Compartment model based analytical PET simulator for PVELab

1
Compartment model based analytical PET simulator
for PVELab
L. Balkay, I. Valastyán, M. Emri, L. Trón UDMHSC,
PET Center, Debrecen, Hungary
2

• Monte Carlo simulators
• (Eidolon PET simulator, SimSET package,)
• tracks each individual g photons from the
  annihilation to the final absorption or escape
• it can take into consideration the PMT
  characteristics and the whole coincidence signal
  processing
• conceptually precise, versatile, but not fast

• Analytical simulators(AS)
• (McConnell Brain Imaging Center ,)
• analytically models the most important photon
  interactions (attenuation, scatter, randoms)
• adding poisson noise to sinograms
• although less versatile, very fast (1min/slice)
  allowing repeated simulation as often as necessary

3
Analytical PET simulator

convolving with the PET PSF

• Pk (TkAkSk)NkRk
• Tk true counts
• Ak attenuation factors
• Sk scatter counts
• Rk random counts
• Nk normalization factors

4
Option kinetic model based simulation
k1, k2, maps
inputs
frames of true activity
distortions
tracer kinetic information

• model
• kinetic const.
• blood curve

. . .
5
Implementation of kinetic modeling

• Generalized matrix representation of dynamic
  system models

X conc. vector ,B matrix, vector of
kin. const. The general solution

• Examples

6
Selectable analytical blood curves
7
Main steps during the simulation
Add the statistical noise (Poisson distribution)
Generate the radioactivity distribution and the
statistical error of the radioactive decay
Transform to the sinogram space
Spatial blur with PSF
Theoretical activity distribution
Simulate the attenuation of the source object
Simulate the Compton scattering
Simulate the random coincidences
Simulate the instrumental and physical effects as
Poisson processes
Apply the scatter correction
Apply the random correction
Apply the attenuation correction
Correct the distortions of the acquisition
Reconstruction, simulated PET image
8
The Matlab GUI
9
GUI to explore the input volume.
10
The GUI for kinetic model definition.
11
The Matlab GUI
12
Validation using the Hoffman slice phantom.
Measured and simulated images
The simulation time of one dynamic slice took
approximately 5 minutes on one 2.8 GHz processor.
13
Simulating C11FCWAY accumulation
14
Simulating FDG accumulation
T1.5 min
T2 min
T7 min
T35 min
T60 min
T120 min
15
Simulating F18-L-Dopa accumulation
Summed image slice from 40 min to 80 min of
simulated dynamic frames
16
Compartment model based analytical PET simulator
for PVELab
The program can be downloaded from http\\pet.dot
e.hu\pveout