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

Computed Tomography Data Path and Signal Preprocessing

• The progression of signal acquisition and reconstruction in conventional computed tomography follows a sequential digital pipeline:
  • Detectors acquire transmitted x-ray photons and convert physical interactions into analog electrical signals.
  • Analog-to-digital converters (ADCs) digitize these signals into measurement data (scan data).
  • The computer stores the digitized projections as planar raw data.
  • Preprocessing algorithms correct the raw measurements for physical degrading factors before reconstruction.
  • Convolution is applied to the corrected raw data using a mathematical filter (reconstruction kernel).
  • Convoluted data undergoes back projection.
  • The final reconstructed matrix is rendered as the actual CT image.

• Key preprocessing corrections executed prior to mathematical reconstruction include:
  • Scatter radiation correction: compensates for Compton interactions that strike detectors outside the primary ray paths.
  • Defective detector calibration: normalizes non-uniform detector sensitivities and accounts for unresponsive channels to prevent ring artifacts.
  • Beam hardening correction: adjusts attenuation profiles to prevent cupping artifacts and edge-darkening caused by preferential absorption of low-energy photons in polychromatic x-ray beams.

Image Reconstruction and Filtered Back Projection

• Reconstruction fundamentals:
  • Back projection is the mathematical technique that projects attenuation profiles across the image matrix at angles corresponding to the acquisition angles.
  • Unfiltered simple back projection produces a characteristic $1/r$ star blur artifact (classical star blur) caused by the constructive superposition of ray sum lines beyond the true boundaries of the object.
  • Analytic reconstruction (filtered back projection, or FBP) applies a digital convolution filter to ray profiles before summation to cancel positive blurring trails with negative side lobes.

• Mathematical foundations of projection measurements:
  • Each individual ray sum represents the total attenuation of an x-ray beam along a line through an array of voxels:     $I = I_0 e^{-\sum_{i=1}^n \mu_i \Delta x}$     where $I_0$ is incident intensity, $I$ is transmitted intensity, $\mu_i$ is the linear attenuation coefficient of voxel $i$, and $\Delta x$ is the path length through the voxel.
  • A standard clinical cross-sectional CT slice collects approximately 1500 distinct angular projections to reconstruct the attenuation distribution without undersampling artifacts.

Digital Convolution Kernels and Spatial Frequency

• Convolution filters alter image characteristics by acting on the spatial frequency domain:
  • Spatial frequency describes the rate of brightness or density change per unit distance across anatomy.
  • High spatial frequency boundaries involve abrupt transitions in attenuation over short distances, such as the interface between cortical bone and adjacent soft tissue (e.g., femur cortex to thigh musculature).
  • Low spatial frequency regions exhibit gradual attenuation changes across homogeneous parenchymal tissue, such as between the liver and spleen.

• High-pass (sharpening / high-contrast / edge-enhancing) algorithms:
  • Suppress or block low spatial frequencies while transmitting high spatial frequencies.
  • Produce pronounced edge definition and maximize spatial resolution.
  • Introduce increased statistical image noise and create a shorter contrast scale.
  • Utilized for high-contrast structures including the temporal bones, internal auditory canals (IAC), musculoskeletal exams, and fine pulmonary interstitium.

• Low-pass (smoothing / low-contrast / noise-reducing) algorithms:
  • Suppress or block high spatial frequencies while transmitting low spatial frequencies.
  • Reduce image graininess (quantum mottle) and optimize low-contrast detectability.
  • Blur sharp boundaries and decrease high-contrast spatial resolution.
  • Clinical applications encompass soft-tissue parenchymal imaging of the brain, abdomen, pelvis, and mediastinum.

Helical and Multi-Detector CT Reconstruction

• Continuous table translation in helical/spiral scanning causes acquired projections to be non-planar, requiring z-axis interpolation:
  • Linear interpolation estimates attenuation values at a specific planar slice position using data collected before and after the chosen table position.
  • 360-degree linear interpolation uses data points separated by $360^\circ$ of rotation, providing lower image noise but wider effective slice sensitivity profiles (SSP).
  • 180-degree linear interpolation uses conjugate rays separated by $180^\circ$, improving z-axis spatial resolution and reducing effective slice thickness broadening at the expense of slightly higher noise.

