|
| |
|
image
quality and image perception |
|
 |
|
In
science and medicine, images are acquired for specific purposes
or tasks, and the ultimate measure of image quality is how well
the task is performed. In mammography, for example, the task is
detection of breast tumors, and in several forms of cardiac imaging
the task is estimation of the cardiac ejection fraction. The term
“observer” refers to the method by which the task
is performed; for tumor detection the observer can be either a
radiologist |
|
| or
a computer algorithm, and for estimation of ejection fraction
it is usually an image analysis program. In either case, the average
performance of the observer is a measure of the quality of the
imaging system for that particular task and observer, and for
the particular class of subjects being imaged. This performance
is influenced by many factors, including the object itself, any
contrast agent used, the nature of the image-acquisition system,
subsequent image reconstruction, processing and display; and the
capabilities of the observer, be it human or machine.
|
|
|
| |
The
Radiology Department at the University of Arizona
has a comprehensive research program that studies
all of these aspects of image quality. Specific
projects include:
Research
Grants
Research
Laboratories
The
Radiology Research Laboratory of the Department of
Radiology 1609 N. Warren Building 211 - The Radiology
Research Laboratory is located in close proximity
to the University Medical Center. It opened in Feb.
1988 and contains 10,000 net assignable square feet
of office and laboratory space.
a. The Psychophysics Laboratory - The facilities for
psychophysical research consist of two large research
laboratories with 413 square feet of floor space.
These laboratories are connected to the research laboratory
network and to the Internet. The rooms are designed
to maintain maximum control of all extraneous variables
(e.g., ambient room light) during experimental sessions.
A standard radiologic viewbox is available in the
psychophysics lab for viewing film images.
Key
to the psychophysics lab is a Series 6000SU Eye-Tracker
(Applied Science Laboratories, Bedford, MA) with a
magnetic head-tracking device. This system is an advanced
eye-tracker that accurately and unobtrusively measures
point of gaze and gaze duration information. The 4000SU
is ideal for use in a clinical radiology setting because
it provides for an unlimited field of view with free
head and body motion. The 4000SU is capable of measuring
point of gaze with a precision of less than one-half
a degree and accuracy of less than one degree. The
system comes with an extensive software library to
collect and analyze data.
A
number of software packages are available for use
in the psychophysics applications. The main statistical
software that Dr. Krupinski uses for analyzing general
data is StatView (SAS Institute). It is used for general
descriptive statistics as well as comparative tests
(ANOVA, Chi-Squared etc.). IN addition Dr. Krupinski
has the software to run a variety of Receiver Operating
Characteristic (ROC) analyses including: CLABROC,
CORROC2, INDROC, LABMRMC, LABROC1, LROC, MRMC, PLOTROC,
ROCFIT, ROCKIT, RSCORE and ROCPWRPC. Dr. Krupinski
maintains the website for the Medical Image Perception
Society (http://www.radiology.arizona.edu/krupinski/mips/rocprog.html)
that provides links to all of the programs from the
various sites that developed them. As new programs
become available they are added to the site.
We
also have customized software for image display and
analysis developed by Dr. Dallas (IMPROCRad). This
software takes any type of image (e.g., DICOM, .tif,
.img) and displays it on any monitor. There are various
navigation functions available (e.g., next image,
zoom, pan) as well as image analysis functions (e.g.,
image statistics such as noise calculations for a
given region of interest). When image processing techniques
are developed for a given study, these too are implemented
in the software for ready use during observer studies.
- 4 Pentium
based Personal Computers (PC's)
- PC based
- A/D converter
(Metrabyte DAS-HERS 16-bit I/0-board)
- Real time
video frame grabber
- X-Y-Z translation
stage with 2-micron position resolution
- Translation
in X (horizontal): 15 inch
- Translation
in Y (vertical): 20 inch
- Translation
in Z (perpendicular to CRT): 3 inch
- A high
performance CCD Camera, controlled by a desk-top
PC and based on a Fairchild-Loral CCD with the following
features:
- 2048
x 2048 pixels of 0.015 mm x 0.015 mm size
- cooled
to -25 0C.
- digitization
to 12 bits
- read-out
noise at 500 K-pixels/sec read-out is about 40
electrons
- A variety
of 35 mm type photographic Nikkon lenses as well
as microscope lenses to permit operating with a variety
of optical magnifications to achieve resolution on
the CRT to well below 0.020 mm
- A high
performance, "portable" CCD Camera,
controlled by a laptop computer and based on a Kodak
CCD with the following features:
- 1317 H
x 1035 V pixels of 0.0068 mm x 0.0068 mm size
- cooled
to -25 0C.
- digitization
to 14 bits
- read-out
noise at 500 K-pixels/sec read-out is about 40
electrons
- A variety
of 35 mm type photographic Nikkon lenses as well
as microscope lenses to permit operating with a variety
of optical magnifications
to achieve resolution on the CRT to well below 0.020
mm
- A mobile
cart with leveling feet, carrying a medium grade
X-Y-Z positioner to position the "portable" CCD
camera in front of most CRT displays used in the clinical
environment.
- Image analysis
software based on the IDL development package is
available on the PCs for:
- data
correction
- data analysis
- Fourier
Transforms (for noise power spectra and MTF measurements;
evaluation of modulation;
Contact
|
|
|
