Transcript left : 0.90 - 0.92 right : 0.95

DATSCAN
Quantitation
Vs
Image Reading
Mike Avison
Bradford Royal Infirmary
Overview
• Simple quantitation of SPECT slices
implemented on Odyssey
• How reproducible is it?
• How does it correlate with radiologist
reports?
The Acquisition Procedure
• Acquisition protocol
–
–
–
–
–
–
–
185MBq imaging 3 hrs PI
LEHR collimators
128x128 matrix
Orbit “as-is” 13cm radius or less
360º data in 120 frames of 20sec
Mag x2 (i.e. 2.34 mm pixel size)
Brain-reach offset in Y direction
Reconstruction (workbench macro)
• Iterative reconstruciton
– OS ML-EM – all
– From vertex to lowest slice
From vertex to lowest slice
Reconstruction (workbench macro)
• Iterative reconstruciton
– OS ML-EM – all
– From vertex to lowest slice
– Check orbit
Arc and Orbit
Reconstruction (workbench macro)
• Iterative reconstruciton
– OS ML-EM – all
– From vertex to lowest slice
– Check orbit
– Check sinogram and cine for
movement
Check Sinogram
Check Cine
Filter
• Always the same
• Butterworth
– Order = 4
– Cutoff = 0.33 cycles/2pixels
–
= 0.7 cycles/cm
–
(min object size ~= 1x0.5/0.7 = 0.7 cm)
Reformatting
Occipito-frontal
orientation
Define
slices
Quantitative Macro (Pixie)
• Features
–
–
–
–
–
Background information poster at the start
Analyses headers compares with specification
Simple to operate
2 rectangular regions pre-set size
Elliptical background region pre-set size
Background
information about the
processing procedure.
Especially the
expected slice and
ROI dimensions
Give the operator
instructions about
selecting slices for
processing
Allow the operator to
modify their selection
Tell the operator how
to place ROIs
Results are printed on
the screen
Expected dimensions
of the volumes of
interest are printed
along side the actual
values
Formulae
Striatal to background uptake ratios
ratioright 
Cright / arearight  Cbg / areabg
ratiocombined 
Cbg / areabg
Ccombined / areacombined  Cbg / areabg
Cbg / areabg
Controversy – why subtract
background?
• FOR
– The box contains some
non-striatal tissues
– The scale is amplified:• No uptake = 0
• Double bg counts = 1
• AGAINST
– Its cross sectional
imaging so there is no
overlying background
Controversy, shaped ROIs?
– Use shaped ROIs to cut
out extra-striatal
tissues?
• Which bits to include?
• What margin?
• Operator dependant!
Reproducibility?
Intra-operator repeatability
– One patient processed 10 times
– LEFT : 0.90 - 0.92
– RIGHT : 0.95 – 0.96
– GLOBAL: 0.93 – 0.94
– Standard deviation
~= 0.07% of the mean
Reproducibility?
Selection of Slice Set
• How sensitive is processing to differences in slice
selection?
• Selected 5 consecutive positions for central slice set
(slices 9-13)
• Range of results = 11% of the mean.
• LESSON: display max and mean counts for each
slice on display during slice selection
Reproducibility?
Inter-operator variation
• 2 operators processed all studies
– Maximum difference = 0.18 (19% of mean)
– Std Dev of differences = 0.10 (11% of mean)
Quantitation vs Qualitative Report
1.8
Combined ratio
1.6
n = 30
1.4
1.2
1
0.8
0.6
0.4
0.2
0
-0.5
Normal
0.5
Abnormal
1.5
Conclusion
• Robust quantitative system with low intra- and
inter-operator variation.
• Good (but not perfect) separation of normal from
abnormal
• Useful when apparent uptake contradicts shape
(dot, comma paradigm)
Next Step
• Audit of patient outcome in progress.
• Compare uptake with outcome instead of
qualitative image interpretation.