• Cone beam geometry:
  • Multi-detector CT (MDCT) systems using wide beam collimation (greater than $20\,\text{mm}$ along the z-axis) produce cone-shaped rather than fan-shaped beams.
  • Standard fan-beam reconstruction introduces cone-beam artifacts (distortion and shading) for detector arrays acquiring more than 4 slices per rotation.
  • Dedicated three-dimensional cone beam reconstruction algorithms (such as the Feldkamp-Davis-Kress or FDK algorithm) account for ray obliquity across outer detector rows.

Digital Image Matrix, Pixels, and Voxels

• Two-dimensional picture elements (pixels):

  • An image matrix comprises a rectilinear grid of rows and columns containing discrete picture elements.
  • Standard acquisition matrices measure $512 \times 512$ ($262{,}144$ pixels) or high-resolution $1024 \times 1024$ ($1{,}048{,}576$ pixels).
  • Increasing matrix dimensions for a fixed field of view reduces pixel dimension, improving spatial resolution but increasing computational reconstruction time.

• Mathematical determination of pixel size:   $\text{Pixel Size} = \frac{\text{Display Field of View (DFOV)}}{\text{Matrix Dimension}}$

  • Example: A $400\,\text{mm}$ DFOV displayed on a $512 \times 512$ matrix:     $\text{Pixel Size} = \frac{400\,\text{mm}}{512} = 0.78\,\text{mm}$

• Three-dimensional volume elements (voxels):

  • A voxel incorporates the two-dimensional pixel area along with depth, where depth corresponds to the reconstructed slice thickness ($z$-axis dimension).
  • Voxel volume calculation:     $\text{Voxel Volume} = \text{Pixel Width} \times \text{Pixel Height} \times \text{Slice Thickness}$
  • Voxel dimensions are calculated as:     $\text{Voxel Dimension} = \frac{\text{DFOV} \times \text{Slice Thickness}}{\text{Matrix Size}}$
  • Example: For a $400\,\text{mm}$ DFOV, $4\,\text{mm}$ slice thickness, and $512$ matrix:     $\frac{400\,\text{mm} \times 4\,\text{mm}}{512} = 3.1\,\text{mm}$
  • An isotropic voxel possesses identical dimensions on all three axes ($x = y = z$), eliminating stair-step artifact degradation when images are reformatted in coronal, sagittal, or oblique planes.

• Bit depth dictates digital dynamic range:
  • Represents the number of binary bits allocated to each pixel to encode attenuation values.
  • The total number of distinct grayscale or brightness shades available is given by $2^n$, where $n$ is the bit depth (e.g., 12-bit CT data encodes $2^{12} = 4096$ discrete shades of gray).

Spatial Resolution and Modulation Transfer Function

• Metrics and evaluation of spatial resolution:
  • Spatial resolution defines the capability of an imaging system to resolve two high-contrast objects positioned in close proximity.
  • Measured experimentally using line-pair phantom inserts containing alternating radio-opaque bars and radiolucent spaces.
  • Expressed quantitatively as line pairs per millimeter ($\text{lp/mm}$) or line pairs per centimeter ($\text{lp/cm}$).
  • Most modern clinical CT scanners can resolve structural details down to $0.3\,\text{mm}$.

• Formula for converting line-pair frequency to resolved object size:

  • For an imaging system capable of resolving $16\,\text{lp/cm}$:     $16\,\text{lp/cm} \times 2 = 32\,\text{lines/cm}$$\frac{10\,\text{mm}}{32\,\text{lines}} = 0.3125\,\text{mm}$

• Quantitative descriptors of spatial resolution:

  • Point Spread Function (PSF): Describes the system response to an infinitely narrow point object.
  • Line Spread Function (LSF): Describes the system response to an infinitely thin line structure.
  • Edge Response Function (ERF): Describes the system response to a sharp step interface between two homogeneous zones.
  • Modulation Transfer Function (MTF): The Fourier transform of the PSF/LSF that measures the ratio of output contrast to input contrast as a function of spatial frequency, where an MTF of $1.0$ indicates complete fidelity and $0.0$ represents zero image contrast transfer.

