NeuroLF_Hoffman_Dataset/README.md
Overall description
This dataset contains a 20 minute acquisition at 39 MBq of a cylindrical Hoffman brain phantom, that simulates an FDG brain scan by having a structure of many plastic slices, the gaps between which are filled with radioactive water.
The dataset was acquired on an R&D model of Positrigo's NeuroLF brain PET scanner using acquisition code that was still under development.
Phantom description
Hoffman brain phantom:
- cylindrical phantom with layered geometry to achieve three levels of activity in brain FDG-like outlines
- filled with F18-FDG solution
- 39 MBq
- 20 minute acquisition
- for purposes of attenuation, the plastic and water were both set to 0.096 cm2/g
- the outside dimensions of the cylinder are 175 mm length and 203 mm diameter and it was approximately centred in the field of view
Institution
University Hospital Leipzig, Department of Nuclear Medicine, Germany.
Scanner model
Positrigo NeuroLF S1 brain scanner
- detectors_per_ring = 256
- number_of_rings = 48
- arccorrected_bins = 180
- inner_ring_radius_mm = 134
- crystal_length_mm = 15.0
- avg_interaction_depth_mm = 8.26007
- crystal_spacing_mm = 3.313
- ring_spacing_mm = crystal_spacing_mm
- bin_size_mm = crystal_spacing_mm / 2
- intrinsic_tilt_rad = -3.1091819
- num_axial_blocks_per_bucket = 6
- num_transaxial_blocks_per_bucket = 4
- num_axial_crystals_per_block = 8
- num_transaxial_crystals_per_block = 8
- num_crystals_per_block = 1
- num_detector_layers = 1
- energy_resolution = 0.25
- reference_energy = 511
- max_TOF_bins = 1
- size_TOF_pos = 0
- timing_resolution = 500
- axial_block_spacing_mm = 27.36
- transaxial_block_spacing_mm = 27.36
- coincidence_window_ns = 3.5
scanner = stir.Scanner(stir.Scanner.User_defined_scanner, "NeuroLF_15mm", detectors_per_ring, number_of_rings,
arccorrected_bins, arccorrected_bins, inner_ring_radius_mm, avg_interaction_depth_mm, ring_spacing_mm, bin_size_mm,
intrinsic_tilt_rad, num_axial_blocks_per_bucket, num_transaxial_blocks_per_bucket,
num_axial_crystals_per_block, num_transaxial_crystals_per_block, num_crystals_per_block, num_crystals_per_block,
num_detector_layers, energy_resolution, reference_energy, max_TOF_bins, size_TOF_pos, timing_resolution,
"BlocksOnCylindrical", crystal_spacing_mm, crystal_spacing_mm, axial_block_spacing_mm, transaxial_block_spacing_mm)
Acquisition Date
04 APRIL 2024
FDG, F-18
Preparation protocol
Some text on set-up, ideally including expected activities in various inserts (cross-calibrated to start of PET scan)
Acquisition protocol
A 20 min PET scan was performed in list-mode.
Example vendor reconstruction settings
- STIR OSMAPOSL with parallelproj
- 50 iterations
- 1 subset
- quadratic prior with penalty factor 5
- decay, randoms, attenuation and scatter correction
VOI description
The phantom model (https://depts.washington.edu/petctdro/DROhoffman_main.html) was first averaged axially across corresponding sets of 10 slices and then registered to the reconstruction using NiftyReg. Then, all voxels with activity above 0.9 were assigned to grey matter (value 3), all voxels below 0.05 to ventricles (value 1) or background (value 0) and the voxels in between to white matter (value 2).
VOI 1: whole phantom
convex hull of the active part of the phantom
VOI 2: ventricles, white and grey matter
0: background, 1: ventricles, 2: white matter, 3: grey matter
Low activity brain regions like the white matter are created by having two thin activity layers instead of 10 - this leads to some visible transaxial patterns in the reconstructed image.
Also some slivers of activity are left in the background, because the plastic slices have gaps in between where water can go in.