Beam Geometry, Slice Collimation & Data Sampling

---

Architectural Evolution: Transaxial Fan Beam to Cone Beam Geometry

The 2D Fan Beam Era

In single-slice CT and early dual-slice systems, the x-ray beam emerged from the tube focal spot and expanded in the transaxial ($xy$) plane with a fan angle of $30^\circ\text{--}50^\circ$, matching the circular opening of the gantry bore. Along the longitudinal ($z$) axis, however, pre-patient collimators narrowed the beam to a thin, nearly parallel slab. Standard 2D reconstruction algorithms (such as conventional 2D Filtered Back Projection [FBP]) operated on the mathematical assumption that every measured x-ray path traveled within a flat plane strictly perpendicular to the axis of gantry rotation ($z$-axis).

The Multi-Detector Cone Beam Architecture

With the introduction of multi-detector CT (MDCT) systems sporting 16, 64, 128, 256, and 320 detector rows, the longitudinal coverage of the detector array expanded dramatically, spanning from $20\text{ mm}$ up to $160\text{ mm}$ at isocenter in a single rotation. The x-ray beam could no longer be treated as a parallel planar fan. Instead, the beam forms a three-dimensional pyramid: the cone beam.

Longitudinal divergence follows the actual geometry

The cone angle depends on longitudinal coverage and source-to-detector geometry. Increasing physical coverage generally increases longitudinal divergence, but a marketed slice count alone does not determine the angle. Different detector configurations can produce different coverage for the same displayed slice count. Use the system's actual geometry rather than one universal angle table for 4-, 16- or 64-slice labels.

Cone-Beam Artifacts

When x-rays strike the outer detector rows at significant oblique angles, they traverse anatomical structures obliquely rather than perpendicularly. If these oblique projection data are reconstructed using standard 2D FBP, two severe artifacts emerge:

• Volume Distortion & Non-Planar Blurring: Structures situated away from the central plane ($z = 0$) appear blurred, distorted, and elongated along the longitudinal axis.
• Cone-Beam Streaking & Shading: Dark shading and bright streaks radiate from high-contrast interfaces, particularly near the skull base, ribs, and vertebral bodies.

Reconstruction must match the geometry

Multidetector rows acquire rays with longitudinal divergence. Appropriate cone-beam or helical reconstruction accounts for that geometry. Feldkamp–Davis–Kress is an approximate cone-beam method commonly associated with circular cone-beam systems; it is not the universal exact algorithm of every diagnostic helical MDCT scanner. Vendors use different rebinning, analytical and iterative methods. The exam principle is that a flat fan-beam assumption becomes insufficient as longitudinal coverage and divergence increase.

Collimation Mechanics: Pre-Patient vs. Detector Collimation

In computed tomography, the definition of "collimation" encompasses two distinct hardware systems: pre-patient mechanical collimation and detector electronic binning.

Pre-Patient Collimation Architecture

The pre-patient collimator assembly is mounted directly at the exit port of the x-ray tube housing:

• Physical Composition: Motorized, high-precision tungsten collimator jaws or shutters.
• Primary Function: It shapes the primary x-ray beam along the longitudinal ($z$) axis before radiation enters the patient's body.
• Determines Total Beam Width ($W$):

$$W = N \times T$$

where $N$ is the number of active data channels and $T$ is the nominal detector width at isocenter.

• Radiation Protection Mandate: By physically blocking off-axis radiation, pre-patient collimators restrict the exposed beam according to the acquisition, including its penumbra and endpoint requirements, minimizing unnecessary patient dose and reducing Compton scatter production.

Post-Patient / Pre-Detector Collimation

Positioned directly in front of the detector array, the pre-detector collimator consists of thin tungsten or lead grids (anti-scatter septa) aligned with the focal spot. An important function is to intercept scattered radiation exiting the patient at oblique angles, preventing Compton scatter from degrading low-contrast resolution.

Acquisition collimation and reconstructed thickness

Detector configuration and supported electronic grouping help establish the acquired longitudinal sampling. The nominal acquisition width is commonly expressed as N × T, with N active acquisition channels and T their nominal width at isocenter. Prepatient collimation defines the exposed beam, including its edge profile.

Reconstructed thickness is chosen from the data the system actually acquired and its supported reconstruction methods. It is not determined solely by electronic binning, and choosing an extremely thin number cannot manufacture independent detail finer than the acquisition supports. Multiple reconstructed thicknesses and intervals may be obtained from one sufficiently fine acquisition without repeating the exposure.

Geometric Radiation Dose Penalties in MDCT: Overbeaming and Overranging

Overbeaming describes exposed beam width beyond the useful detector region, while overranging describes extra helical exposure beyond the planned image range. They are especially relevant in multidetector CT, but penumbra and helical endpoint requirements are not concepts exclusive to MDCT.