• Physical and technical determinants of spatial resolution:

  • Focal spot size: Smaller focal spot geometry diminishes geometric penumbra and improves spatial resolution.
  • Detector aperture dimensions: Smaller active detector elements sample the projection beam at higher spatial frequencies.
  • Geometry: Decreasing focal spot-to-patient distance or patient-to-detector distance reduces geometric magnification blur.
  • Reconstructed slice thickness: Thinner slices reduce partial volume averaging along the $z$-axis.
  • Subject contrast: Spatial resolution is maximized in regions of high inherent subject contrast (e.g., bone-to-air or bone-to-soft tissue interfaces) and lowest in soft-tissue boundaries (e.g., renal parenchyma vs. splenic parenchyma).

Scan Field of View vs. Display Field of View and Targeting

• Scan Field of View (SFOV):

  • Defines the circular diameter within the gantry opening from which physical detector channels record attenuation data.
  • Calibrated to match the anatomical boundaries of the patient.
  • The selected SFOV determines the total active detector span and calibration file application.

• Display Field of View (DFOV):

  • Designates the portion of raw data reconstructed and mapped onto the monitor display matrix.
  • Must always be equal to or smaller than the SFOV ($\text{DFOV} \le \text{SFOV}$).
  • A smaller DFOV reconstructs a smaller anatomical area over the fixed $512 \times 512$ matrix, resulting in smaller individual pixels and higher spatial resolution (true magnification/retrospective targeting).
  • In contrast, optical magnification (zoom) merely enlarges existing pixels without altering pixel dimensions or improving resolution.

Linear Attenuation Coefficients and CT Number Calculations

• Linear attenuation coefficients ($\mu$):

  • The fractional reduction in x-ray beam intensity per unit thickness of absorber, expressed in $\text{cm}^{-1}$.
  • Determined by the average effective atomic number ($Z$), physical tissue density ($\text{g/cm}^3$), material thickness, and incident photon energy spectrum ($\text{kVp}$).

• Reference linear attenuation coefficients at clinical effective beam energies ($\sim 70\,\text{keV}$ equivalent to $120\,\text{kVp}$ operation):

• Conversion of attenuation to Hounsfield Units (HU):
  • The CT number is calculated via the normalized formula:     $\text{CT Number (HU)} = \frac{\mu_{\text{tissue}} - \mu_{\text{water}}}{\mu_{\text{water}}} \times K$     where $K$ is the international scaling factor equal to $1000$.
  • Pure water is mathematically defined as $0\,\text{HU}$.
  • Standard air is defined as $-1000\,\text{HU}$.
  • Dense cortical bone is mapped to $+1000\,\text{HU}$ (with dense metals and hyper-dense objects extending past $+3000\,\text{HU}$).
  • Any measured tissue attenuation greater than $+100\,\text{HU}$ represents calcification or contrast enhancement.

• Typical clinical ranges of CT numbers:
  • Cortical bone: $+1000\,\text{HU}$
  • Acute clotted hemorrhage: $+65\,\text{HU}$ to $+95\,\text{HU}$
  • Cerebral gray matter: $+30\,\text{HU}$ to $+40\,\text{HU}$
  • Cerebral white matter: $+20\,\text{HU}$ to $+30\,\text{HU}$
  • Muscle: $+20\,\text{HU}$ to $+40\,\text{HU}$
  • Cerebrospinal fluid (CSF): $0\,\text{HU}$ to $+15\,\text{HU}$
  • Simple cyst fluid: $-10\,\text{HU}$ to $+20\,\text{HU}$
  • Adipose tissue: $-30\,\text{HU}$ to $-70\,\text{HU}$ (pure lipid approaches $-100\,\text{HU}$)
  • Air: $-1000\,\text{HU}$

Window Width and Window Level Fundamentals

• Psycho-physical principles of contrast manipulation:
  • While a reconstructed CT dataset contains thousands of discrete Hounsfield units, human visual perception distinguishes approximately 20 to 60 gray shades.
  • Windowing maps a specific clinical segment of CT numbers to the full display range from black (minimum optical density) to white (maximum optical density).

