|
Visual Numerics in Medicine
Bristol-Myers Squibb Uses PV-WAVE to Help Develop Safer, Stronger Medicines
Designing effective new drugs is a difficult, complicated and
expensive task. Identifying compounds that are active against
various diseases and then understanding the compounds' mechanism
of activity and how they cure diseases are enormous undertakings
that require years of research. Because of the incredible complexity
and delicate internal chemical balances in the body, it takes
trained experts and intense, careful study to ensure the safety
of these medicines.
Over the last decade a powerful imaging technique called Magnetic
Resonance Imaging (MRI) has emerged as an important medical diagnostic
tool, especially for tissues that are otherwise impossible to
study, such as those deep within the brain. MRI exploits the fact
that different tissues behave differently in strong magnetic fields.
With computerized imaging technology, researchers can observe
this behavior in the body's tissues under different conditions
without surgery or lengthy hospital stays.
Dr. Ron Behling, a researcher at Bristol-Myers Squibb's Pharmaceutical
Research Institute in New Jersey, is using Visual Numerics' PV-WAVE
visual data analysis software and MRI to study problems in drug
discovery and design. Behling uses MRI to measure the effects
of drug activity and to understand how drugs work within the body.
One disadvantage of MRI is that a large amount of data is created.
One experiment can easily
produce more than four megabytes of data, and there are often
multiple experiments in a study. To interpret these images, Behling
needs powerful analysis and imaging techniques that can handle
this volume of data and still remain interactive.
In Behling's work, MRI data is sent to his Silicon Graphics
workstation where it is read into PV-WAVE. According to Behling,
"Although there are no standard data formats in MRI, it is trivial
for me to read my data into PV-WAVE since the command language
has the flexibility to read virtually any data." In a few commands,
the image is quickly displayed. A quick visual inspection determines
if the data are suitable for further study. In further analysis,
Behling uses many of PV-WAVE's graphics and analysis features.
For example, color tables are interactively manipulated to bring
out hidden details in the images. Regions of interest are selected
for more detailed statistical analysis. Using various PV-WAVE
analysis functions, like FFTs and curve fits, custom procedures
are developed to assess diffusion rate constants in the tissues.
Contour plots provide a closer look at relative scales and intensities
in the images. And finally, binary files and hardcopy output to
TIFF and PostScript formats help to share findings with colleagues
and professional publications.
Flexibility is important in Behling's work because an analysis
or visualization technique appropriate for one type of data may
be completely unsuitable for another. PV-WAVE, with its highly
interactive command-based interface and its powerful macro capability,
provides the flexibility to easily modify existing analysis methods
and add new algorithms. This is absolutely vital for successful
data analysis. Reports Dr. Behling, "While many packages provide
imaging capability, the flexibility to analyze the images in any
way is a special strength of PV-WAVE. I process the data as I
wish, visualize them quickly, search for interesting regions,
analyze those regions and do further analysis based on those results.
Because of the ease of visualization and analysis, I can analyze
data quickly and do analyses that would otherwise be quite difficult
to perform."
With MRI and PV-WAVE, Behling and his colleagues have the tools
they need to study the behavior of drugs and thereby design medicines
that are safer and more effective for future patients.
IMSL Libraries Used in the Study of HIV and AIDS at the Centers
for Disease Control and Prevention
"I use
the IMSL Libraries to reduce my program development time,
so that I can concentrate on those aspects of a problem
that are novel and require statistical or AIDS-related
expertise."
|
At the Centers for Disease Control and Prevention
(CDC) in Atlanta, Georgia, mathematical statistician Dr. Glen
Satten is using the IMSL FORTRAN Numerical Libraries to model
the natural history of the human immunodeficiency virus type
1 (HIV-1). Satten is associated with several projects at the
CDC's Division of HIV/AIDS, where he is working to elucidate
the course of this complicated disease.
HIV is an epidemic with varying methods of transmission
accounting for its spread worldwide.In Africa and Asia, which
accounts for 60% of the world's population, heterosexual contact
is the primary source of HIV transmission.
"Homosexual transmission and shared needles among
injecting drug users are still the primary sources of transmission
in North America, but the proportion of new AIDS cases accounted
for by heterosexual transmission is increasing steadily," Satten
said. A unique study group of male military inductees in Thailand
provided an opportunity for Satten and his colleagues to develop
models of the probability of female-to-male HIV transmission
per sexual contact.
The analysis was possible because the situation in Thailand allowed
for the assumption that HIV-1 men
in the study had been infected through sex with female prostitutes
since 1988. "The Thai study group is unique because we can determine
how and when the HIV was most likely transmitted and then chart
the infection from its onset and continue as it progresses through
stages," said Satten. "Identifying the stages of HIV development
will assist in treatment of the infection."
Although prostitution in Thailand is not legal,
it is socially accepted. The military inductees provided information
on their frequency of contacts with prostitutes, which would
be difficult to obtain in other settings. In addition, the Thai
Ministry of Public Health has been monitoring the increase of
HIV among female prostitutes since mid-1989, when the prevalence
was 3.2 percent. By mid-1992, the prevalence had increased to
23 percent.
