Download Chapter 7 Digital Radiographic Image Processing and

Digital Radiographic Image Processing
and Manipulation
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
1
Objectives









Describe formation of an image histogram.
Discuss automatic rescaling.
Compare image latitude in digital imaging with
film/screen radiography.
List the functions of contrast enhancement
parameters.
State the Nyquist theorem.
Describe the effects of improper algorithm
application.
Explain modulation transfer function.
Discuss the purpose and function of image
manipulation factors.
Describe the major factors in image management.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
2
Key Terms












Archive query
Automatic rescaling
Contrast manipulation
Edge enhancement
High-pass filtering
Histogram
Image annotation
Image orientation
Image sampling
Image stitching
Latitude
Look-up table











Low-pass filtering
Magnification
Manual send
Modulation transfer
function
Nyquist theorem
Patient demographics
Shutter
Smoothing
Spatial frequency
resolution
Window level
Window width
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
3
Digital Radiographic Image Processing
and Manipulation

In cassette-based and cassetteless systems,
once the x-ray photons have been converted
into electrical signals, these signals are
available for processing and manipulation.

Pre-processing
• Takes place in the computer where the algorithms
determine the image histogram

Postprocessing
• Done by the technologist through various user functions
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
4
CR Reader Functions

The computed radiography (CR) imaging plate
records a wide range of x-ray exposures.



Data recognition program searches for anatomy
recorded on the imaging plate as follows:




If the entire range of exposures were used in image
formation, values at the extremely high and low ends of the
exposure range would result in a low-density resolution.
To avoid this, exposure data recognition processes only the
optimal density exposure range.
Finding collimation edges
Eliminating scatter outside the collimation
Failure of the system to find the collimation edges
can result in incorrect data collection.
Images may be too bright or too dark.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
5
CR Reader Functions

Histogram


A graphical representation of exposures values collected
from the imaging plate.
Because information within the collimated area is the
signal used for image data, the information is also the
source for a vendor-specific exposure data indicator.
Histogram
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
6
CR Image Sampling




Plate is scanned.
Image location and orientation is determined.
Size of the signal is determined.
Value is placed on each pixel.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
7
CR Image Sampling


Histogram is generated that allows system to find useful
signal by locating the minimum (S1) and maximum (S2)
signal within the anatomic regions of interest in the
image.
Histogram identifies all densities on the imaging plate in
the form of a graph:



X-axis is related to amount of exposure.
Y-axis displays the number of pixels for each exposure.
Graphic representation appears as a series of peaks and valleys
and has a pattern that varies for each body part.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
8
CR Image Sampling



Low energy (kilovoltage peak) gives a wider
histogram.
High energy (kilovoltage peak) gives a narrow
histogram.
Histogram shows the distribution of pixel values for
any given exposure.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
9
CR Image Sampling

For example:



Pixels have a value of 1, 2, 3, and 4 for a specific exposure.
Histogram shows the frequency of each of those values and
actual number of values.
Histogram sets the minimum (S1) and maximum (S2)
“useful” pixel values.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
10
Histogram Analysis



Analysis is complex.
Shape of the histogram stays fairly constant for each
part exposed (anatomy specific).
For example:

Shape of histogram for a chest radiograph on a large adult
patient looks different from a knee histogram generated from
a pediatric knee exam.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
11
Histogram Analysis


It is important to choose the correct anatomic region
on the menu before exposing the patient.
Raw data used to form the histogram are compared
with a “normal” histogram of the same body part by
the computer.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
12
The Nyquist Theorem

Theorem states that when sampling a signal, the
sampling frequency must be greater than twice the
bandwidth of the input signal so that the
reconstruction of the original image will be nearly
perfect.

At least twice the number of pixels needed to form the image
must be sampled.

If too few pixels are sampled, the result is a lack of
resolution.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
13
The Nyquist Theorem

The number of conversions in CR—electron to light, light to
digital information, digital to analog signal—results in loss of
detail.

Some light is lost during the light-to-digital conversion because of
the spreading out of light photons.
 Because there is a small distance between the phosphor plate
surface and the photosensitive diode of the photomultiplier, some
light spreads out there as well, resulting in loss of information.

The longer the electrons are stored, the more energy they lose.

When laser stimulates electrons, some lower-energy electrons
escape the active layer.
 If enough energy was lost, some lower-energy electrons are not
stimulated enough to escape and information is lost.
 All manufacturers suggest that imaging plates be read as soon as
possible to avoid this loss.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
14
DR and The Nyquist Theorem

Indirect and direct radiography lose less signal to
light spread than conventional radiography.