• Window Width (WW):

  • Defines the total range of CT numbers mapped across the grayscale display.
  • Directly controls image contrast:
  • As WW increases, contrast decreases (longer grayscale, displaying more subtle transitional shades).
  • As WW decreases, contrast increases (shorter grayscale, moving rapidly between black and white).
  • All CT numbers above the upper window threshold appear pure white; all numbers below the lower window threshold appear pure black.
  • The lower and upper window bounds are given by:     $\text{Lower Limit} = \text{WL} - \frac{\text{WW}}{2}$$\text{Upper Limit} = \text{WL} + \frac{\text{WW}}{2}$

• Window Level (WL) or Window Center (WC):

  • Specifies the midpoint CT number of the chosen Window Width.
  • Controls overall display brightness:
  • Increasing the Window Level darkens the image (requires higher HU values to produce visible brightness).
  • Decreasing the Window Level brightens the image (allows lower HU values to appear gray and white).
  • The level should generally match the average CT number of the target organ or pathological process being evaluated.

• Standard clinical window settings:

Image Acquisition Parameters and System Performance

• Tube current-time product ($\text{mAs}$):

  • Dictates the total quantity of x-ray photons emitted per rotation.
  • Higher $\text{mAs}$ increases detector photon fluence, decreasing statistical quantum noise (quantum mottle) and improving low-contrast resolution.
  • Patient radiation dose is directly proportional to $\text{mAs}$.
  • When scanning speed must be accelerated to minimize motion artifacts, the tube current ($\text{mA}$) must be elevated proportionally to preserve total photon fluence and SNR.

• Tube potential ($\text{kVp}$):

  • Determines the maximum and mean energy of the polychromatic photon spectrum (typically operated between $100\,\text{kVp}$ and $140\,\text{kVp}$).
  • Directly influences beam penetrability, scatter-to-primary ratios, and tissue contrast.
  • Lowering $\text{kVp}$ increases photoelectric interactions for high-$Z$ elements (e.g., iodine contrast), dramatically elevating contrast conspicuity at the expense of higher noise.
  • Patient radiation dose varies exponentially with beam potential: approximately as $\text{Dose} \propto (\text{kVp})^2$.
  • Adjustment of $\text{kVp}$ does not alter physical geometric spatial resolution.

• Slice thickness and collimation:

  • In single-detector row CT, pre-patient collimators define the physical x-ray beam thickness, and the acquired slice thickness equals the reconstructed slice thickness.

• In multi-detector CT (MDCT), wide cone-beam collimation exposes multi-row detector channels simultaneously.
• The acquired slice thickness is dictated by the specific detector configuration selected (e.g., individual channel widths along the $z$-axis).
• Rebinning combines digital outputs from adjacent detector channels to synthesize thicker reconstructed slices from thin raw acquisitions.
• Reconstructed slice thickness can be equal to or greater than the acquired detector width, but can never be thinner than the minimum acquired detector channel.

• Partial volume averaging artifact:
  • Occurs when distinct tissues of diverging attenuation occupy the volume of a single voxel (frequently along organ margins or fine skeletal edges).
  • The computer assigns a single CT number to the pixel representing the weighted mathematical average of all materials within that voxel.
  • Thin acquisition slices reduce partial volume averaging, improving diagnostic accuracy for small pulmonary nodules, nondisplaced cortical fractures, and intracranial structures.
  • Thicker display slices exhibit reduced image noise and superior low-contrast detectability for broad parenchymal structures (such as liver and spleen) at the cost of partial voluming along boundary margins.

• Table increments, overlap, and pitch:
  • Table increment defines the longitudinal distance (in millimeters) traveled by the patient couch during one full $360^\circ$ rotation of the gantry.
  • Contiguous scanning occurs when table increment matches slice thickness.
  • Overlapping scans occur when the table increment is less than the slice thickness, reducing partial volume artifacts and optimizing 3D reformations.
  • Gapped scans occur when the table increment exceeds slice thickness, decreasing patient radiation dose at the cost of anatomical data omission.
  • Pitch in helical scanning is defined as:     $\text{Pitch} = \frac{\text{Table Movement per Rotation (mm)}}{\text{Total Collimated Beam Width (mm)}}$
  • Pitch values greater than $1.0$ reduce radiation dose and scan time, but slightly broaden the slice sensitivity profile (SSP).

