Singlephoton Emission Computed Tomography SPECT

1
Single-photon Emission Computed Tomography(SPECT)

• Old Chapter 9, New Chapter 11

2
Single-photon Emission Computed Tomography

• Also known as SPECT
• Is a radionuclide imaging technique which
  produces a picture of the radionuclide
  distribution within a thin section of the patient

3
SPECT

• Advantages
• Ability to show radionuclide distribution in any
  slice of the body-coronal, sagittal, transverse,
  or oblique plane
• Great at detecting and accurately locating
  lesions in a complex anatomical structure
• Yields a higher image contrast when compared with
  planar images because activities lying outside
  the plane of interest are reduced

4
Instrumentation-System Description

• SPECT system consists of two basic components
• A radiation detection device for measuring the
  radioactivity profiles at various angles around
  the patient
• A computer for processing the projection profiles
  to form an image of the cross-section

5
Instrumentation-System Description

• Conventional gamma camera on a gantry that
  rotates around the patient to obtain projection
  profiles

6
Instrumentation-System Description

• Most systems use the step and shoot technique, in
  which the camera moves in stepwise rotation
• Image acquisition is done with the camera stopped
  at regular angular intervals around the patient

7
Instrumentation-System Description

• About 30 minutes are required to acquire one
  complete set of projection data for
  reconstruction
• Why 30 minutes ???

8
Instrumentation-The Computer System

• Triggers the gantry at regular time intervals to
  rotate for a precise angular increment
• Must have a large CPU or buffer memory and ample
  disk storage
• Most have an array processor, which are hardware
  devices designed to do only a limited variety of
  arithmetic operations on matrices and highly
  ordered array of numbers

9
Simple Backprojection Method

• Take an image of a point source from different
  angles, divide each of these images into a number
  of strips and select the strip that passes
  through the image of the point source
• Plot the number of counts in each pixel across
  the selected strip
• This is a projection profile

10
Simple Backprojection Method

• The number of counts at each point in the
  projection profile is called the ray-sum because
  it represents the total number of counts seen by
  the gamma camera along a line of view

11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
Simple Backprojection Method

• Advantage-the ease with which we can construct a
  transverse tomogram without complicated
  mathematics or even a computer
• Disadvantage- the starburst artifact in the
  reconstructed image

15
The Filtered BackProjection Method

• Modify the original projection profiles with a
  number scheme to produce a projection profile
  with a negative lobe on each side of the peak.

16
(No Transcript)
17
(No Transcript)
18
The Filtered BackProjection Method

• Filtered backprojection method- a convolution of
  the projection profile with a spatial filter and
  the image reconstruction scheme

19
The Filtered BackProjection Method

• The ideal reconstruction filter is the ramp
  filter with multimillion counts in each
  projection profile
• Frequency Domain
• Ramp filter-a straight line that starts from zero
  and goes up linearly with the spatial frequency

20
The Filtered BackProjection Method

• Ramp filter-puts a low emphasis on the
  low-frequency components of the projection
  profile as compensation for redundant sampling of
  low-frequency data
• Heavily weighs the high frequency components to
  produce high resolution in the reconstructed image

21
(No Transcript)
22
The Filtered BackProjection Method

• Nyquist frequency-the maximum frequency
  resolvable by the camera computer system
• Imposing a range of frequencies to be included in
  the reconstruction is modifying the ramp filter
  with a window function

23
The Filtered BackProjection Method

• If count statistics are low, the fraction of
  noise is high
• The ramp filter amplifies high frequencies, so
  the noise is amplified
• As result, we end up with a high resolution image
  mingled with an exaggerated amount of noise

24
The Filtered BackProjection Method

• Noise can be reduced without introducing
  artifacts by modifying the ramp filter with a
  window function that keeps the ramp filter intact
  in the low-frequency region
• Purpose of a window is to limit the
  reconstruction to data within a range of
  frequencies

25
(No Transcript)
26
The Filtered BackProjection Method

• Butterworth window
• In the low frequency region the coeficients on it
  is equal to 1.0
• At halfway up to the Nyquist frequency, the curve
  takes on a fractional value and descends rapidly
  before gradually rolling to zero at the Nyquist
  frequency

27
The Filtered BackProjection Method

• Butterworth window
• When the ramp filter is modified by the
  coefficient of the Butterworth window at the
  corresponding frequency, the ramp remains
  unchanged in the low-frequency region because any
  number multiplied by one is still the same number.