The Nyquist theorem is still applied to ensure that sufficient
signal is sampled.
Because sample is preprocessed by the computer
immediately, signal loss is minimized but still occurs.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
15
Aliasing

Occurs in DR when:



Spatial frequency is greater than the
Nyquist frequency.
Sampling occurs less than twice per cycle.
• Information is lost.
• Fluctuating signal is produced.
Wraparound image is produced.
• Image appears as two superimposed
images slightly out of alignment.


Aliasing results in a moiré effect.
Aliasing can be problematic because of
the same effect occurring with grid errors.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
16
Automatic Rescaling


Exposure is greater than or less than what is needed
to produce an image.
Automatic rescaling occurs in an effort to display the
pixels for the area of interest.


Problems occur with rescaling:
•
•


Images are produced that have uniform brightness and
contrast regardless of the amount of exposure.
When too little exposure is used.
When too much exposure is used, it will result in a loss of
contrast and loss of distinct edges because of detector
saturation.
Rescaling is no substitute for appropriate technical
factors.
Danger exists of using higher than necessary mAs
values to avoid quantum mottle.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
17
Look-Up Table

Look-up table (LUT)



A table of the luminance values derived during image
acquisition.
Used as a cross-reference to transform the raw information.
Has a mapping function:
• All pixels are changed to a new gray value.
• Image will have appropriate appearance in brightness and
contrast.

Provided for every anatomic part.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
18
Look-Up Table

LUT can be graphed as follows:

Plotting the original values ranging from 0 to 255 on the horizontal
axis
 Plotting new values, also ranging from 0 to 255 on the vertical axis


Contrast can be increased or decreased by changing the slope
of this graph.
Brightness (density) can be increased or decreased by moving
the line up or down the y-axis.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
19
Latitude

Exposure Latitude


Refers to the range of exposures that can be used and still
result in the capture of a diagnostic quality image.
The greater exposure latitude is due to the higher dynamic
range of the receptors.
• Dynamic range refers to the image receptors ability to respond
to the exposure.



In CR, if exposure is more than 50% below ideal exposure,
quantum mottle results.
If exposure is more than 200% above ideal exposure,
contrast loss results.
Biggest difference between digital and film/screen
radiography lies in the ability to manipulate the
digitized pixel values, which results in what seems
like greater exposure latitude.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
20
Enhanced Visualization Image Processing

Carestream

Enhanced Visualization Image Processing (EVP)
• Increases latitude while preserving contrast
• Decreases windowing and leveling
• Virtually eliminates detail loss in dense tissues
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
21
Modulation Transfer Function

Modulation transfer function (MTF)

The ability of a system to record available spatial
frequencies.
• Sum of the components in a recording system cannot be
greater than the system as a whole.


MTF is a way to quantify the contribution of
each system component to the overall
efficiency of the entire system—for example,
ratio of the image to the object.
A perfect system would have an MTF of 1, or
100%.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
22
Modulation Transfer Function

Digital detectors





X-ray photon energy
excites a phosphor.
Phosphor produces light.
Spreading out of the light
will always occur.
Light spread reduces
system efficiency.
The more light spread,
the lower the MTF.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
23
Quality Control Workstation Functions




Image processing parameters
Contrast manipulation
Spatial frequency resolution
Spatial frequency filtering
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
24
Image Processing Parameters

Digital systems have greater dynamic range
than film/screen imaging.

Initial digital image appears linear when graphed
because all shades of gray are visible.
 Digitalization gives the image a wide latitude.
 If all shades were left in the image, contrast would
be too low.
 To avoid this, digital systems make use of various
contrast-enhancement parameters.
• Names differ by vendor; Agfa uses MUSICA, Fuji uses
Gradation, and Carestream uses Tonescaling.
• Purpose and effects are basically the same.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
25
Contrast Manipulation

Contrast-enhancement parameters convert
the digital input data to an image with
appropriate brightness and contrast.