Temporal Resolution and Dynamic Imaging

• Core definition and clinical significance:
  • Temporal resolution denotes the capability of an imaging system to freeze physiological motion and image rapidly moving structures without blur.
  • Critical in coronary CT angiography and cardiac functional evaluation.
  • If an acquisition has a temporal resolution of $250\,\text{ms}$, an individual projection set is recorded within a $250\,\text{ms}$ temporal window of the cardiac cycle (permitting 4 phases per beat at a heart rate of $60\,\text{bpm}$).
  • Modern dual-source CT platforms achieve temporal resolutions down to $42\,\text{ms}$, resolving coronary arteries during diastole without pharmacologic beta-blockade.

• Determinants of temporal resolution in CT:
  • Gantry rotation speed: Higher mechanical rotational velocities directly reduce the physical time required to collect a projection arc.
  • Reconstruction schemes: Segmental reconstruction combines data segments acquired across multiple consecutive heartbeats (half-scan or quarter-scan algorithms).
  • Electrocardiographic gating modalities:
  • Prospective ECG triggering: Activates the x-ray tube only during targeted quiescent diastolic cardiac phases, markedly reducing radiation exposure.
  • Retrospective ECG gating: Continuously irradiates throughout the cardiac cycle while logging the patient's ECG trace, permitting dynamic post-hoc phase selection and cine loop analysis at higher radiation dose levels.
  • Pitch factor: Lower pitch values ensure redundant angular data sampling during cardiac cycles.

Advanced Post-Processing and Three-Dimensional Visualization

• Multiplanar Reconstruction (MPR):
  • Reformats transaxial contiguous or overlapping slices into coronal, sagittal, oblique, or curved diagnostic projections.
  • Relies on volumetric datasets composed of isotropic voxels to eliminate geometric stair-step distortion along reformatted planes.

• Maximum Intensity Projection (MIP):
  • Traces mathematical projection rays through a 3D data volume, selecting and rendering only the voxel containing the highest CT attenuation number encountered along each ray path.
  • Excellent for visualizing contrast-opacified vascular trees in CT angiography (CTA) and tracking pulmonary vessels.
  • Disadvantages: Depth perception is entirely lost along the projection axis; higher-attenuation voxels visually occlude lower-attenuation structures behind them; vessels closer to the viewer and those farther away appear indistinguishable.

• Volume Rendering Technique (VRT):
  • Processes the entire 3D voxel volume rather than a subset.
  • Assigns individual opacity (transparency) ratings ($0\%$ to $100\%$) and RGB color values to specific attenuation intervals.
  • Produces an interactive 3D model that preserves structural depth perception, permitting visualization through translucent outer tissue layers to reveal underlying vascular and skeletal relationships.

• Systematic technical comparison between VRT and MIP:

• Shaded Surface Display (SSD):

  • Uses a thresholding algorithm to delineate anatomical boundaries (e.g., identifying all voxels above $+150\,\text{HU}$ as bone).
  • Generates an external 3D virtual shell based on simulated light ray reflection from the identified boundary.
  • Internal voxel attenuation data located beneath the defined surface boundary is discarded.

• Virtual Endoscopy (CT Colonography):

  • Uses volume-rendered perspective transformations within lumen-segmented hollow organs.
  • Generates a dynamic "fly-through" visualization simulating optic endoscopic navigation to detect colonic polyps and mucosal abnormalities.

Quantitative CT Analysis and Radiomics

• Clinical quantitative morphology:

  • Automated tools enable precise anatomical dimension tracking, such as evaluating aortic aneurysms or pulmonary hypertension via vascular caliber ratios.
  • Morphometric measurements: Ascending aorta diameter (normal threshold example: $38.6\,\text{mm}$) compared to main pulmonary artery diameter (e.g., $25.7\,\text{mm}$).