28
Butterworth Filter

• Is the composite filter
• The noise is eliminated, but the spatial
  resolution is reduced

29
Optimal Reconstruction Filter

• A function of the count statistics
• The higher the count statistics, then more high
  frequencies can be included in the reconstruction
• Start the reconstruction with a high cut-off
  frequency if too fuzzy and grainy, drop the
  cut-off frequency to produce a smoother image
• If over smoothed, raise the cutoff frequency and
  try again

30
Physical Factors Affecting te Quality of SPECT
Images

• During acquisition we need to endure the accuracy
  of the gantry positioning and to optimize the
  angular and linear sampling frequencies for the
  type of clinical studies.
• Additional corrections
• Attenuation
• Scatter
• Collimator resolution with depth
• Poor counting statistics
• Operational characteristics of the
  camera-computer system

31
Linear Sampling Criteria

• Linear sample - the same as the linear dimension
  of the image matrix (number of pixels in a row
  across the matrix, 64X64, 64 linear samples).
• The number of counts recorded in each pixel is
  called the ray-sum of the linear sample
• A plot of the number of counts in each pixel
  along the row is called the projection profile

32
Linear Sampling Criteria

• Nyquist sampling theorem- states that the maximum
  spatial frequency that we can observe is only
  half of the sampling frequency that we use to
  acquire the image
• If we take two measurements in each centimeter,
  the Nyquist sampling theorem says that the
  smallest lesion we can observe is one centimeter.

33
Nyquist Sampling Theorem

• The smallest lesion we can observe is one
  centimeter. Lesions smaller than one centimeter
  may not only be unresolvable due to inadequate
  sampling, but may introduce artifacts known as
  aliases.
• Sampling frequency, f, is equal to the n number
  of pixels in the row divided by the diameter, D,
  of the filed of view
• F n/D

34
Nyquist Sampling Theorem

• 40 cm field of view
• 64 X 64 matrix
• F 64/40 cm or 1.6 cycles per centimeter, space
  interval between samples is 0.625 cm
• Nyquist can only see 0.8 cycles per centimeter
  or a sample spacing of 1.25 cm

35
Nyquist Sampling Theorem

• The spatial resolution of the gamma camera
  ultimately determines the maximum observable
  spatial frequency in the image
• The matrix size should be consistent with the
  spatial resolution of the gamma camera
• Compromise to keep statistical noise, acquisition
  time and computation time reasonable

36
Angular Sampling Criteria

• Number of angular samples is a compromise between
  the count density needed in each projection to
  keep image noise at an acceptable level and
  patient motion.
• 2 10 degree samples
• 180 360 degrees
• 30 minutes

37
Angular Sampling Criteria

• 2 degrees tend to be fuzzy because of limited
  counts
• Greater than 10 degrees cause streak artifacts
• 6 degrees tend to give best compromise between
  statistical noise and streak artifacts

38
Photon Attenuation Correction

• As the photons travel from their origin toward
  the camera, some are scattered away from the
  camera and some are absorbed by the intervening
  body tissues.
• The ray-sum is not linearly proportional to the
  quantity of radionuclide along the ray
• Hot rim artifact at boundaries

39
Photon Attenuation Correction

• There are three approaches to correct this
• Incorporate a photon attenuation correction
  factor directly in the reconstruction algorithm
• Correct the projection data for attenuation prior
  to reconstruction, by acquiring data through 360
  degree and use geometric mean to calculate
  (simplest)
• Apply a correction factor to each pixel in the
  reconstructed image

