COMPUTED AND

DIGITAL
RADIOGRAPHY
Carlos F. Romasanta Jr, RRT, MMHoA.
PROCESSIN
G OF
DIGITAL
IMAGE
PRE-PROCESSING
• Defined as all corrections that are made to the “raw” digital
image data is due to physical flaws in image acquisition that are
inherent to the x-ray beam, the elements and circuitry of the
particular image receptor system, or the physical elements and
circuitry of the processor.
• Preprocessing may also be termed acquisition processing.
EXPOSURE FIELD RECOGNITION
▸ Ability of a digital imaging
system to identify the borders of
a collimated x-ray exposure field,
so that data outside the field may
be excluded from histogram
analysis and exposure indicator
calculations.
▸ The first order of business for the
processing computer is to sort out how
many separate images are on the plate and
identify where each one begins and ends
so that they are not all processed together
as a single image.
▸ In computed radiography (CR), more than one
image can be recorded on a single PSP plate
(photostimulable phosphor plate).
▸ DR units allow only one exposure on each
detector “plate,” which is sent to the computer
before another exposure can be taken on the 5
FIELD UNIFORMITY CORRECTIONS
• Several flaws are found in the electronics of receptor systems as well
as the optical components of a CR reader (such as lenses and optic
fibers).
• These limitations are inherent in every image acquisition system, and
result in an uneven distribution of the “background” density of the
resulting images
• Adjustments for flaws include:
• inherent in projecting the x-ray signal to the detectors (such as
the anode heel effect)
• flaws in digital acquisition hardware
• pixel drop-out effects..
 CORRECTING DEAD
DEXEL AND MOTTLE
• In the second step for digital preprocessing,
the computer looks for “dead pixels,” where
individual detector elements (dexels) in the
image receptor might not report any data due
to electronic failure.
• To avoid blank spots showing up in the image
the computer uses noise reduction software
(i.e. Kernel).
• Kernel may be defined as a sub-matrix that is
passed over the larger image executing some
mathematical function on the pixels.
• The process of averaging the pixels
Correction for a dead or stuck pixel: A nine-cell kernel first sums
surrounding the malfunctioning component the pixel values for the surrounding eight pixels (top). These
values are then averaged to “fill” the dead or stuck pixel (bottom).
and fill this “dead” space in the image is
called interpolation.
 MOTTLE
• Mottle is a form of noise that manifests as a
grainy appearance to the image, consisting of
very small freckle-like blotches of dark and
light throughout the image
• Radiographers are particularly concerned with
two very common types of mottle, quantum
mottle and electronic mottle,
• Periodic mottle is best corrected using during
frequency processing,
• Quantum mottle it is better to use a kernel for
correction.

Quantum mottle, A, results from naturally occurring randomness

in the distribution of x-rays within the beam. Electronic mottle,
B, manifests on a television or LCD monitor screen from small
fluctuations in current that occur in all electronic devices
TYPES OF IMAGE NOISE
 ▸ Is actually a bar graph depicting the
pixel count throughout the image for
IMAGE HISTOGRAM each pixel value or density (or
brightness).

▸ Before rescaling of the image can take

place, the computer must build up a
histogram of all the image data, in
which a count is made of all pixels
sharing the same pixel value (density,
or brightness).

▸ There is no indication of the location of

these pixels in the image or what
anatomy they represent, just a simple
count for each “brightness” value..

10
IMAGE HISTOGRAM
Single Lobe
histogram

11
HISTOGRAM ANALYSIS

▸ The computer analyzes the histogram using processing algorithms and

compares it with a pre-established histogram specific to the anatomic part
being imaged.

▸ The computer software has stored histogram models, each having a shape
characteristic of the selected anatomic region and projection.

▸ These stored histogram models have values of interest (VOI), which

determine the range of the histogram data set that should be included in the
displayed image.
HISTOGRAM ANALYSIS

▸ The computer begins by effectively comparing

the actual histogram from the exposed image to
an expected histogram shape for that
procedure.

▸ If the general shape matches, analysis can

proceed without errors.

▸ The initial purpose of analyzing the histogram is

to eliminate extreme data that will skew the
rescaling of the image, making it come out too
dark or too light.

▸ “bin” actually represents a particular pixel

value, density or brightness
13
HISTOGRAM ANALYSIS
▸ Histogram analysis, the computer always
“scans” into the histogram from the left
and from the right, looking for data set
“landmarks.”

▸ One type of landmark is to set a

threshold to the pixel count that must be
exceeded to keep that data for analysis.

▸ Having eliminated both background

densities and noise from the data set, the
remaining data between points SMIN and
SMAX all represent true anatomical
structures, which should result in proper
rescaling of the image.

14
TYPES OF HISTOGRAM
ANALYSIS
▸ Type 1 - is designed to
analyze two-lobe histograms
with a “tail-spike”
representing background
densities.

▸ Type 2 analysis is designed

for a single-lobe histogram.