• Radiomics principles:

  • Radiomics involves extracting, mining, and analyzing high-throughput quantitative features (texture, shape, wavelet transformations) from medical images.
  • Yields phenotypic information imperceptible to visual inspection, facilitating objective lesion classification, oncologic staging, and therapeutic response prediction.

• Coronary Artery Calcium (CAC) Scoring:

  • Identifies non-contrast hyper-dense calcified plaques ($> +130\,\text{HU}$ with area $\ge 1\,\text{mm}^2$) within the coronary arterial tree.
  • Plaque calcification is quantified using the Agatston scoring method, which weights calcified plaque area by peak attenuation:
  • Attenuation factor of $1$: $130\,\text{HU}$ to $199\,\text{HU}$
  • Attenuation factor of $2$: $200\,\text{HU}$ to $299\,\text{HU}$
  • Attenuation factor of $3$: $300\,\text{HU}$ to $399\,\text{HU}$
  • Attenuation factor of $4$: $\ge 400\,\text{HU}$
  • The calculated Agatston score correlates directly with total atherosclerotic burden and cardiovascular event risk.

• Clinical reference dataset for Agatston calcium scoring evaluation:

Computed Tomography Data Path and Signal Preprocessing

• Digital Data Flow:
  1. Detectors: Convert transmitted x-ray photons into analog electrical signals.
  2. Analog-to-Digital Converters (ADCs): Digitize electrical signals into measurement/scan data.
  3. Preprocessing: Corrects raw data for beam hardening, bad detector channels, and scatter.
  4. Convolution: Applies a mathematical reconstruction filter (kernel) to raw data.
  5. Back Projection: Projects attenuation profiles onto the image matrix to generate the CT image.

Image Reconstruction and Filtered Back Projection

• Simple Back Projection: Adds attenuation profiles together across the image matrix, causing a classic $1/r$ star blur artifact.
• Filtered Back Projection (FBP): An analytic reconstruction technique applying a convolution filter with negative lobes to eliminate star blurring before summation, producing a sharp image.

Digital Convolution Kernels and Spatial Frequency

• Spatial Frequency: The rate of attenuation change per unit distance.
  • High Spatial Frequency: Sharp transitions across anatomy (e.g., cortical bone to soft tissue).
  • Low Spatial Frequency: Subtle, gradual transitions (e.g., liver parenchyma to spleen).
• Filter Types:
  • High-Pass (Bone / Edge-Sharpening Algorithm): Enhances spatial resolution and edge detail; increases image noise and reduces contrast.
  • Low-Pass (Soft Tissue / Smoothing Algorithm): Reduces noise and improves contrast resolution; softens sharp edges.

Helical and Multi-Detector CT (MDCT) Reconstruction

• Z-Axis Interpolation: Compensates for table motion during spiral scanning by estimating planar attenuation:
  • $180^\circ$ Linear Interpolation: Uses conjugate rays separated by $180^\circ$; improves z-axis spatial resolution and narrows effective slice thickness, with slightly higher noise.
  • $360^\circ$ Linear Interpolation: Uses data separated by $360^\circ$; produces lower noise at the cost of a wider slice sensitivity profile (SSP).
• Cone-Beam Reconstruction: Applied in MDCT when detector collimation exceeds $20\,\text{mm}$.

Digital Matrix, Pixels, and Voxels

• Matrix: Grid of rows and columns (typically $512 \times 512$ or $1024 \times 1024$). Larger matrices decrease pixel dimensions and improve spatial resolution.
• Pixel Size Formula:
  $\text{Pixel Size} = \frac{\text{DFOV}}{\text{Matrix Size}}$
• Voxel: 3D volume element combining pixel area with slice thickness ($z$-axis):
  $\text{Voxel Volume} = \text{Pixel Width} \times \text{Pixel Height} \times \text{Slice Thickness}$
• Isotropic Voxel: Dimensions are equal on all axes ($x = y = z$), preventing stair-step artifacts on multiplanar reformats.

