Transcript Slide 1

IMAGE RECONSTRUCTION
ALGORITHM-A SET OF RULES
OR DIRECTIONS FOR
GETTING SPECIFIC OUTPUT
FROM SPECIFIC INPUT
ALGORITHM MUST
TERMINATE AFTER FINATE
NUMBER OF STEPS
DETECTORS
PRE-PROCESING
REFORMATTED RAW DATA
CONVOLUTION WITH FILTER
IMAGE RECONSTRUCTION ALGORITHM
IMAGE:
RECONSTRUCTED IMAGE
STORAGE
DISPLAY
RECORDING
ARCHIVING
IMAGE RECONSTRUCTION
PERFORMED BY:
• ARRAY PROCESSOR OR
PROCESSORS
RECONSTRUCTION WHEN
ONE OR MORE TECHNICAL
FACTORS ARE MODIFIED IS
CALLED:
• RETROSPECTIVE RECONSTRUCTION
FACTORS THAT CAN BE
CHANGED DURING
RECONSTRUCTION
•
•
•
•
•
DFOV
MATRIX
SLICE THICKNESS
SLICE INCREMENTATION
ANGLE
SFOV
LARGE DFOV
SFOV
SMALL DFOV
TARGETED (ZOOMED) RECON
SFOV
VERY SMALL DFOV
TARGETED (ZOOMED) RECON
DIFFERENT MATRICES
80 X 80
512 X 512
USUALLY MATRIX IS PERMANENTLY SET AT 512 X 512
SLICE THICKNESS
• RECONSTRUCTION IN DIFFERENT
THICKNESS ONLY POSSIBLE IN THE
MULTISLICE UNIT
SLICE INCREMENTATION
• RECONSTRUCTION IN DIFFERENT
SLICE INCREMENTATION POSSIBLE IN
SINGLE AND MULTISLICE UNITS
INCREMENTATION
SLICE
2MM
SLICE
2MM
SLICE
2MM
INCREMENT
2MM
INCREMENT
1MM
INCREMENT
4MM
CONTIGUOUS
50% OVERLAP
100% GAP
SLICE
vs
RECONSTRUCTION
INCREMENTATION
• SLICE THICKNESS
• IMAGE QUALITY
• RECON INCREMENT
• IMAGE QUALITY
TOMO ANGLE 360º
TUBE
TOMO ANGLE 180º
TUBE
TOMO ANGLE
• ANGLE
IMAGE QUALITY
TYPES OF DATA
•
•
•
•
MEASUREMENT DATA
RAW DATA
CONVOLVED DATA
IMAGE DATA
PREREQUISITE FOR DATA
RECONSTRUCTION
•
RAW
DATA AVAILABLE
MEASUREMENT DATA
(SCAN DATA)
• DATA THAT ARISES FROM
DETECTORS. IT NEEDS TO BE
PREPROCESSED TO ELIMINATE
ARTIFACTS.
RAW DATA
• IT’S THE RESULT OF SCAN DATA
BEING PRE-PROCESSED
CONVOLVED DATA
• FILTERED BACKPROJECTION IS THE
ALGORITHM USED BY MODERN CT.
• IT REQUIRES FILTERING AND THEN
BACKPROJECTION. RAW DATA IS
FILTERED USING MATHEMATICAL
FILTER.(CONVOLUTION)
• IT REMOVES BLURR. CONVOLUTION
CAN ONLY BE APPLIED TO RAW DATA.
MATHEMATICAL FILTER CAN
ALSO BE CALLED:
• KERNEL
• ALGORITHM
• PASS FILTER
TYPES OF FILTERS
•
•
•
•
SMOOTH
STANDARD
SHARP
EXTRA SHARP
TYPES OF FILTERS (SIEMENS)
• B20 (SMOOTH)
• B40 (STANDARD)
• A80 (SHARP)
• A91 (VERY SHARP)
SMOOTH
SHARP
IMAGE DATA
(RECONSTRUCTED DATA)
• CONVOLVED DATA THAT HAVE BEEN
BACKPROJECTED INTO THE IMAGE
MATRIX TO CREATE CT IMAGES
DISPLAYED ON THE MONITOR.
