Computed tomography is fundamentally different from conventional radiography. Instead of collapsing three-dimensional anatomy onto a flat detector, CT measures X-ray attenuation from hundreds of angles and mathematically reconstructs cross-sectional images. This article covers the essential physics every CT technologist needs to know.
The gantry of a CT scanner is the circular structure that houses the imaging components. Understanding its anatomy is essential because every major CT physics concept ties back to how these components work together.
Mounted on the rotating gantry and powered through slip-ring technology, the tube makes repeated rotations around the patient. Tube ratings and heat-management designs vary by scanner and protocol.
Detector elements are arranged across the fan angle and in rows along the z-axis. Physical rows, active data channels, acquired sections, and reconstructed images are related but are not interchangeable counts; detector configuration and reconstruction determine section thickness and interval.
Transfers electrical power and data across stationary and rotating gantry components, allowing repeated continuous rotation without winding cables. Slip-ring systems enabled clinical helical/spiral CT.
Moves smoothly through the gantry aperture during helical scans. Table speed, combined with beam collimation, determines the pitch.
The X-ray tube rotates around the patient while a fan beam—or a wider cone-shaped beam in multidetector systems—passes through the body. Opposing detectors sample transmitted intensity from many angles. The system converts these measurements to attenuation projections and reconstructs cross-sectional images, commonly with filtered back projection, iterative reconstruction, or a validated deep-learning reconstruction method.
Each reconstructed voxel is assigned a Hounsfield unit (HU), also called a CT number; a displayed pixel represents the voxel value for that image section. HU describes X-ray attenuation relative to water, not mass density alone. Values can support tissue characterization, but must be interpreted as ranges because beam energy, material composition, patient size, calibration, reconstruction kernel/algorithm, partial-volume effects, artifacts, and iodinated contrast can change a measurement.
▸ Reference scale: water is defined as 0 HU and air is approximately −1000 HU. Displayed limits and measured material values vary; dense bone may exceed +1000 HU.
| Tissue | Hounsfield Units (HU) |
|---|---|
| Air (reference) | Approximately −1000 |
| Aerated lung | Often about −900 to −500 |
| Adipose tissue | Often about −190 to −30 |
| Water (reference) | 0 |
| CSF / simple fluid | Often near water (roughly 0 to +20) |
| White matter | Often about +20 to +30 |
| Gray matter | Often about +30 to +45 |
| Muscle / soft tissue | Often about +30 to +60 |
| Acute clotted blood | Often about +50 to +100 |
| Cortical bone | Typically several hundred to more than +1000 |
| Iodine or metal | Highly protocol/material dependent; may exceed the displayed HU range |
The formula for calculating a CT number is:
HU = 1000 × (μtissue − μwater) / μwater
where μ is the linear attenuation coefficient of the tissue.
Never diagnose from a single memorized HU cutoff without the protocol and clinical context. Measure a sufficiently large, homogeneous region of interest while avoiding edges and partial volume, and compare the appropriate phase/series. Dual-energy and spectral CT can provide additional material information, but conventional HU remains energy dependent.
Reconstructed CT data can contain many more values than a display can present as visibly distinct grays at once; stored bit depth and HU range vary by system and image format. Windowing maps a selected HU interval to the available display grayscale.
The total range of HU values displayed as shades of gray. Narrow window = high contrast (good for subtle differences). Wide window = low contrast (good for seeing many tissue types at once).
The center HU value of the window. This is the value that maps to middle gray. Adjusting the level shifts which tissue appears brightest.
| Preset | Window Width | Window Level | Best For |
|---|---|---|---|
| Lung | 1500 | −500 | Pulmonary parenchyma, airways |
| Soft Tissue (Abdomen) | 400 | +40 | Liver, spleen, kidneys, pancreas |
| Brain | 80 | +35 | Differentiating gray/white matter |
| Bone | 2000 | +500 | Osseous structures, fractures |
| Subdural / Brain Blood | 150 | +40 | Acute hemorrhage detection |
| Angiography | 600 | +200 | Contrast-filled vessels |
These values are illustrative starting points, not universal standards. Presets and optimal adjustments vary by vendor, monitor, anatomy, reconstruction, and diagnostic task.
Example: Under the usual linear window convention, a soft-tissue window of WW 400/WL 40 has nominal limits near −160 and +240. Values below the lower limit map to black and values above the upper limit map to white; exact endpoint mapping and the number of displayed gray steps depend on implementation and display.
Before the 1990s, CT scanners used axial (step-and-shoot) mode: rotate, stop, table advances, repeat. Helical (spiral) CT changed everything by enabling continuous rotation with simultaneous table motion.
For helical multidetector CT (MDCT), pitch is the ratio of table movement per rotation to total nominal acquired beam collimation. Total collimation is the number of simultaneously acquired data channels (N) multiplied by the nominal width of each channel (T). In older single-detector helical CT, pitch used nominal slice thickness in the denominator.
