centre of rotation: is there a problem in the y dimension? stephen brown - southend mike avison -...
TRANSCRIPT
Centre of Rotation:Is there a problem in the Y
dimension?
Stephen Brown - Southend
Mike Avison - Bradford
Y axis
Tc99m point source positioned on-axis
In the ideal situation the source would appear in the central pixel in both heads
Y axis
Tc99m point source positioned on-axis
-the problem of sag (affects y alignment)
Y axis
Y error should be proportional to radius
Sometimes we used the 3 point IRC source
Y error on rear source is less than front source-the radii are more similar for rear source
Y axis
Actual Variation of Y offset with radius
0
2
4
6
8
10
0 20 40 60
Detector radius
Max peak2peak y-deviation
(mm)
Spot 1
Spot 2
Spot 3
Linear (Spot 3)
Linear (Spot 2)
Linear (Spot 1)
(mid)(rear)
(front)
How did we measure IRC Y error
• Acquire dual head 360° of data (point source)
• Use Display A - FWHM – Place ROI over point, read Y centroid– Enter data in spread sheet
• Interfile Export to Park– Automatic processing checks X too
• Odyssey IRC test – gives misleading and ill defined results
IRC Y Deviants form mean MEGP
-1.5
-1
-0.5
0
0.5
1
1.5
0 20 40 60 80 100 120
Projections
Co
ord
ina
te f
o c
en
tro
id
Spot 1
Spot 2
Spot 3
How did we measure IRC Y error
• Acquire dual head 360° of data (point source)
• Use Display A - FWHM – Place ROI over point, read Y centroid– Enter data in spread sheet
• Interfile Export to Park independent computer– Automatic processing checks X too
• Odyssey IRC test – gives misleading and ill defined results
How did we measure IRC Y error?
• Acquire dual head 360° of data (point source)
• Use Display A - FWHM – Place ROI over point, read Y centroid– Enter data in spread sheet
• Interfile Export to Park– Automatic processing checks X too
• Odyssey IRC test – gives misleading and ill defined results
<0.5<0.5<0.65
<1.0<1.0<1.5
Ref. Ref.
Is this X or Y or a combination of both?It’s a long way from 5mm so what does it mean?
I think I was !!!
Results(heads at max radius 38.6 cm)
Camera (collimator) Maximum Y deviation (mm)Front Mid Rear
Axis Southend (LEHR) 7.7 5.5 3.6
Axis 1 BRI (LEGP) 5.0
Axis 2 BRI (LEHR) 7.4 5.3 5.2
Axis 1 BRI (MEGP) 7.6 7.1 7.1
Acceptance criteria: typically 1 or 2mm maximum Y deviation
mean = 5.7
Is 5 mm too big?
• We are hoping to resolve objects of about 10mm in SPECT of trunk sized volumes
• What resolution do we aspire to for DAT scan?
• Guidance from professional bodies state – 1mm or 2mm (range) as acceptable
• How do other gamma cameras perform?
Is this bad performance?
Comparison with other brands of camera
Results - by brand (LE collimators)
Camera Max Y deviation (mm) Radius cm
Axis (avg) 5.7 38
E Cam 3.1 33
Hawkeye 1.7 33
Argus 4.8 (single head) 33
Forte A5.4 34
Forte B5.4 34
Results - by brand: NormalisedNormalised to 33cm radius
(LE collimators)
Camera Max Y deviation (mm)
Axis (avg) 5.0
E Cam 3.1
Hawkeye 1.7
Argus 4.8 (single head)
Forte A 5.2
Forte B 5.2
Range of Y COR (IRC) Error
0
1
2
3
4
5
6
Axis Forte ECam Hawkeye
Camera
mm
Y axis
word of warning ...
Symmetrical constant misalignment (not sag).
As the gantry rotates, y-centroids remain fixed therefore no error detected
We tested with source on axis
Y axis
Tc99m point source positioned off-axis
-Source off centre in Y vertical direction
- as the gantry rotates, y-centroids on each detector move
-therefore the error is detected
General Causes of COR errors
• Misalignment of electronic and physical axis. (Electronic might change with angle)
• Sag of detectors (Physical change with angle)
• Detector misalignment
• Collimators not fixed firmly
• Collimators warp under gravity
• Non-linearity
What should we be testing?• Maybe …
– acquisition of IRC jig 10cm lat. from iso-centre– radius 33cm– dual head
• 180° • 102° non-circular orbit
• Write macro to do analysis (X and Y errors)
• When you get home try the display A method for Y errors only
Summary 1
• If you are using Odyssey IRC test you are probably being misled into believing performance is much better than it really is.
• Philips should supply better documentation.
Summary 2• Philips should improve correction:
– Forte and Axis. (Worst in class).
• Our measurements indicate:– If Philips modified the IRC cal. so that it took mean Y
offsets for both heads over 360° and used the means to correct data, then offset errors could be reduced to 2.1 mm (Axis)
– If Philips derived a variable correction as a function radius and angle: error could be eliminated at COR but linearly increase with radius (to 2.1 mm at 33cm)
– This would probably meet all aspirations.
• Further improvement would require gantry re-engineering
Axis 1 X centroids
raw fitted
Y centroids raw
mean
X error Y error
Axis 2X centroids
raw fitted
Y centroids raw
mean
X error Y error
ECamX centroids
raw fitted
Y centroids raw
mean
X error Y error
HawkeyeX centroids
raw fitted
Y centroids raw
mean
X error Y error
Forte AX centroids
raw fitted
Y centroids raw
mean
X error Y error
Forte BX centroids
raw fitted
Y centroids raw
mean
X error Y error