"The Thai study is unique because we know the
mode of transmission can determine the number of contacts in
which transmission could have occurred, but most important because
these contacts occurred with people who were likely to be recently
infected. Widespread transmission of HIV in Thailand was a fairly
recent phenomenon at the time the study was conducted. Our study
supports the hypothesis that individuals who are recently infected
are themselves more infectious," said Satten.
A FORTRAN programmer, Satten first used the IMSL
Libraries during postdoctoral work in biostatistics at the University
of North Carolina at Chapel Hill in 1988. The models he develops
at the CDC rely primarily on the optimization routines in the
IMSL Libraries to estimate the values of the parameters that
best describe the observed data, which are called maximum likelihood
estimates. The CDC relies primarily on an IBM mainframe and
PCs, although Satten has recently received authorization to
obtain a DEC Alpha 3000-600 workstation to run both the IMSL
FORTRAN Numerical Libraries and IMSL Exponent
Graphics software. Increased speed is the primary
reason for his move to a UNIX-based system.
"I use the IMSL Libraries to reduce my program development time,
so that I can concentrate on those aspects of a problem that
are novel and require statistical or AIDS-related expertise."
The IMSL Libraries were used to model the data,
and the findings were presented to researchers at the IX International
Conference on AIDS in June 1993 in Berlin. A mathematical model
was developed to estimate the probability of HIV-1 transmission
per sexual contact. Data included the inductee's age at first
sexual contact, frequency of sex with female prostitutes, province
of origin and province-specific HIV-1 seroprevalence among prostitutes.
PV-WAVE Used to Develop Radiotherapy Application at the Paul Scherrer
Institute in Switzerland
As part of the radiotherapy project at the Paul Scherrer Institute
(PSI) in Villigen, Switzerland, a comparative assessment of radiation
therapy techniques is being undertaken. The goal of this work
is to be able to present -- in easily understood but convincing
format -- the improved distributions to both target and normal
tissue that will result from the spot-scanning method of proton
radiotherapy.
The aim of the work is to:
- identify those clinical indications in which a definitive
difference can be shown between the efficacy of proton radiotherapy
and the routinely available photon therapy techniques,
- compare proton dose distributions with state-of-the-art photon
planning method and
- assess the relative merits of spot-scanning methods of proton
beam delivery with the more conventional passive-scattering
method of proton delivery.
For accurate comparison of the efficacy of photon and proton
treatments, it is imperative that identical data are available
to all the planning systems to be used in the comparisons. To
facilitate this, a set of common file formats has been developed
that allows data to be transferred easily between different planning
and computer systems. Formats have been defined to represent both
CT and dose distributions as 3D
distributions and to store Volumes of Interest (VOIs) in both
vertex and voxel representations. All file types consist entirely
of integer values and are read and written as integer stream files
with no system- or language-dependent record structure.
A model for the dual display and analysis of dose distributions
from different plans has been developed using Visual Numerics'
PV-WAVE data-visualization package. The main advantages of PV-WAVE
in this application are that it provides optimized array-manipulation
capabilities, allowing for fast processing of two- and three-dimensional
data sets, and a command language that can easily be transferred
between VMS and UNIX with little or no modification.
Together these features provide a high-level tool with which development
times can be significantly reduced. The combination of stream-data
file formats and the PV-WAVE shell provides a very portable system
that can be easily distributed over different hardware platforms
in the future.
The analysis software provides a tool for presenting separate
dose distributions on the same display for direct visual comparison.
Transaxial and coronal slices of the CT and dose data sets can
be interactively selected and displayed, with the position of
the orthogonal cuts being defined using cross hair cursors controlled
by a mouse. The dose distribution can be represented by either
overlaying the dose on the CT data using color wash and/or by
the use of isodose contours. Full interactive control of the CT
window, dose banding and display levels are all provided by the
use of widget-type control boxes.
The functionality of the system has been written using PV-WAVE
and its associated routines. PV-WAVE is ideally suited to such
applications, with the slicing of the 3D volume data into orthogonal
planes being achieved with simple, single-line commands. Similarly,
the overlaying of dose and VOI information onto the CT data can
be efficiently accomplished using look-up tables and the array
operations provided by the command language.
Two main analysis tools have been included in the comparison
software. The simplest allows users to plot profiles of both distributions
along any of the three axes defined by the cross hair cursors.
A more three-dimensional analysis of the plan is provided by the
Dose-Volume Histogram (DVH) facility. DVHs provide a method of
representing 3D dose distributions for selected VOIs in a convenient
2D format.
In brief, the DVH curve shows the volume (ordinate) of the selected
VOI irradiated to a level equal to or greater than any given dose
value (abscissa). Although spatial information is invariably lost,
the DVH provides a succinct summary of the dose in the selected
VOI. All DVH calculations and geometric manipulations are performed
using C functions called from the PV-WAVE shell using the LINKNLOAD
dynamic linking function. Although many of the calculations necessary
to produce the DVH plots could be achieved using PV-WAVE, the
use of external C routines gives additional flexibility.
Extracted from "A Tool for the Comparative Analysis of Radiation
Therapy Treatment Plans" by A. Lomax, G. Munkel, E.Pedroni and
H. Blattmann. Used with permission.
|
