Image Reconstruction
Chapter 3
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Reconstruction Terminology
• Algorithms are a finite set of unambiguous steps
performed in a prescribed sequence to solve a problem
• Fourier transform is an important image-processing tool
that decomposes an image into components (waveform=
sum of sines and cosines)
• Interpolation is a mathematical method of estimating the
value of an unknown function using the known value on
either side of the function
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Image Reconstruction: Major Components
• Hardware
– Parts of the computer that can be physically touched
• Software
– Instructions that tell the computer what to do and
when to do it
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Computer Hardware
• Hard disk (Ref. Romans page 24)
• Input devices (e.g., keyboard)
• Output devices (e.g., monitor)
• Central processing unit(The central processing unit, or
CPU, interprets computer program instructions and
sequences tasks. It has been referred to as the “brain” of
the CT system.)
– Contains the microprocessor, the control unit, and
the primary memory
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Computer Memory
• Three principal types of
memory
– Read-only memory (ROM)
– Random-access memory
(RAM)
– Write-once read many
times memory (WORM)
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Raw Data
• The process of using raw data to create an image is
called image reconstruction
– Prospective reconstruction is that which is
automatically produced during scanning
– Retrospective reconstruction refers to using the same
raw data later to create a new image
• Raw data storage requires much more computer storage
space than that of image data
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Image Data
• Once the computer has processed the data into pixels
• Image data requires about one fifth of the computer
space as does raw data
• If only image data are available, data manipulation is
limited
• Image data allow measurements such as Hounsfield
units, standard deviation (see Chapter 4), and distance,
but anything not seen on the image is unavailable for
analysis.
• Image data are those which result once the computer
has processed the raw data. One Hounsfield unit value
(avrg) is assigned to each pixel.
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Image Reconstruction Process
• The DAS “reads” each arriving ray and measures how
much of the beam is attenuated
– This is a ray sum
• Many ray sums make up a view
• Many views are needed to create an image
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Image Reconstruction Process (cont’d)
• An attenuation profile is created for each view
• Information from all the profiles is projected onto a
matrix
– The process of converting the data from the
attenuation profile to a matrix is known as back
projection
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Filter Functions
• Filtering is used to minimize streaks on the image that
result from back projection
• Filtering is done by a series of complicated mathematic
steps, often referred to as a reconstruction algorithm
• Applying a filter function to an attenuation profile is
called convolution
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Filter Functions (cont’d)
• Many different filters are available
– They use different algorithms depending on which
parts of the data must be enhanced or suppressed
• Some will “smooth” the data, reducing the
difference between adjacent pixels
• Reduces artifacts but also reduces spatial
resolution
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Filter Functions (cont’d)
• Others enhance contrast by accentuating the difference
between neighboring pixels
– Improves spatial resolution, but at the cost of low-
contrast resolution
• Filter functions may be referred to as algorithms,
convolution filters, or kernels
• Filter functions can only be applied to raw data
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Reconstruction Algorithm vs Window
Setting
• Changing the algorithm changes the way the raw data
are manipulated to reconstruct the image
• This contrasts with changing the window setting, which
merely changes the way the image is viewed
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Adaptive Statistical Iterative
Reconstruction
• Newer method of image reconstruction
• Complex method that computes projections from the
image, compares it with the original projection data, and
updates the image based on the difference
• Can reduce image noise
• Shown to reduce the radiation dose to the patient by as
much as 50%
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Scan Field of View
• Selecting SFOV determines the area, within the gantry,
from which the raw are acquired
• SFOV is in the isocenter of the gantry
• SFOV selection determines the number of detector cells
collecting data
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Scan Field of View (cont’d)
• In addition to size, factors such as calibration vectors and
types of image processing are often incorporated into
SFOV
• Anything outside the SFOV is not imaged
• To produce the highest quality image, select the SFOV
that comes closest to just encompassing the patient
• Parts of the patient located outside the SFOV may cause
out-of-field artifacts on the image
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Display Field of View
• DFOV selection determines how much of the collected
raw data are used to create an image
• DFOV is also called zoom or target
• Changing the DFOV affects image quality by changing the
pixel size
• DFOV works in a manner similar to the zoom on a
camera
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Display Field of View (cont’d)
• Because the data selected for the DFOV are a subset of
all the scan data available, the DFOV cannot be larger
than the SFOV
• Choosing the optimal DFOV improves the detectability of
abnormalities
– A DFOV that is too large makes anatomic structures
unnecessarily small
– A DFOV that is too small may exclude important
patient anatomy
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Image Center
• Image coordinates allow the operator to specify the area
within the SFOV that will be displayed on the image
• Some manufacturers use x and y preceding the number
to designate the direction of the coordinates
• Some manufacturers precede the number by R (right), L
(left), A (anterior), P (posterior), S (superior), and I
(inferior)
– Often referred to as RAS coordinates
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Image Center (cont’d)
Image center adjusted using traditional x,y coordinates
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins
Image Center (cont’d)
Image center adjusted using RAS coordinates
Copyright © 2011 Wolters Kluwer Health | Lippincott Williams & Wilkins