Scanner models

Overview

There are no preset factory functions in the package. The clinical systems below are the documented parameter sets that the example notebooks pass to the constructors, reproduced keyword for keyword, so any of them can be pasted into a session. Parameters a vendor does not publish are marked as modelling assumptions, as the notebooks mark them.

Anything that changes between scans (kVp, mA, views, rotation time, collimation, pitch, added filters such as a tin shield) belongs in CTProtocol, not in the scanner. A dual-source system is modelled as two acquisitions: one scanner per tube/detector chain, each with its own protocol.

Constructor defaults

Every keyword has a default (a CatSim-like research scanner), so EICTScanner() alone is valid; PCCTScanner additionally requires energy_thresholds. Distances are in mm, detector sizes are at the isocentre.

Geometry — both families

EICTScanner — scintillator

PCCTScanner — direct conversion

Detectors, bowties, filters

Energy-integrating detector materials

With SimOptions(use_detector_efficiency = true) (the default) the energy-integrating forward model weights every energy by a Monte Carlo absorbed-fraction table η(E), 1–140 keV. detector_efficiency_mode = :beer_lambert switches to the analytical fallback, which is the only place detector_depth matters. Any other material symbol is an error.

The garnet table carries the Tb (52.0 keV) and Lu (63.3 keV) K-fluorescence escape dips; the two UFC tables carry the Gd K-edge escape at 50.2 keV. The Flash crystal is thinner than the Force's: the two tables agree below 30 keV and differ by −28 % at 140 keV (0.588 vs 0.816), and a test forbids aliasing them.

Photon-counting sensors

:CdTe, :CZT and :Si. The photon-counting chain always applies a Monte Carlo detector-response matrix (photoelectric, Compton and Rayleigh transport, Fano noise, charge sharing, K-fluorescence) before thresholding into bins, then the Monte Carlo pile-up and the optional pile-up and scatter corrections set on the scanner.

Bowtie filters

Filtration

The scanner's flat filter is flat_filter_material (:aluminum, :copper or :titanium) at flat_filter_thickness mm. Per-scan layers go on the protocol: CTProtocol(additional_filters = [("Ti", 0.9), ("Sn", 0.6)]) accepts "Al", "Cu", "Sn", "Ti", "C" and "W". Source spectra are the bundled IPEM tables at 8° and 10° anode angles (CTProtocol(anode_angle = 8)), up to 140 kVp.

GE Revolution Apex Elite

A 256-row wide-cone system with 0.625 mm rows and a curved Gemstone (Lumex) garnet array. Notebooks 01, 02 and 05 add clinical DAS noise (3500 e⁻); the VMI and CatSim notebooks (03, 06, 07) set electronic_noise = 0.

scanner = BS.EICTScanner(
    source_to_isocenter = 625.6,  source_to_detector = 1100.0,
    detector_rows = 256,          detector_cols = 834,
    detector_row_size = 0.625,    detector_col_size = 0.6,
    detector_col_offset = 0.25,   detector_shape = :arc,     # the defaults, written out
    focal_spot_width = 1.0,       focal_spot_length = 1.0,   target_angle = 10.0,
    flat_filter_material = :aluminum, flat_filter_thickness = 2.5,
    bowtie_filter = :ge_revolution_large,
    detector_material = :lumex,   detector_depth = 3.0,
    fill_factor_row = 0.9,        fill_factor_col = 0.9,
    electronic_noise = 3500.0,    # e⁻, clinical DAS readout (0 in the VMI notebooks)
    detection_gain = 10.0,        # e⁻/keV
)

Rapid kVp switching is each energy's duty cycle of the view: 0.65 of the view period at 80 kVp and 0.35 at 140 kVp, for the tube current and for the arc the view integrates over (view_arc). DOSE_SCALE scales both to the physical scan's CTDIvol of 10.07 mGy. Every notebook integrates each view over its arc with SimOptions(; view_samples = 5) (CatSim's comparison in 06 uses point views on both sides).