1. Overbeaming Mechanics & Dose Penalty

• Physical Origin: The x-ray tube focal spot is not an infinitesimal point source; it has finite dimensions (typically $0.6\text{ mm} \times 0.7\text{ mm}$ for small focus, $1.1\text{ mm} \times 1.2\text{ mm}$ for large focus). Because of this finite focal spot size, the collimated beam exhibits an optical penumbra (region of partial x-ray illumination) at its outer margins flanking the central umbra (region of uniform full intensity).
• The Overbeaming Requirement: To ensure that the useful detector region is appropriately illuminated without signal drop-off at the edges, the pre-patient collimator must be opened slightly wider than the active detector array. The radiation penumbra that extends beyond the active detector elements is termed overbeaming.
• Wasted Dose: Overbeaming radiation penetrates the patient's body (contributing to absorbed dose) but falls completely outside the active detector elements, contributing zero diagnostic data to image reconstruction.
• Geometric Dose Efficiency:

$$\text{Geometric Efficiency} = \frac{\text{Active Detector Array Width at Isocenter}}{\text{Total Collimated Beam Width (including Penumbra)}} \times 100\%$$

For a hypothetical active width of 5 mm and total exposed width of 6.5 mm, the simple useful-width ratio is 5/6.5 = 76.9%. With 40 mm active and 41.5 mm exposed, it is 40/41.5 = 96.4%. These illustrate a smaller relative edge penalty for a wider beam under the stated identical excess width. They are not measurements of a named scanner's total patient dose; actual profiles and geometry determine the result.

Overranging and dynamic longitudinal collimation

Helical reconstruction can require measurements beyond the first and last displayed image positions. The additional longitudinal irradiation is overranging. Its extent depends on detector width, pitch, reconstruction and scanner design, so a single formula is not universally valid. The relative penalty can be larger in a short scan than in a long scan.

For an idealized length-only example, 40 mm of extra irradiation compared with a 50 mm planned range is 40/50 = 80% additional length; compared with 800 mm it is 5%. Those fractions are not automatically the patient's dose increase, because actual beam profiles and dynamic collimation matter. Dynamic z-collimation can reduce unnecessary endpoint exposure in supported systems, with the benefit depending on the protocol.

Data Sampling Theory & Nyquist Limits

Accurate image reconstruction without aliasing artifacts requires satisfying the Nyquist-Shannon Sampling Theorem: the sampling frequency ($f_s$) must be at least twice the highest spatial frequency ($f_{\max}$) present in the scanned object:

$$f_s \ge 2 f_{\max}$$

Angular Sampling

• Definition: The number of projection views acquired per $360^\circ$ gantry rotation (typically $1,000$ to $2,400\text{ views/rotation}$, corresponding to thousands of view samples per second).
• View Aliasing: If angular sampling is insufficient (too few projection views per rotation relative to gantry rotation speed and reconstruction matrix size), high-frequency signals fold back into the image. This manifests as fine, dense pinwheel streaks radiating from sharp, high-density structures (such as cortical bone edges or surgical clips).

Ray Sampling (Transaxial In-Plane Sampling)

• Definition: The number of individual detector measurements per projection view across the transaxial fan angle. It is governed by detector element spacing and physical width.
• Radial Aliasing: Insufficient ray sampling causes fine radial streaks and fundamentally limits transaxial ($xy$) in-plane spatial resolution.
• Quarter-Detector Offset: In third-generation CT scanners, shifting the detector array laterally by one-quarter of a detector element width ($1/4 \Delta d$) relative to the rotation axis interleaves ray samples acquired $180^\circ$ apart. This doubles the effective radial sampling density and can reduce ray-sampling aliasing.

Flying focal spots and interleaved rays

Supported focal-spot deflection can interleave transaxial or longitudinal ray measurements. It can improve sampling and reduce selected aliasing or helical artifacts, but does not guarantee elimination or manufacture unlimited fine detail. A quarter-detector offset can likewise interleave opposite-view samples to improve ray sampling; it is not a correction for detector-gain ring artifacts.

Worked geometry and sampling check

For 64 nominal acquisition channels each 0.625 mm wide, total nominal collimation is 40 mm. If table travel is 48 mm per rotation, beam pitch is 48/40 = 1.2. At a rotation time of 0.5 second, table speed is 48/0.5 = 96 mm/s. These calculations describe geometry and speed, not the reconstructed slice thickness or patient absorbed dose.

The Nyquist frequency for a sampling spacing Δd is 1/(2Δd). For 0.5 mm spacing, that is 1 cycle/mm or 10 cycles/cm. Actual spatial resolution may be lower because of focal spot, detector aperture, reconstruction and noise. Adequate sampling is necessary but not sufficient for resolving a clinical structure.