CT Physics Registry Review: Image Reconstruction, Quality, and Post-Processing

Comprehensive vocabulary flashcards covering computed tomography physics, including data acquisition paths, back projection algorithms, digital filtering, matrix physics, spatial and temporal resolution, window settings, and advanced 3D post-processing methods.

Filtered Back Projection (FBP)

An analytic image reconstruction process that applies a mathematical convolution filter (kernel) to raw projection data to eliminate classic star blur before summing attenuation profiles to form the final CT image.

Preprocessing

The preparatory data processing stage where detector measurement data is corrected for scatter, bad detectors, and beam hardening prior to generating the raw data file.

Convolution

The mathematical operation in CT image reconstruction of applying a filter or kernel to raw projection data to modify its spatial frequency characteristics.

Classical Star Blur

The characteristic radial streak artifact and smearing that occurs when simple back projection is performed without applying a convolution filter.

High-Pass Filter

A sharpening reconstruction algorithm (also known as a bone algorithm) that suppresses low spatial frequencies to enhance edges and spatial resolution at the cost of increased image noise and decreased contrast resolution.

Low-Pass Filter

A smoothing reconstruction algorithm (such as soft tissue, brain, and mediastinum algorithms) that suppresses high spatial frequencies to reduce image noise and improve low-contrast resolution.

Interpolation

A mathematical technique utilized in spiral/helical CT that estimates unknown data values from known data points on either side of a plane to compensate for continuous table motion.

$180^\top$ Interpolation

A spiral CT interpolation method that samples data points separated by $180^\top$ of gantry rotation, providing superior $z$-axis resolution compared to $360^\top$ interpolation.

Cone Beam Reconstruction

A specialized MDCT reconstruction algorithm required for wide beam collimations (greater than $20\,\text{mm}$) to handle beam divergence and acquire more than $4$ slices per rotation.

Pixel

A two-dimensional picture element arranged within a matrix of rows and columns, possessing dimensions of height and width.

Image Matrix

A two-dimensional grid of columns and rows of pixels into which slice data is divided (commonly $512 \times 512$ with $262{,}144$ pixels or $1024 \times 1024$ with $1{,}048{,}576$ pixels); larger matrices yield superior spatial resolution but require longer reconstruction times.

Pixel Size Formula

The formula used to determine pixel dimensions: $\text{Pixel size} = \frac{\text{DFOV}}{\text{Matrix Size}}$. For example, a $400\,\text{mm}$ DFOV over a $512$ matrix produces a pixel size of approximately $0.78\,\text{mm}$.

Spatial Resolution

The ability of an imaging system to resolve the smallest possible object, expressed in line pairs per millimeter ($\text{lp/mm}$) or line pairs per centimeter ($\text{lp/cm}$); most CT scanners can resolve objects down to $0.3\,\text{mm}$.

Voxel

A three-dimensional volume element possessing height, width, and depth, where depth corresponds to the slice thickness.

Bit Depth

The number of bits allocated to each pixel determining the number of gray shades or brightness levels that can be demonstrated, calculated as $2^{\text{bit depth}}$.

Scan Field of View (SFOV)

The circular anatomical acquisition diameter determined by the coverage of the X-ray fan beam over the detector array.

Display Field of View (DFOV)

The field of view of the displayed reconstructed image; selecting a smaller DFOV increases image display size (true magnification) and improves spatial resolution by sampling less tissue volume.

Targeting

A CT post-processing method that magnifies specific anatomical regions of interest by decreasing the DFOV without losing spatial resolution.

Linear Attenuation Coefficient ($\mu$)

A numerical value representing the rate at which an X-ray beam is attenuated per centimeter of tissue ($\text{cm}^{-1}$), determined by average atomic number, tissue density, thickness, and $\text{kVp}$.

CT Number Conversion Formula

The mathematical formula: $\text{CT Number} = \frac{\mu_t - \mu_w}{\mu_w} \times K$, where $\mu_t$ is the tissue attenuation coefficient, $\mu_w$ is the linear attenuation coefficient of water ($0.206\,\text{cm}^{-1}$), and $K$ is a constant equal to $1000$.

Hounsfield Unit (HU)

The standardized unit for CT numbers, calibrated so that bone is $+1000\,\text{HU}$, water is $0\,\text{HU}$, and air is $-1000\,\text{HU}$; values above $+100\,\text{HU}$ typically indicate calcification.

Window Width (WW)

The specific range of CT numbers allocated to the displayed grayscale, which controls image contrast; increasing WW decreases contrast (more shades of gray), whereas decreasing WW increases contrast (more black and white).

Window Level (WL)

The midpoint or central CT number of the selected window width, which controls image brightness; increasing WL darkens the image, while decreasing WL brightens it.

Brain Window Settings

Standard CT window parameters configured at a Window Width of $80\,\text{HU}$ and a Window Level of $40\,\text{HU}$ to differentiate gray and white matter.

Lung Window Settings

Standard CT window parameters configured at a wide Window Width of $1500\,\text{HU}$ and a Window Level of $-400\,\text{HU}$ (or $-600\,\text{HU}$) to visualize pulmonary parenchyma.

Bone Window Settings

Standard CT window parameters configured at a wide Window Width of $1500\,\text{HU}$ and a Window Level of $300\,\text{HU}$ to resolve fine bony detail.

Rebinning

The process in MDCT of combining signals from smaller adjacent detector rows along the $z$-axis to synthesize a thicker acquired slice.

Partial Volume Averaging

An artifact that occurs when anatomically distinct tissues with different attenuation coefficients occupy the same voxel, causing the computer to assign an average CT number to the corresponding pixel.

Table Increment

The distance the patient table travels during scanning; an increment equal to slice thickness produces contiguous slices, less movement causes overlap (reducing partial volume averaging), and greater movement causes gaps.

Temporal Resolution

The ability of a scanner to freeze motion and resolve fast-moving objects (such as the heart), capable of reaching values as low as $42\,\text{ms}$ on modern dual-source scanners.

Quantum Mottle

A grainy image noise artifact produced by photon starvation or photon insufficiency when an inadequate number of X-ray photons reach the detector array.

Multiplanar Reconstruction (MPR)

A post-processing technique that uses contiguous helical data to reconstruct non-axial viewing planes (coronal, sagittal, or oblique), optimized when voxels are isotropic.

Isotropic Voxel

A three-dimensional voxel possessing identical measurements on all three dimensions ($x = y = z$), enabling high-quality multiplanar reformations without loss of spatial resolution.

Maximum Intensity Projection (MIP)

A 3D post-processing technique that casts rays through a data volume and displays only the voxels with the highest attenuation values along each ray path, highlighting contrast-enhanced vessels while losing depth perception.

Volume Rendering Technique (VRT)

A computationally intensive 3D post-processing technique that incorporates all voxels within a dataset, assigning color and transparency to preserve depth perception and display multiple tissue types simultaneously.

Shaded Surface Display (SSD)

A 3D surface modeling method that calculates simulated light reflections off object surfaces based on preselected CT number threshold boundaries.

Radiomics

The practice of extracting, analyzing, and interpreting high-throughput quantitative data and image features from medical images that are imperceptible to the human eye to assist in disease diagnosis and prognosis.

Agatston Score

A standardized quantitative scoring metric utilized in cardiac CT calcium scoring to quantify coronary artery calcification volume and density.