Spatial and Temporal Resolution

• Spatial Resolution: Ability to distinguish two small, adjacent high-contrast objects.
  • Measured in line pairs per millimeter ($\text{lp/mm}$) or line pairs per centimeter ($\text{lp/cm}$).
  • Clinical systems resolve down to approximately $0.3\,\text{mm}$.
  • Evaluated via the Modulation Transfer Function (MTF).
• Temporal Resolution: Ability to freeze physiological motion (vital for cardiac imaging).
  • Dual-source scanners achieve speeds down to $42\,\text{ms}$.

Field of View (FOV) and Targeting

• Scan FOV (SFOV): Circular area inside the gantry actively read by detectors.
• Display FOV (DFOV): Raw data area mapped to the monitor ($\text{DFOV} \le \text{SFOV}$).
• Targeting / Retrospective Reconstruction: Reducing DFOV maps a smaller area onto the fixed $512 \times 512$ matrix, decreasing pixel size and improving spatial resolution (true magnification).

CT Numbers and Attenuation

• CT Number Calculation:
  $\text{CT Number} = \frac{\mu<em>{\text{tissue}} - \mu</em>{\text{water}}}{\mu_{\text{water}}} \times 1000$
• Standard CT Values (HU):
  • Bone: $+1000\,\text{HU}$
  • Muscle: $+50\,\text{HU}$
  • White Matter: $+45\,\text{HU}$
  • Gray Matter: $+40\,\text{HU}$
  • Blood: $+20\,\text{HU}$
  • Cerebrospinal Fluid (CSF): $0\,\text{HU}$ to $+15\,\text{HU}$
  • Water: $0\,\text{HU}$
  • Fat: $-100\,\text{HU}$
  • Lung: $-200\,\text{HU}$ to $-700\,\text{HU}$
  • Air: $-1000\,\text{HU}$

Window Width (WW) and Window Level (WL)

• Window Width (WW): Range of CT numbers assigned to the grayscale spectrum. Controls image contrast:
  • Narrow WW: High contrast, fewer gray shades (used for brain and soft tissue).
  • Wide WW: Low contrast, wide gray scale (used for bone and lung).
• Window Level (WL): Midpoint CT number of the window. Controls image brightness (increasing WL makes the image darker).
• Preset Reference:
  • Brain: WW $80\,\text{HU}$, WL $+40\,\text{HU}$
  • Chest: WW $400\,\text{HU}$, WL $+40\,\text{HU}$
  • Abdomen: WW $400\,\text{HU}$, WL $+60\,\text{HU}$
  • Lung: WW $1500\,\text{HU}$, WL $-400\,\text{HU}$ to $-600\,\text{HU}$
  • Bone: WW $1500\,\text{HU}$ to $2500\,\text{HU}$, WL $+300\,\text{HU}$ to $+500\,\text{HU}$

Technical Acquisition Parameters

• Tube Current-Time ($\text{mAs}$): Controls x-ray photon quantity. Higher $\text{mAs}$ reduces quantum mottle (noise); radiation dose is directly proportional to $\text{mAs}$.
• Tube Potential ($\text{kVp}$): Dictates beam energy and penetrability. Lowering $\text{kVp}$ increases photoelectric contrast for iodine; dose scales exponentially as $\text{Dose} \propto (\text{kVp})^2$. Adjusting $\text{kVp}$ does not change spatial resolution.
• Slice Thickness: Thin slices reduce partial volume averaging but increase image noise.

3D Post-Processing Techniques

• Multiplanar Reconstruction (MPR): Re-slices transaxial data into coronal, sagittal, or oblique planes; requires isotropic voxels to prevent stair-step artifacts.
• Maximum Intensity Projection (MIP): Projects only the highest attenuation voxels; ideal for contrast-enhanced vascular maps, but lacks 3D depth.
• Volume Rendering Technique (VRT): Uses all voxels within the 3D volume with variable opacities; maintains depth perception and displays multiple tissue classes simultaneously.
• Shaded Surface Display (SSD): Uses an attenuation threshold to construct a 3D surface model; discards internal voxel data.
• Virtual Endoscopy: Simulates intraluminal perspective for hollow organs (e.g., CT colonography).