Siemens NAEOTOM Alpha

The photon-counting system. Native CdTe dexels of 0.275 × 0.322 mm at the detector face are binned 2 × 2, which at the 1113/610 magnification gives 0.301 × 0.353 mm pixels at the isocentre and 144 rows; notebooks 04 and 08 both span the 50 cm scan field (1659 columns). Four thresholds at 20, 35, 55 and 70 keV make the bins 20–35, 35–55, 55–70 and > 70 keV.

scanner = let native_col = 0.275, native_row = 0.322,   # mm at the detector face
               sid = 610.0, sdd = 1113.0, bf = 2
    col_iso = native_col * bf / (sdd / sid)                # 0.301 mm
    row_iso = native_row * bf / (sdd / sid)                # 0.353 mm
    BS.PCCTScanner(
        source_to_isocenter = sid, source_to_detector = sdd,
        detector_rows = 144, detector_cols = ceil(Int, 500.0 / col_iso),   # the 50 cm scan field
        detector_row_size = row_iso, detector_col_size = col_iso,
        detector_shape = :arc, detector_col_offset = 0.25,
        focal_spot_width = 0.4, focal_spot_length = 0.5, target_angle = 7.0,
        gantry_rotation_time = 0.5, scan_diameter = 500.0, gantry_aperture = 820.0,
        flat_filter_material = :aluminum, flat_filter_thickness = 3.0,
        bowtie_filter = :large_body,                        # both notebooks
        detector_material = :cdte, detector_depth = 1.6,
        fill_factor_row = 0.95, fill_factor_col = 0.95,
        energy_thresholds = [20.0, 35.0, 55.0, 70.0],
        energy_resolution = 10.0, charge_sharing_fwhm = 0.08, dead_time_ns = 5.0,
        pixel_mode = :standard,
        native_dexel_col_mm = native_col, native_dexel_row_mm = native_row, binning_factor = bf,
        pileup_correction = true, scatter_correction = true, noise_reduction = 0.0,
    )
end

Both notebooks scan at 140 kVp and 174 mA × DOSE_SCALE (the physical CTDIvol of 10.12 mGy), 1200 views in 0.5 s, 5 mm collimation, with the tube's 0.9 mm titanium window as additional_filters = [("Ti", 0.9)] on top of the 3 mm aluminium, exact Poisson counts, and view integration on the photon-counting path (view_samples = 5).

Siemens SOMATOM Force

Third-generation dual source: 96 rows of 0.6 mm (57.6 mm), 920 channels of 0.054° (0.561 mm at the isocentre). Both tubes use tube A's geometry so the low/high sinogram pair is co-registered ray by ray; the Gammex body fits inside detector B's real 35.5 cm field, so no ray used would be missing on the real detector.

scanner = BS.EICTScanner(
    source_to_isocenter = 595.0, source_to_detector = 1085.6,
    detector_rows = 96,          detector_cols = 920,
    detector_row_size = 0.6,     detector_col_size = 0.561,
    focal_spot_width = 0.8,      focal_spot_length = 1.2,  target_angle = 8.0,
    flat_filter_material = :aluminum, flat_filter_thickness = 3.0,
    bowtie_filter = :large_body,                 # assumption: Siemens profile unpublished
    detector_material = :ufc,    detector_depth = 1.4,
    fill_factor_row = 0.9,       fill_factor_col = 0.9,
    electronic_noise = 0,        detection_gain = 10.0,
)
protocol_low  = BS.CTProtocol(kVp = 100, mA = 380.0, views = 1160, rotation_time = 0.5,
                              collimation_mm = 4.8, anode_angle = 8,
                              additional_filters = [("Ti", 0.9)])
protocol_high = BS.CTProtocol(kVp = 140, mA = 190.0, views = 1160, rotation_time = 0.5,
                              collimation_mm = 4.8, anode_angle = 8,
                              additional_filters = [("Ti", 0.9), ("Sn", 0.6)])

The clinical abdomen pairs are x/Sn150; the bundled spectra stop at 140 kVp, so the notebook runs 100/Sn140 at the published 2:1 current ratio. Assumptions: 8° anode, 3.0 mm Al + 0.9 mm Ti flat filtration (the Vectron-family stack), the CatSim large-body bowtie, 1.4 mm crystal and zero electronic noise. The 0.88 mm tube-B z-offset is applied by shifting the phantom.