DETECTORS
SCAN DATA
PREPROCESSING
RAW DATA
CONVOLUTION
CONVOLVED DATA DATA
BACK
PROJECTION
BEAM GEOMETRIES AND DATA
ACQUSITION
• PARALLEL –SLOW
• FAN -FASTER
Algorithms applicable to CT
Back projection
Iterative methods
Analytic methods
BACK PROJECTION
• ALSO CALLED
SUMMATION
METHOD OR
LINEAR
SUPERPOSITION
METHOD
Example image of how a cube generates different
projections depending on the angle of projection
ITERATIVE ALGORITHMS
• SIMULTANEOUS ITERATIVE
RECONSTRUCTION TECHNIQUE
• ITERATIVE LEAST SQUARES
TECHNIQUE
• ALGEBRAIC RECONSTRUCTION
TECHNIQUE
ITERATIVE RECON
DEFINITION
• IT STARTS WITH ASSUMPTION AND
COMPARES THIS ASSUMPTION WITH A
MEASURED VALUE, MAKES
CORRECTIONS TO BRING THE TWO
VALUES IN AGREEMENT. THIS
PROCESS REPEATS OVER AND OVER.
ART USED BY HOUNSFIELD
IN HIS FIRST EMI BRAIN
SCANNER
ITERATIVE TECHNIQUES ARE
NOT USED IN TODAY’S
COMMERCIAL SCANNERS
THEY ARE VERY SLOW
BETTER THAN FILTERED BACK PROJECTION
IN METAL ARTIFACT
REDUCTION AND NOISE REDUCTION
ANALYTIC RECONSTRUCTION
ALGORITHMS
• FOURIER RECONSTRUCTION
• FILTERED BACK PROJECTION
USED IN MODERN CT SCANNERS
FILTERED BACK-PROJECTION
CONVOLUTION METHOD
• PROJECTION PROFILE IS FILTERED
OR CONVOLVED TO REMOVE THE
TYPICAL STAR LIKE BLURRING THAT IS
CHARACTERISTIC OF THE SIMPLE
BACK PROJECTION.
STEPS IN FILTERED BACK
PROJECTION
• ALL PROJECTION PROFILES ARE OBTAINED
• THE LOGARITHM OF DATA IS OBTAINED
• LOGARITHMIC VALUES ARE MULTIPLIED BY
DIGITAL FILTER
• FILTERED PROFILES ARE BACKPROJECTED
• THE FILTERED PROJECTIONS ARE SUMMED
AND THE NEGATIVE AND POSITIVE
COMPONENTS ARE CANCELLED
FOURIER RECONSTRUCTION
• USED IN MRI
• NOT USED IN CT BECAUSE OF
COMPLICATED MATHEMATICS
ALGORITHM (FILTER) vs
NOISE
DETAIL
HIGH PASS FILTER
SHARP
EDGE-ENHANCEMENT
STANDARD
LOW PASS FILTER
SMOOTHING
NOISE
RECONSTRUCTION IN SPIRAL
CT
• FILTERED BACK PROJECTION USED +
INTERPOLATION BACAUSE OF THE
CONTINUOUS MOVEMENT OF THE
PATIENT IN THE Z-DIRECTION.
( TO ELIMINATE MOTION BLURR)
RECONSTRUCTION IN
MULTISLICE SPIRAL CT
• INTERLACED SAMPLING
• LONGITUDINAL INTERPOLATION
• FAN BEAM RECONSTRUCTION
RECONSTRUCTION ALGORITHM
COMPARISON
• ANALYTIC METHODS ARE FASTER
THAN ITERATIVE ALGORITHMS.
• FILTERED BACK PROJECTION IS USED
IN MODERN CT SCANNERS
• ITERETIVE METHODS ARE BETTER
THAN FILTERED BACK PROJECTION IN
METAL ARTIFACT REDUCTION AND
NOISE REDUCTION
MPR
MULTIPLANAR
RECONSTRUCTION
(REFORMATTING)
IT USES
IMAGE
DATA
STACKS OF IMAGES ONLY IN
ONE PLANE CAN BE USED
FOR ONE TYPE OF MPR
DIFFERENT PLANE IMAGES
CAN NOT BE COMBINED
ONLY IMAGES IN ONE PLANE CAN BE COMBINED!
CURVED MPR
FOR BEST QUALITY IMAGES
• USE VOLUMETRIC DATA ACQUSITION
• RECON YOUR ORIGINAL THICKER SLICES
INTO VERY THIN SLICES (2mm OR 1 mm)
• SELECT RECON INCREMENT THAT WOULD
CREATE AT LEAST 50% OVERLAP BETWEEN
SLICES
EXAMPLE:
2 MM SLICE THICKNESS
1 MM SLICE INCREMENT
STAIRCASE ARTIFACT
3-D
EXTRUSION
IS A MODELING TECHNIQUE THAT
GENERATES A 3-D OBJECT FROM A
2 –D PROFILE ON THE COMPUTER
SCREEN.