MDCT formula: Pitch = table travel per rotation / (N × T)
| Pitch Value | Type | Effect |
|---|---|---|
| < 1.0 | Denser helical sampling | More overlapping acquisition data; usually slower coverage. Dose rises if other exposure settings are held fixed, but tube-current modulation may alter this relationship. |
| 1.0 | Table travel equals total nominal collimation | A geometric reference—not a guarantee of contiguous reconstructed sections. |
| > 1.0 | Sparser helical sampling | Faster z-axis coverage; may affect noise, artifacts, and z-resolution. Dose falls only if other factors are not automatically compensated. |
Useful pitch ranges are scanner- and task-specific. Cardiac systems may use low-pitch retrospectively gated helices or specialized high-pitch modes; brain perfusion is commonly axial/cine, shuttle, or volume acquisition rather than a routine pitched helix. Pediatric and trauma protocols require protocol-specific optimization—not a universal “high-pitch = low-dose” rule.
Acquisition collimation, reconstructed slice thickness, and reconstruction interval are different settings. Thinner reconstructed sections improve z-axis detail and reduce partial-volume averaging, but each image contains fewer detected photons and is noisier when acquisition and reconstruction method are otherwise unchanged. Thicker sections reduce noise by averaging more data but can obscure small structures. Overlapping reconstruction intervals improve reformats and 3D sampling without increasing radiation dose when reconstructed from the same raw acquisition; acquiring additional overlapping data can increase dose.
| Artifact | Cause and practical response |
|---|---|
| Motion | Patient, respiratory, cardiac, or bowel motion causes blur/streaks; use communication, immobilization, appropriate timing/gating, and the shortest suitable acquisition. |
| Beam hardening / photon starvation | Preferential removal of low-energy photons or too few transmitted photons can create cupping or streaks, especially through dense anatomy; positioning, protocol selection, filtration, correction software, and sometimes more appropriate exposure help. |
| Partial volume | Different materials averaged within one voxel produce an intermediate CT number; thinner sections and careful ROI placement reduce it. |
| Metal | Severe attenuation, scatter, beam hardening, and incomplete data cause streaks; optimized acquisition, spectral techniques, and validated metal-artifact reduction may help but can introduce new artifacts. |
| Ring / cone-beam / helical | Detector calibration errors, cone geometry, or interpolation/sampling can create characteristic patterns; correct centering, calibration/service, and suitable acquisition/reconstruction are important. |
CT commonly delivers more radiation than a single conventional radiographic exposure, although dose varies widely by examination and protocol. Dose monitoring and optimization are therefore important. Three useful output/estimate metrics are:
An index of scanner radiation output for a protocol, representing average dose in a standardized 16-cm or 32-cm PMMA phantom and incorporating pitch for helical scans. It is not the dose to an individual patient.
CTDIvol × irradiated scan length. It characterizes output integrated along z; it is not absorbed energy or an individual patient's dose. Region-specific coefficients can give a rough population-level effective-dose estimate, with substantial limitations.
CTDIvol multiplied by a size-conversion factor, preferably based on water-equivalent diameter and the correct phantom. SSDE better reflects patient size, but remains an estimate of average absorbed dose in the scanned region—not organ dose or effective dose.
Effective dose (mSv) is a radiation-protection quantity for comparing typical population exposures; it is not a patient-specific risk or dose. DLP × k estimates depend on anatomic region, age/model, and coefficient set and should not be presented as precise individual values.
Use a justified, task-specific protocol and scan only the required range and phases. Center the patient, select appropriate kV, quality-reference mAs/noise target, rotation time, pitch, and collimation, and use correctly configured tube-current/organ-dose modulation plus suitable iterative or deep-learning reconstruction. In children and small adults, size protocols rather than merely copying adult settings. Avoid repeats and unnecessary multiphase acquisitions. Image quality must remain adequate for the diagnostic task—optimization is not simply choosing the lowest dose.
Projection data must be reconstructed into viewable images. Reconstruction families include analytical filtered back projection, iterative methods, and newer data-driven/deep-learning methods.
Commercial deep-learning reconstruction can suppress noise and improve selected image-quality measures, but “AI” does not guarantee a fixed dose saving or superior diagnosis. Performance depends on training, implementation, dose level, anatomy, and task; protocols should be validated locally and images reviewed for altered texture, bias, or loss/fabrication of subtle features.
Iodine increases attenuation, especially at lower photon energies and near its K-edge, so enhancement depends on iodine concentration/delivery, tube potential, scan timing, patient circulation and size, and reconstruction. Contrast-enhanced tissue has no single HU range. Scan phase and protocol must accompany any HU interpretation. Contrast administration itself does not create X-rays, but a contrast-enhanced examination may involve additional acquisitions; radiation dose is determined by those scan parameters and phases.
Want to compare CT with other modalities? Read CT vs MRI: When to Use Which or explore our CT Scan modality overview.
Try these ARRT-style multiple choice questions based on this article. Click an option to check your answer — correct answers turn green, wrong ones turn red.