Embed Size (px)
Citation preview
Phantom to test MUGA software
Bill Thomson , Joe O’BrienNigel Williams
Not New!
Phantom developed some years ago
need to test amplitude and phase component
not designed for ejection fraction
Normal / Abnormal MUGA scan
Single Harmonic Fourier
At each pixel position, we fit a curve based on first term of Fourier series
a + b.sin(t) + c.cos(t)
= A.sin(t + )
the first harmonic is a sinusoid function with the fundamental frequency and a phase angle .
Sum of Cos and Sin terms
+ =
Sum of Cos and Sin terms
+ =
Sum of Cos and Sin terms
• Frequency does not change• phase change only
+ =
Sum of Cos and Sin terms
• Frequency does not change• phase change only
+ =
Single harmonic Fourier fit
0
20
40
60
80
100
120
1 5 9 13 17 21
A
For each pixel,
AAmplitude functional imageextent of change
Phase functional imagetiming
Amplitude and Phase
Amplitude
how much movement
Phase
onset of contraction
15% threshold (from amplitude)
histogram of phase in -
whole image
ED L ventricle region
Test Procedure
Need to input sinusoid terms
want amplitude range 0 - 100%
want phase values 0 - 360
Sinusoid shape
A A
Sinusoid shape
A A
Curves generated
Curves generated
rotating plane phantom
Amplitude and phase
Amplitude and phase
Phase direction
note 90 degree shift in phase angle
Phase should follow the phantom
Phase is actually going in opposite direction
Change of ED ROI
Effect of change in ED position
Different ROI position
Phantom for EF curve creation
• Perimeter of phantom based on EF curve
• similar effects seen
• shows robust Fourier fit at low ejection fraction and counts
Previous work - different processor
Area of LV was delayedphase shows early contraction!!phase was reversed
Previous work - different processor
Area of LV was delayedphase shows early contraction!!phase was reversed
Corrected program
Higher Harmonic testing
1st harmonic
Higher Harmonic testing
1st harmonic 2nd harmonic
Higher Harmonic testing
1st harmonic
3rd harmonic
2nd harmonic
3 Harmonic phantom
3 harmonic phantom
Fourier curve fit tries to fit first EF type component only
3 harmonic phantom , different ROI
3 harmonic phantom , different ROI
3 harmonic fit to noisy data
Issues
• Phase angle appears to run in wrong direction
• Will not affect aneurism detection
• Amplitude and phase dependent on ROI
• Fourier fits to CDBS curve
• Time to Peak Fill is based on Fourier curve
• Suggest use 4 harmonic fit routinely