Image contrast is controlled by using a parameter
that changes the steepness of the exposure
gradient.
Brightness can be varied at the toe and shoulder
of the curve, removing the extremely low and
extremely high density values using a different
parameter.
No amount of adjustment takes the place of
proper technical factor selection.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
26
Workstation Screen Showing Contrast
Manipulation Choices
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
27
Spatial Frequency Resolution

Spatial Frequency Processing


Controls are available for the following:





Controls detail or sharpness of the image
Structure to be enhanced
Degree of enhancement for each density to reduce image
graininess
Adjust how much edge enhancement is applied
If improper algorithms are applied, image formation is
affected.
It is possible to degrade image information if
algorithms are improperly applied.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
28
Spatial Frequency Filtering


Edge enhancement
Smoothing
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
29
Edge Enhancement

When the signal is obtained, averaging of the signal
occurs to shorten processing time and storage.

The more pixels involved in the averaging, the smoother the
image appears.
• Signal strength of one pixel is averaged with the strength of
adjacent pixels or neighborhood pixels.
• Edge enhancement occurs when fewer pixels in the
neighborhood are included in the signal average.
• The smaller the neighborhood, the greater the enhancement.

When frequencies of areas of interest are known, they can
be amplified and other frequencies can be suppressed.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
30
Edge Enhancement

Amplification, also known as high-pass filtering, results in
an increase of contrast and edge enhancement.
• Suppression of frequencies, also known as masking, can result
in loss of small details.
• This technique is useful for enhancing large structures such as
organs and soft tissues but can be noisy.
Without Enhancement
With Enhancement
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
31
Smoothing

Smoothing




Also known as low-pass filtering.
Results from averaging of the frequency of each pixel with
surrounding pixel values to remove high-frequency noise.
Result is a reduction of noise and contrast.
Low-pass filtering is useful for viewing small structures such
as fine bone tissues.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
32
Basic Functions of the
Processing System


Image manipulation
Image management
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
33
Image Manipulation






Window level and window width
Background removal or shutter
Image orientation
Image stitching
Image annotation
Magnification
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
34
Window Level and Window Width

Window Level


Window Width



Controls how light or dark the image is.
Controls the ratio of black to white, or contrast.
User is able to manipulate quickly through use of the
mouse.
Movement of the mouse in one direction (vertical or
horizontal) controls brightness, and the other
direction controls contrast.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
35
Background Removal or Shutter


Unexposed borders around the collimation
edges allow excess light to enter the eye.
Veil glare




Causes oversensitization of a chemical within the
eye called rhodopsin.
This results in temporary white light blindness.
Eye recovers quickly enough so that viewer
recognizes only that the light is very bright.
Glare is a great distraction that interferes with
image reception by the eye.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
36
Background Removal or Shutter

In CR, automatic shuttering is used to blacken out the white collimation
borders.

Shuttering is a viewing technique only.
 Shuttering should never be used to mask poor collimation practices.
 Removal of the white unexposed borders results in an overall smaller
number of pixels.
 This reduces the amount of information to be stored.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
37
Image Orientation

Vendors mark the cassettes in different ways



Cassette must be oriented so that the image is
processed to display as expected.
• Fuji uses a tape-type orientation marker.
• Carestream uses a sticker.
Reader must be informed of the orientation of the
anatomy with respect to the reader.
Technologist will have the ability to re-orient
the image while preparing the image at the
quality control workstation.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
38
Image Stitching

Stitching is used for anatomy or areas of interest too
large to fit on one cassette.





Multiple images can be “stitched” together.
Sometimes, special cassette holders are used and
positioned vertically, corresponding to foot to hip or entire
spine radiography.
Images are processed in computer programs that nearly
seamlessly join the anatomy.
Computer displays one single image.
Process eliminates the need for large (36-inch) cassettes
previously used in film/screen radiography.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
39
Image Stitching
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
40
Image Annotation


Information other than standard identification must be
added to the image.
In screen/film radiography, additional information is
marked by the following:



Time and date stickers
Grease pencils
Permanent markers
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
41
Image Annotation

Annotation function





Allows selection of preset terms
and/or manual text input.
Can be useful when such additional
information is necessary.
Overlay the image as bitmap
images.
May not transfer to picture archival
and communication system (PACS).
Input of annotation for
identification of the patient’s left
or right side should never be
used as a substitute for
technologist’s anatomy markers.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
42
Magnification

Two basic types of magnification techniques are
standard with digital systems:

One type functions as a magnifying glass:
• A box is placed over a small segment of anatomy on the main
image.
• Box shows a magnified version of the underlying anatomy.
• The size of the magnified area and the amount of magnification
can be made larger or smaller.

