seasonal patterns or abnormal

patterns for seasonal diseases can be easily detected.

For patients with multiple hospital visits at different

times for the same or different conditions, curves are

drawn to connect these multiple occurrences by the

same patient.

The base spiral curve is:



sin

  cos

where  is a monotonic continuous radius

function of angle θ. When  is a linear function









, the gap between the spirals is a

constant 2, which can be estimated based on the

maximum cumulative width of the themes (Fig. 5).

When plotting patient data within each theme,

the width of the theme at a particular angle is

determined by the total occurrence of the disease at

that particular time. The boundary curve of each

theme can then be interpolated by spline curves.

This interpolation is done by splitting the time axis

into a fixed number of segments. The maximum

width of each segment is used as an interpolation

point. This leads to a discrete set of interpolation

points from which the spline curve can be generated

as the boundary curve of the theme. When plotting a

point for each patient, the width of the theme needs

to be computed first in order to determine the proper

IVAPP 2016 - International Conference on Information Visualization Theory and Applications

218

radius of the point. Although this width information

can theoretically be computed from the spline

representations, we found that it is more efficient to

simply check the color values along the normal

direction of the spiral curve to estimate the width of

a theme at each angle.

Lines are drawn between points representing

multiple occurrences of the same patient. Such lines

sometimes can become very dense leading to a

cluttered image. We implemented an edge bundling

strategy to bundle these connecting lines for each

pre-defined time interval (Fig. 6). Figure 7 show a

periodical (seasonal) pattern of Flu over 4 years.

Figure 5: Spiral Theme Plot for Hepatitis A, B, and C over

four years.

Figure 6: Spiral Theme Plot with bundled links.

Figure 7: Seasonal pattern of Flu.

6 SYSTEM INTERFACE AND

EVALUATION

The system is implemented using Javascript in an

HTML5 canvas. The visualization algorithms are

implemented using HTML, CSS, SVG, and WebGL

technologies with a number of open-source

Javascript libraries.

The user interface uses multiple split windows so

that multiple types of visualizations can be applied

and compared for the same dataset. Fig. 8 show a

screen shot of three visualizations for a dataset

selected from an association map. Visualization

results can also be saved into a slider bar, with time

stamps, and be brought back later (Fig. 9). This

provides a flexible workspace for health

administrators or physicians to explore and compare

different scenarios for health policy planning,

decision making, resource management, etc.

Figure 8: A screen shot of a split window interface.

Healthcare Data Visualization: Geospatial and Temporal Integration

219

Figure 9: System interface with saved working windows.

To evaluate the system, we adopted the National

Institute of Standards and Technology (2007)

definition of usability for our participants. Using an

unstructured qualitative interview process, we

explored dimensions of effectiveness, efficiency,

and satisfaction. Due to the data privacy policy

provisions of the institutional review board research

process, we used obfuscated de-identified clinical

data for the usability assessment.

Prior to reviewing the interviewees were oriented

to a few detailed dimensions of the application: The

interviewees’ responses can be summarized as

follows:

• Users were pleased with the abilities to quickly

identify associations of different terms and form

subnetworks. Some felt that the visualization has

the potential to make them think about things

that they wouldn't otherwise, and that has value

to them.

• Some users felt that they may not use

visualization to identify disease outbreaks, but

would instead use this visualization after an

outbreak has been detected through other means

in order to explore the relationships and

characteristics of individuals within an outbreak

in order to identify potential risk factors and

target interventions.

• Users felt that this visualization system was very

complex and exhibited high information density,

which sometimes can obfuscated important

information. More in-line guidance or pop-up

descriptions (e.g., mouse-overs) would be

helpful.

• For geospatial data visualization, some suggested

adding a nonlinear scaling to highlight details in

lower prevalence regions, or presenting the data

as incident rates. Epidemiologist interviewees

requested extended functionality to visualize the

highest prevalence diseases in each county.

