Conventional Reconstruction Followed by Kinetic Modelling PET Scanner

Conventional Reconstruction Followed by Kinetic Modelling PET Scanner Kinetic Model k 3 K 1 P F k 2 Back-Project sinograms for each time-frame separately; A-1(Y) = F(t, j) Draw regions, make time-activity curves. B k 4 Estimate parameters, j, for a single region: j = [K 1 = ? , k 2 = ? , k 3 = ? , k 4 = ? ]

Step-by-step: Conventional Recon Followed by Kinetic Modelling PET Scanner Kinetic Model k 3 K 1 P F k 2 Back-Project sinograms for each time-frame separately; A-1(Y) = F(t, j) Draw regions, make time-activity curves. B k 4 Estimate parameters, j, for a sinlge region: j = [K 1 = ? , k 2 = ? , k 3 = ? , k 4 = ? ]

Direct Reconstruction Scheme measured data k 3 PET Scanner K 1 Kinetic Model k 3 K 1 k 2 P F k 1 k 2 F B k 4 Guess Parametric Images, Solve Model for Emission Forward Project Emisisons, F E[Y |F(j)] = AF(j) + m Calculate Fit of Sinograms min {-LL(Y|j)}

Step-by-step: Direct Reconstruction measured data k 3 PET Scanner K 1 Kinetic Model k 3 K 1 k 2 P F k 1 k 2 F B k 4 Guess Parametric Images, Solve Model for Emission Forward Project Emisisons, F E[Y |F(j)] = AF(j) + m Calculate Fit of Sinograms min {-LL(Y|j)}

Direct Reconstruction Scheme measured data k 3 PET Scanner K 1 Kinetic Model k 3 K 1 k 2 P F k 2 F B k 4 Guess NEW Parametric Images, Solve Model for Emission Forward Project Emisisons, F E[Y |F(j)] = AF(j) + m Calculate Regularized Fit of Sinograms min {-LL(Y|j) + S(j)}

Direct Recon. with Regularization measured data k 3 PET Scanner K 1 k 2 Kinetic tric Model F e ! m ra. K 1 gesk 3 a PP Ima. F B k 2 k 4 ! e c n k 4 n Co Guess Regularized Parametric Images, Solve Model for Emission e g r ve Forward Project Emisisons, F E[Y |F(j)] = AF(j) + m Calculate Regularized Fit of Sinograms min {-LL(Y|j) + S(j)}

Conventional Recon. applied pixel-by-pixel to Rat Phantom Images Kamasak, Bouman, Morris, Sauer (2005) Direct Reconstruction of Kinetic Parameter Images from Dynamic PET Data. IEEE Trans Med Imag (In Press)

Direct Recon. applied to Rat Phantom Images Kamasak, Bouman, Morris, Sauer (2005) Direct Reconstruction of Kinetic Parameter Images from Dynamic PET Data. IEEE Trans Med Imag (In Press)

Direct Recon. Monkey Data 18 F-fallypride K 1 k 3 K 1 P F k 2 B k 3 k 4 4 -18. 2; Moderate Regulz’n on K 1, k 2, k 3, k 4 none on BP or VD data care of Dr B Christian, Kettering Med Center k 2 Recons care of M Kamasak, C Bouman of Purdue k 4

Direct Recon. Monkey Data 18 F-fallypride K 1 k 3 K 1 P F k 2 B k 3 k 4 4 -18. 2; High Regularization on K 1, k 2, k 3, k 4 none on BP or VD data care of Dr B Christian, Kettering Med Center k 2 Recons care of M Kamasak, C Bouman of Purdue k 4

Where do we go from here?

Combine Direct Recon. with “nt. PET” model k 3 K 1 Dopamine Take-off Time P F k 2 10, 000 Microdialysis Probes in every scan! Plasma K 1 k 2 B Dopamine Peak height k 4 Free Bound DA DA Free k 3 k 4 Bound Dopamine Peak Time

What about validation of parameter estimates in direct recon?

Recall that Multi-injection experiments can be used to get highly precise parameter estimates. P F B N P F B F N incoporate multi-species model into direct reconstruction framework compare estimates by region to estimates at each voxel