Siemens SOMATOM Definition Flash

Second-generation dual source (2008). Two STRATON MX P tubes and two UFC detectors, 95° apart. The detector efficiency is its own Monte Carlo table, UFC_FLASH_MC_EFFICIENCY_LUT (detector_material = :ufc_flash), not the Force's: same Gd₂O₂S material, thinner crystal. The sourced specification follows; unpublished items are listed as assumptions.

Identification

Gantry geometry

Detector

Tubes and filtration

Acquisition

The model

scanner = BS.EICTScanner(                     # tube/detector A, used for both tubes
    source_to_isocenter = 595.0, source_to_detector = 1085.6,
    detector_rows = 64,          detector_cols = 736,
    detector_row_size = 0.6,     detector_col_size = 0.70473,
    detector_shape = :arc,
    focal_spot_width = 0.7,      focal_spot_length = 0.7,  target_angle = 7.0,
    gantry_rotation_time = 0.5,  scan_diameter = 500.0,    gantry_aperture = 780.0,
    flat_filter_material = :aluminum, flat_filter_thickness = 8.4,
    bowtie_filter = :large_body,                 # assumption: profile unpublished
    detector_material = :ufc_flash,              # the Flash table, never :ufc
    detector_depth = 1.0,
    fill_factor_row = 0.9,       fill_factor_col = 0.9,
    electronic_noise = 1500,     # e⁻, Stellar DAS floor (pre-2011 builds ≈ 3500)
    detection_gain = 10.0,
)
# dual energy: 80 kV on tube A, Sn140 on tube B; a full 1152-view rotation each, both currents
# scaled by DOSE_SCALE to the physical scan's CTDIvol (10.01 mGy)
protocol_low  = BS.CTProtocol(kVp = 80, mA = 470.0 * DOSE_SCALE, views = 1152, rotation_time = 0.5,
                              collimation_mm = 4.8, anode_angle = 8)
protocol_high = BS.CTProtocol(kVp = 140, mA = 182.0 * DOSE_SCALE, views = 1152, rotation_time = 0.5,
                              collimation_mm = 4.8, anode_angle = 8,
                              additional_filters = [("Sn", 0.4)])

The heel effect is off and every view integrates over its arc (view_samples = 5). Notebook 12 also runs the regular dual-power mode, 120 kV on both tubes at 420 mA × DOSE_SCALE each, whose two independent chains reduce noise by √2. The bundled spectra reach 140 kVp, the Flash's top kV, so the clinical pair runs without substitution; they come at 8° and 10° anodes, so the published 7° is run as 8°.

Modelling assumptions

Research geometries

Two notebooks use generic geometries and leave the rest at the defaults: the titanium-implant study (a compact 16-row scanner) and the helical study (a 256 × 0.625 mm, 16 cm wide-cone volume scanner, scanned with 20 mm collimation at pitch 1.0).

scanner_implant = BS.EICTScanner(            # notebook 10
    source_to_isocenter = 540.0, source_to_detector = 950.0,
    detector_rows = 16, detector_cols = 512,
    detector_row_size = 1.0, detector_col_size = 1.0,
    detector_material = :lumex, detector_depth = 3.0,
)
scanner_helical = BS.EICTScanner(            # notebook 11
    source_to_isocenter = 541.0, source_to_detector = 949.0,
    detector_rows = 256, detector_cols = 512,
    detector_row_size = 0.625, detector_col_size = 1.0,
)
helical = BS.CTProtocol(kVp = 120.0, mA = 200.0, views = 360, rotation_time = 0.5,
                        collimation_mm = 20.0, pitch = 1.0, n_rotations = 16)

Modelling a new scanner

What a new system needs, in three tiers.

Check a protocol against the hardware with validate_protocol(protocol, scanner), and the detector rows an axial reconstruction needs with required_axial_detector_rows(scanner; fov_cm, z_cm).