Table 2.

Reconstructions that contributed to the study categorized according to ToF and non-ToF systems from specific vendors

System	Reconstruction algorithm	Reconstruction parameters	Filter	Correction	Number of reconstructions
GE Healthcare ToF	3D OSEM	MLEM 24 – 72a	All-pass Gaussian	ToF, RR ToF ToF, RR ToF –	7 2 7 3 2
GE Healthcare non-ToF	2D OSEM 3D OSEM	MLEM 16 – 64a	All-pass Gaussian	– – RR – – –	4 5 2 3f 5 4f
Philips ToF	BLOB OS TFb	4i8s 3i33s or 4i33s	–	ToF ToF	2 6
Philips non-ToF	3D RAMLAc		–	–	1
Siemens ToF	3D OSEM	1i21s 2i21s 3i21s	All-pass Gaussian All-pass Gaussian All-pass Gaussian	ToF, RR ToF, RR ToF, RR ToF, RR ToF, RR ToF, RR	9 3 4 2 5 5
Siemens non-ToF	2D OSEM 3D OSEM	MLEM 16 – 84a ‘NETTRUES’d ‘PROMPTS + RANDOMS’e		– RR RR	7 9 12

ToF’ time-of-flight, RR’ resolution recovery (point spread function recovery)

^aMaximum likelihood expectation maximization “equivalent number” = no. of iterations × no. of subsets

^bA list mode algorithm that uses spherical (‘blob’-shaped), as opposed to voxel-shaped, basis functions to enhance signal and suppress noise; the final image is produced through interpolation from overlapping blobs to voxels, which acts as a filter [35]

^cRow-action maximum-likelihood algorithm that operates using a relaxation parameter to control the amount of correction applied in each iterative update [36]

^dThe acquisition mode on the Siemens Biograph series that performs direct subtraction of delayed coincidences event-by-event [37]

^eThe acquisition mode on the Siemens Biograph series that stores delayed coincidences as a separate acquisition for subtraction from prompt events at a later time [37]

^fDenotes scanners with LYSO crystal material; all other scanners in this category (GE non-ToF) used BGO material