40
Photon Attenuation Correction

• Reconstruction algorithm using reconstructed
  transmission image and attenuation coefficients
• 153-Gd, 133-Ba, 137-Cs

41
Compton-Scattered Photons

• Low energy photons lose only a small fraction of
  their energy as a result of small-angle scatter
• Consequently, much of these small-angle scattered
  photons are able to pass through the pulse-height
  analyzer and get included in the projection
  images.
• Degrades image resolution

42
Compton-Scattered Photons

• Because scattered photons increase the number of
  counts recorded in the ray-sum, the observed rate
  of a point source does not drop off as rapidly
  with depth in the tissue as predicted by the
  exponential attenuation equation

43
Compton-Scattered Photons

• Eliminate those unwelcomed counts form getting
  recorded in the ray-sum,-2 methods
• 1. Use an asymmetric energy window that is
  shifted toward the high energy end of the
  spectrum
• 2. Subtract from the projection image an image of
  the scattered-photon distribution using a window
  in the Compton region of the spectrum

44
Non-Circular Detector Motion

• Each projected ray on a gamma camera is defined
  by the field of view of channels in the
  collimator.
• The field of view of a collimator channel
  diverges as the distance from the face of the
  collimator increases

45
Non-Circular Detector Motion

• The structures deep inside the body are not as
  well resolved as those near the body surface
• If the camera follows a circular orbit around the
  patient, the patients position if first adjusted
  so that the axis of rotation coincide with the
  midline of the body

46
Non-Circular Detector Motion

• The radius of rotation (the distance of the
  camera from the axis of rotation) is adjusted to
  clear the patients shoulders or hips
• Elliptical orbit- to continuously adjust the
  detectors radius of rotation to minimize
  patient-to-detector distance.

47
Non-Circular Detector Motion
48
Non-Circular Detector Motion

• Axis of rotation must remain fixed at one
  position inside the body
• The face of the collimator must be perpendicular
  to the radius of rotation at all angular
  positions
• Each angular increment of the detector must be
  equal

49
Image Noise

• The major source of image noise is the limited
  number of counts in a reconstructed image.
• Patients undergoing SPECT scans are routinely
  given a dose of 30 - 50 higher than that for a
  planar imaging

50
Quality Assurance of the SPECT System

• Camera Field Uniformity
• Center of Rotation
• Pixel Size
• Orthogonality and Centering

51
Camera Field Uniformity

• 1 nonuniformity can be amplified as much as 20
  nonuniformity near the axis of rotation in
  reconstructed images
• Correction matrix matirx of multiplication
  factors to modify the number of counts in
  corresponding pixels
• Uniform flood source with less than 1 variation
  in activity 57-Co
• 64 X64 30 million counts
• 128 X 128 120 million counts

52
Center of Rotation

• The center of rotation in the projection image
  can be derived from two images of a point source,
  each taken 180 degrees apart.
• The center of rotation is located at the pixel
  halfway between the two pixels occupied by the
  point source image in the opposing views
• 30 pairs of images are acquired to calculate the
  center of rotation.

53
Pixel Size

• Accurate calibration of the pixel size is
  necessary for photon attenuation correction and
  for measurement of organ or lesion volume.
• The dimensions of a pixel can be easily measured
  by taken an image of two point sources located at
  a known distance apart

54
Pixel Size

• Once an image of the two point or line sources
  has been acquired, the pixel location of each
  source in the computer image is identified from
  the activity profile.
• Pixel dimension is calculated by dividing the
  separation distance between the two point sources
  by the number of pixels between the two peaks in
  the activity profile of the image

55
Orthogonality and Centering

• By adjusting the offset control of the x and y
  positional signals from the camera, the axes of
  the image matrix can be calibrated to intersect
  at the exact midpoint
• Can be done using five Co57 sources