7 CONCLUSIONS

We present a health data visualization system which

emphasizes the integration of geospatial and

temporal information in healthcare data. We focus

on two new visualization methods we developed

specifically for public health data: Spatial Textures,

and Spiral Theme Plot. Spatial Texture approach is

effective because geospatial visualization

intrinsically provides additional screen space

(surface areas) that can be taken advantages of to

encode additional data and attributes. The Spiral

Theme Plot technique is a combination of several

information visualization methods including Theme

River, Spiral Plot and Scatter Plot. For public health

data with large patient databases, this particular

combination satisfies several key requirements for

visualizing time-variant patient records. With the

rich set of tools available to support web based user

interface, graphics, and data communications, we

also feel that it is as efficient to develop a web based

visualization system as in a traditional programming

environment.

In the future, we would like to continue refining

and expanding this visualization system by adding

new visualization tools and improving the existing

ones, in particular, the desired features and

improvements suggested by the evaluators. We

would also like to develop a configurable user and

data interface so that the system can be easily

configured for other types of use cases in public

health applications.

ACKNOWLEDGEMENTS

The project is supported by US Department of the

Army, award W81CWH-13-1-0020. We gratefully

acknowledge the contribution of Jennifer Williams;

and the Regenstrief Institute’s data engineers.

REFERENCES

Grossman, C., Powers B., McGinnis J. M., 2011. Digital

infrastructure for the learning health care system: the

foundation for continuous improvement in health and

health care. The National Academies Press.

IVAPP 2016 - International Conference on Information Visualization Theory and Applications

220

Plaisant, C., Milash, B., Rose, A., Widoff, S.,

Shneiderman, B., 1996. Lifeline: Visualizing Personal

Histories, CHI, pp. 221-227.

Wang, T. D., Plaisant, C., et al. 2008. Aligning Temporal

Data by Sentinel Events: Discovering Patterns in

Electronic Health Records, CHI’08, pp. 457-466.

Bui, A., Aberle, D. R., Kangarloo, H, 2007. Timeline:

Visualizing Integrated Patient Records. IEEE Trans.

Information Technology in Biomedicine 11(4):462-

473.

Klimov D., Shahar Y, 2005. A framework for intelligent

visualization of multiple time-oriented medical

records. AMIA Annual Symp Proc.; 405–409.

Klimov D., Shahar Y, Taieb-Maimon M, 2009. Intelligent

interactive visual exploration of temporal associations

among multiple time-oriented patient records. Methods

Inf Med. 48: 254–262.

Hallett, C., 2008. Multi-Modal Presentation of Medical

Histories. IUI’08: 13th International Conference on

Intelligent User Interfaces. pp. 80-89.

Bade, R., Schlechtweg, S. and Miksch, S., 2004.

Connecting time-oriented data and information to a

coherent interactive visualization. CHI’04. 105-112.

Maciejewski, R., Rudolph, S. and Hafen, R. 2009. A

Visual Analytics Approach to Understanding

Spatiotemporal Hotspots. IEEE Transactions on

Visualization and Computer Graphics, 16, 205-220.

Carroll LN et al. 2014. Visualization and analytics tools

for infectious disease epidemiology: A systematic

review. J Biomed Inform, 51:287-98.

Chittaro, L., 2001. Information Visualization and Its

Application to Medicine. Artificial Intelligence in

Medicine, 22(2), 81-88.

Gemmell, J., Aris A., Lueder R. 2005. Telling stories with

MyLifeBits. IEEE International Conference on

Multimedia and Expo, pp. 1536–1539.

The New York Times Company: Openpaths, Feb. 2013.

URL: https://openpaths.cc.

Google: Latitude, 2013. http://www.google.com/latitude/

Eccles, R., Kapler T., Harper, R., Eright, W. 2007. Stories

in GeoTime. In IEEE VAST, 19–26.

Kraak, M. 2003. The Space Time Cube Revisited from a

Geovisualization Perspective, Proc. 21st Int’l

Cartographic Conf., 1988-1996.