Other technique is “zoom.”
• Zoom allows magnification of the entire image.
• Image can be enlarged enough that only parts of it are visible
on the screen.
• Those parts can be seen through mouse navigation.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
43
Magnification
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
44
Image Management



Patient demographics input
Manual send
Archive query
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
45
Patient Demographics Input

Proper identification of the patient is even more
critical.


Retrieval can be nearly impossible if image is not properly
and accurately identified.
Demographic information about the patient includes
the following:







Name
Health care facility
Patient identification number
Date of birth
Exam date
Other pertinent information
Input or linked via barcode label scans, before the start of
the exam and before the processing phase
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
46
Patient Demographics Input


Problems occur if the patient name is entered
differently from visit to visit or exam to exam.
For example:





Patient’s name is Jane A. Doe and is entered that way.
Name must be entered that way for every other exam.
If name is entered as Jane Doe, then system will save it as a
different patient.
Merging of files can be difficult.
Problems:




Several versions of the name are given.
Suppose the patient gives a middle name on one visit but
has multiple exams under his or her first name.
Retrieval of previous files will be difficult.
The right images must be placed in the correct data files.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
47
Manual Send


Because the quality control workstation is networked
to the PACS, it also has the capability to send images
to local network workstations.
The manual send function allows the quality control
technologist to select one or more local computers to
receive images.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
48
Archive Query


PACS archive can be queried for historical images.
Function allows retrieval of images from the PAC
system based on the following:






Date of exam
Patient name or number
Exam number
Pathologic condition
Anatomic area
Example:



Technologist could query PACS to retrieve all chest
radiographs for a particular date or range of dates.
Technologist could query retrieval of all of a patient’s
images.
Multiple combinations of query fields are possible:
• Can generate general retrieval
• Specific recovery of images
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
49
Summary





Recognition of exposure data involves placing all exposures detected in
a historgram. The histogram then undergoes an analysis.
The plate is scanned, and the image location is determined. A value is
place on each pixel, and the histogram is generated displaying the
minimum and maximum diagnostic signal.
The histogram is anatomic region specific and remains fairly constant
from patient to patient.
Automatic rescaling allows pixel display for the area of interest,
regardless of the amount of exposure.
There is no substitute for proper kilovoltage peak and milliamperesecond settings. Images cannot be created from nothing; that is,
insufficient photons, insufficient penetration, or overpenetration will
result in loss of diagnostic information that cannot be manufactured by
manipulation of the image parameters.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
50
Summary







Exposure latitude is slightly greater with digital imaging than that of
film/screen imaging because of the wide range of exposures recorded
with digital systems.
Contrast-enhancement parameters allow enhancement of the image by
controlling the steepness of the exposure gradient, density variance, and
contrast amount.
Spatial frequency resolution is controlled by focal spot, object image
distance, and computer algorithms.
The Nyquist theorem is applied to digital images to ensure that sufficient
signal sampling occurs so that maximum resolution is achieved.
MTF refers to the contribution of all system components to total
resolution. The closer the MTF value is to 1, the better the resolution.
Edge enhancement is accomplished by limiting the number of pixels in a
neighborhood of the matrix. Known area of interest frequencies can be
amplified or high-pass filtered to increase contrast and edge
enhancement.
Suppression of frequencies of lesser importance, known as masking,
can cause small detail loss.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
51
Summary






Low-pass filtering or smoothing is the result of pixel averaging to remove
high-frequency noise. Contrast and noise are decreased, allowing small
structures to be seen.
Window and level parameters control pixel brightness and contrast.
Shuttering is a process that removes or replaces the background in
order to block distracting light surrounding a digital image. This does not
take the place of proper collimation and can be removed to show proper
collimation.
Digital imaging cassettes are marked for orientation to the top and right
sides. This ensures that images are displayed correctly.
Image stitching is a computer program process that allows multiple
images to be joined when the anatomy is too large for one exposure.
The result is a nearly seamless, single image.
Magnification techniques are available with digital systems that allow
small area enlargement or whole image enlargement.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
52
Summary



Proper patient demographic input is the responsibility of the
technologist performing the exam. Any alterations of patient
demographics should be avoided unless absolute
identification is possible.
The manual send function allows images to be sent to one
or more networked computers.
Historical study of patient exams can be accomplished
through the archive query function. Retrieval of
radiographic studies can be specific as to patient name,
date, and exam or broad such as date ranges and
combinations of anatomic areas.
Elsevier items and derived items © 2008 by Mosby, Inc., an affiliate of Elsevier Inc.
53