▸ Type 3 analysis operates on

the assumption that there
will be three lobes
Volume of Interest (VOI) – Portion of an image histogram
that contains data useful for digital processing and for
calculating an accurate exposure indicator.
15
AUTOMATIC RESCALING
▸ The computer rescales the
image based on the
comparison of the
histograms, which is
actually a process of
mapping the grayscale to
the VOI to present a
specific display of
brightness

▸ Problems occur with

rescaling:
▹ Too little exposure
results in quantum
mottle.
Automatic rescaling is employed during histogram analysis to maintain
consistent image brightness despite overexposure or underexposure of ▹
Too much exposure
the IR results in a loss of
contrast and loss of
distinct edges
because of detector
the permanent LUT (initial gradation) from the histogram analysis will be used for
saturation.
rescaling 16
LOOK-UP TABLE (LUT) Straight line
graph
demonstrating no
change in the
pixel values from
the original to the
▸ Provide the means to alter the processed image
brightness and grayscale of the
digital image using computer
algorithms.

▸ They are also sometimes used to

reverse or invert image grayscale.

▸ With the processing of each image,

a table is used rather than the
formula in order to save processing Graph demonstrating a
time in the computer. change in pixel values from
the original image to the
processed image.

17
LOOK-UP TABLE (LUT)

The original low-contrast chest

image is altered to display a
higher-contrast chest image. The
graph shows a change in the
pixel values from the original
image

18
POST-PROCESSING

▸ All manipulation and adjustments of the digital image

(whether by default settings in the processor or by an
operator) made after corrections have been made for data
acquisition.
▸ These operations are targeted at refinement of the image,
and although they may be performed as part of the default
processing of the image, are also somewhat subject to
personal preference.
GRADIENT PROCESSING
▸ Its purpose is to tailor the final image
brightness and contrast according to
the anatomy and predominant
Windowing:
pathologies to be displayed, customizing
them to the procedure. ▹ Window level - controls the
average or overall brightness of
▸ An anatomical LUT is stored by the
the image
computer for each specific type of
radiographic procedure. ▹ Window width - controls the gray
scale, (the inverse of contrast).
▸ When the radiographer enters a
particular procedure at the console, the
anatomical LUT is automatically
selected. It will customize the gray scale
and brightness of the image according
to the specific anatomy and most
common pathologies expected to be
displayed

20
FREQUENCY
PROCESSING
▸ allows an image to be sharpened,
improving its appearance. It also
allows noise to be blurred so that it is
less visible.

▹ Smoothing
▹ Edge Enhancement
▹ Equalization

▸ The ability of frequency processing to separate

structures according to their size is provided by a
mathematical process called Fourier Transform.

21
 SMOOTHING

▸ Smoothing algorithms remove some of There are several results:

the smallest details (highest-frequency ▸ First, the edges of bony structures
layers) from the image. appear “softened” due to a slight
reduction in local contrast.
▸ Smoothing is recommended to correct ▸ Second, noise is reduced in the image
for moderate amounts of mottle by mathematical interpolation which
appearing in the image. However, bear averages the pixel value of small white
in mind that severe mottle indicates or black specks with their surrounding
pixels, making them “blend in.”
gross underexposure to the IR for the
▸ Third, this same interpolation
initial projection. (averaging) process corrects for dead or
stuck dexels in the image receptor

22
SMOOTHING
ALSO KNOWN AS LOW PASS FILTER

23
 ▸ is used when the
EDGE ENHANCEMENT radiologist needs to
better visualize small
details at the edges of
bones, cartilage, or
organs such as the
kidneys.
▸ It is simply an increase in
local contrast that can
make small pathological
changes, such as a
hairline fracture, more
visible to the human eye.
▸ Also known as High pass
filtering

24
EDGE ENHANCEMENT

▸ The trade-off for using

edge enhancement is
that it also enhances
image noise

25
EQUALIZATION

▸ The application of specific image

processing that alters the pixel
values across the image to present a
more uniform image appearance.
▸ The pixel values representing low
brightness are made brighter, and
pixel values with high brightness are
made to appear less bright.

26
OTHER POSTPROCESSING FEATURES
Image reversal
• sometimes called “black bone,”
• All of the pixel values within the
image are simply changed from
high to low numbers and vice
versa. This results in “positive”
black-on-white” image rather than
the standard “negative” white-on-
black image used for conventional
radiographs.
OTHER POSTPROCESSING FEATURES
Image stitching
• is a great invention that
replaces the need for the
heavy, oversized cassettes
that used to be used for
scoliosis series.
OTHER POSTPROCESSING FEATURES
Dual-energy subtraction
• It can separate the
original image into a
tissue only image and a
bony image.
• This allows us to
determine whether a
pathological lesion is in
the soft tissue just behind
or in front of a bone or
within the bone itself.
OTHER POSTPROCESSING FEATURES

Grid line suppression

• Is a software was
developed to allow the use
of stationary grids in
mobile and trauma
radiography, and then
“erase” the grid lines they
cause with digital
processing.