Kraak, M. J., and P. F. Madzudzo, 2007. Space time

visualization for epidemiological research.

Proceedings of the 23nd international cartographic

conference: Cartography for everyone and for you.

Kraak, M. J. and A. Kousoulakou, 2004. A visualization

environment for the space-time-cube. Developments in

spatial data handling. 11th International Symposium

on Spatial Data Handling. 189-200.

Kwan, M. P., 2000. Interactive geovisualization of activity

travel patterns using 3D geographical information

systems: a methodological exploration with a large

data set. Transportation Research C 8: 185-203.

Andrienko, N., G. L. Andrienko, et al., 2003. Visual data

exploration using space-time cube. 21st International

Cartographic Conference, Durban, South Africa.

Tominski, C., Schulze-Wollgast, P. and Schumann, H.,

2005. 3D information visualization for time dependent

data on maps. IEEE Information Visualization’05,

175-181.

Tufte, E. R., 1983. The Visual Display of Quantitative

Information. Graphics Press.

Havre, S., Richland, W. A., Hetzler, B., 2000.

ThemeRiver: visualizing theme changes over time.

IEEE Information Visualization, 115-123.

Cui, W., S. Liu, et. al., 2011. Textflow: Towards better

understanding of evolving topics in text. IEEE Trans.

on Visualization and Computer Graphics, 17 (12),

2412-2421.

Weber, M., Marc Alexa, Wolfgang Müller. 2001.

Visualizing Time-Series on Spirals. IEEE Information

Visualization, 7-13.

Tominski, C., Schumann, H., 2008. Enhanced Interactive

Spiral Display, Annual SIGRAD Conference Special

Theme: Interaction. 53-56.

Cabral, B., and L. C. Leedom. 1993. Imaging Vector

Fields Using Line Integral Convolution. In Poceedings

of ACM SIGGRAPH, 263–272.

Stalling, D. and H. Hege. 1995. Fast and Resolution

Independent Line Integral Convolution. In

Proceedings of ACM SIGGRAPH 95, 249–256.

Laramee, R. S., et al., 2004. The State of the Art in Flow

Visualization: Dense and Texture-Based Techniques.

Computer Graphics Forum, 3(2):203–221.

McGraw, T., M. Nadar. 2007. Fast Texture-Based Tensor

Field Visualization for DT-MRI. 4th IEEE

International Symposium on Biomedical Imaging:

Macro to Nano, 760-763.

Auer, C., Stripf, C., Kratz, A., Hotz, I., 2012. Glyph- and

Texture-based Visualization of Segmented Tensor

Fields. Proc. Int. Conf. on Information Visualization

Theory and Applications, 670-677.

Overhage J. M, Grannis S. J., McDonald C. J., 2008. A

comparison of the completeness and timeliness of

automated electronic laboratory reporting and

spontaneous reporting of notifiable conditions. Am J

Public Health. 98(2):344-50.

Stephen, G., Kobourov, 2012. Spring Embedders and

Force Directed Graph Drawing Algorithms. arXiv:

1201.3011.

Humphreys, B. L., D. A. Lindberg, H. M. Schoolman, G.

O. Barnett, 1998. The unified medical language

system: An informatics research collaboration J. Am.

Med. Inform. Assoc., 5 (1), 1–11.

Osinski, S., D Weiss, 2005. A concept-driven algorithm

for clustering search results - Intelligent Systems,

IEEE Intelligent Systems, 48-54.

Gossett, N., Baoquan Chen, 2004. Paint Inspired Color

Mixing and Compositing for Visualization. IEEE

Symposium on Information Visualization. 113-117.

Perlin, K., 1982. An image synthesizer. In Proceedings of

SIGGRAPH’85

, 287–296.

Rosenfeld, A. and A.C. Kak, 1982. Digital Picture

Processing. Academic Press, New York.

Hoschek, J., 1988. Spline Approximation of Offset

Curves. Computer Aided Geometric Design. 33–40.

Healthcare Data Visualization: Geospatial and Temporal Integration

221