7.4 Reconstruction Parameters — Slice Thickness, Interval & Interpolation

Why Reconstruction Parameters Are Tested

The final three leaf items in ARRT's Image Reconstruction subcategory (1.D.6-8) are reconstruction slice thickness, reconstruction interval, and interpolation. These are the parameters a technologist actually sets after a scan is acquired, and they generate exactly the kind of "given this raw dataset, what would you reconstruct" scenario question that recurs throughout Image Production and Procedures alike — especially anywhere thin-section, multiplanar reformat (MPR), or 3D post-processing work is required (Chapter 8), or wherever pitch and image quality intersect (Chapter 9's artifact content).

Reconstruction Slice Thickness vs. Acquisition Slice Thickness

It is essential to separate two similarly named but distinct concepts. Acquisition (detector) slice thickness is fixed at scan time by the detector row configuration and collimation — it cannot be changed after the patient leaves the table. Reconstruction slice thickness is chosen afterward, from the same raw helical (or volumetric) data, and can be set at or thicker than the acquired detector row width, but never thinner than it — a scanner cannot manufacture spatial resolution finer than what was physically acquired.

Worked example: A chest CT is acquired with 0.625 mm detector collimation. From that single raw dataset, the same scan can be reconstructed at 0.625 mm (thinnest possible, used for MPR and 3D rendering in Chapter 8), at 2-3 mm (a common "standard" viewing thickness that balances detail against noise and image count), or at 5 mm (fastest to review, lowest noise, but coarsest z-axis detail) — all without rescanning the patient or adding any dose, because reconstruction slice thickness is a post-processing choice applied to already-acquired raw data.

Reconstruction Interval

Reconstruction interval is the spacing between the centers of consecutive reconstructed images, and it is set independently of slice thickness. When the interval is smaller than the slice thickness, consecutive reconstructed images overlap — a practice called overlapping reconstruction. Overlapping reconstruction produces smoother multiplanar reformats and improves small-lesion conspicuity and measurement accuracy, because no anatomy falls into an unimaged "gap" between two non-overlapping slices. Critically, because this is applied to raw helical data that has already been acquired, overlapping reconstruction adds zero additional patient dose — the dose was determined entirely by the acquisition parameters (kVp, mA, pitch), not by how many overlapping images are later generated from that same raw dataset.

Worked example: A dataset acquired at 1 mm slice thickness is reconstructed with a 0.5 mm interval, producing 50% overlap between adjacent images — useful when a small pulmonary nodule or a subtle fracture line needs to be tracked continuously through the volume without being "split" between two adjacent non-overlapping slices.

Interpolation

Because helical acquisition data traces a continuous spiral relative to the patient rather than lying flat in any single axial plane, the scanner must mathematically estimate ("interpolate") what the attenuation data would have been at a precise axial location, using the surrounding helical data points collected just before and just after that position. Two general interpolation families are used:

• 360° linear interpolation — averages data collected one full gantry rotation apart, using a relatively wide window of source data.
• Z-filtering (weighted) interpolation — uses a narrower, more tightly weighted window of data closest to the desired z-position, improving z-axis (longitudinal) resolution at the cost of somewhat higher image noise compared to 360° interpolation.

Interpolation quality is directly linked to pitch (Section 7.1): as pitch increases, each z-position's interpolated estimate must draw on raw data collected across a wider table-travel distance and gantry angular range, which broadens the effective slice sensitivity profile and increases the risk of visible artifact on multiplanar reformats — including stair-step artifact along oblique or curved reformats, and, on wide cone-angle scanners at high pitch, the characteristic windmill artifact (spoke-like streaks radiating from high-contrast edges). This is the same underlying mechanism, viewed from the reconstruction side, as the overranging dose penalty discussed in Chapter 5 (viewed from the acquisition/dose side) — both are consequences of how helical data must be extended and estimated beyond the exact prescribed scan range.

Exam Scenario

A radiologist reviewing a routine 5 mm abdominal CT requests thin-section reformats to characterize a subtle liver lesion. Because the study was originally acquired with 0.625 mm detector collimation, the technologist can generate 1 mm reconstruction slice thickness with a 0.5 mm reconstruction interval (50% overlap) directly from the existing raw data — no additional scan, no additional dose. On a separate high-pitch trauma chest/abdomen/pelvis acquisition, a radiologist notes spoke-like streak artifact radiating from the sternum on coronal reformats; the technologist recognizes this as windmill artifact, a consequence of the wide cone angle and high pitch straining the interpolation algorithm's ability to estimate data cleanly at each reconstructed position.

Key Takeaways

• Acquisition (detector) slice thickness is fixed at scan time; reconstruction slice thickness is a post-scan choice from the same raw data and can never be thinner than the acquired detector width.
• Reconstruction interval controls overlap between consecutive images; overlapping reconstruction (interval < slice thickness) improves 3D/MPR quality and small-lesion detection at zero added dose.
• Interpolation (360° linear vs. z-filtering) estimates data at each axial position from the surrounding helical raw data; interpolation demand — and related artifact risk (stair-step, windmill) — rises with pitch.
• All three parameters can be applied and reapplied to the same raw dataset after the scan without rescanning the patient or adding dose, reinforcing why raw data retention (Section 7.3) matters operationally.