IT USES IMAGE DATA TO
BUILD 3-D OBJECT
EXTRUSION
PIXEL AREA
B
A
A= WIDTH
B= HEIGTH
AREA OF THE PIXEL = A x B
VOXEL VOLUME
B
A
C
A= WIDTH
B= HEIGTH
C-DEPTH (SLICE THICKNESS)
VOLUME OF THE VOXEL = A x B x C
DATA ACQUSITION FOR 3-D
• CONVENTIONAL SLICE BY SLICE
• VOLUME DATA ACQUSITION
PROBLEMS WITH CONVENTIONAL
SLICE BY SLICE ACQUISITION IN 3-D
GENERATION
• MOTION - STAIR-STEP ARTIFACT
• MIREGISTRATION
STAIR-STEP ARTIFACT
SEVERE STAIR-STEP
ARTIFACT
PROCESSING FOR 3-D
•
•
•
•
SEGMENTATION
TRESHOLDING
OBJECT DELINEATION
RENDERING
SEGMENTATION
• PROCESSING TECHNIQUE USED TO
IDENTIFY THE STRUCTURE OF
INTEREST IN A GIVEN IMAGE. IT
DETERMINES WHICH VOXEL ARE PART
OF THE OBJECT AND SHOULD BE
DISPLAYED AND WHICH ARE NOT AND
SHOULD BE DISCARDED.
SEGMENTATION
THRESHOLDING
• METHOD OF CLASSIFYING THE TYPES
OF TISSUES REPRESENTED BY EACH
OF THE VOXELS. CT NUMBER IS USED
TO DETERMINE THIS.
TRESHOLDING
(IN SEGMENTATION)
DELINEATION
• BOUNDARY EXTRACTION
• VOLUME EXTRACTION
DELINEATION
RENDERING TECHNIQUES
• SURFACE RENDERING – SHADED
SURFACE DISPLAY (SSD)
• VOLUME RENDERING
SURFACE RENDERING-SSD
• SIMPLER OF THE TWO METHODS.
DISPLAYS THE IMAGE ACCORDING TO
ITS CALCULATIONS OF HOW THE
LIGHT RAYS WOULD BE REFLECTED
TO THE VIEWERS EYES.
• COMPUTER CREATES INTERNAL
REPRESENTATION OF SURFACES
ADVANTAGE OF SSD
• NOT MUCH COMPUTING POWER
REQUIRED
• ONLY CONTOUR IS USED
DISADVANTAGES OF SSD
• INFO OF STRUCTURES INSIDE OR
BEHIND THE SURFACE IS NOT
DISPLAYED!!
VOLUME RENDERING
• SOPHISTICATED TECHNIQUE. 3-D
IMAGES HAVE BETTER QUALITY THAN
IN SURFACE RENDERING. USES
ENTIRE DATA SET FROM 3-D SPACE. IT
REQUIRES MORE COMPUTING
POWER.
ADVANTAGES OF VOLUME
RENDERING (VR)
• UNLIKE SSD, VOLUME RENDERING
ALLOWS SEEING THROUGH
SURFACES. IT ALLOWS THE VIEWER
TO SEE BOTH INTERNAL AND
EXTERNAL STRUCTURES.
DISADVANTAGE/S
• IT REQUIRES GREAT COMPUTING
POWER – SOPHISTICATED COMPUTER
EQUIPMENT
MAXIMUM INTENSITY
PROJECTION
• VOLUME RENDERING 3-D TECHNIQUE
THAT IS NOW FREQUENTLY USED IN
CTA ( CT ANGIO) IT USES LESS THAN
10% OF DATA IN 3-D SPACE. IT DOES
NOT NEED SOPHISTICATED
COMPUTING.
IT ORIGINATED IN MRA
MIP ALLOWS ONLY THE
VOXEL WITH THE
BRIGHTEST VALUE TO BE
SELECTED
VR vs MIP
ADVANTAGES OF MIP
• NO NEED FOR SOPHISTICATED
COMPUTER HARDWARE- IT USES
LESS THAN 10% OF DATA
DISADVANTAGE/S OF MIP
• ARTIFACT- STRING OF BEADS
• NO SUPERIMPOSED STRUCTURES
DEMONSTRATION
ARTIFACTS
• SSD – MANY FALSE SURFACES
• MIP – MIP ARTIFACT
• VR - FEW