How CT Detectors and Hounsfield Units Work

How a CT scanner measures X-ray attenuation, corrects thousands of detector channels, and turns those measurements into Hounsfield Units that can distinguish air, fat, water, soft tissue, and bone

A CT detector does not actually see organs, blood, bone, or a tumor. It sees X-rays.

Back to Biomed Basics

What This Page Explains

This page covers:

The Simple Version

A CT detector measures the amount of X-ray energy reaching it after the beam passes through the patient. Tissue that attenuates more radiation allows fewer photons to reach the detector. Tissue that attenuates less allows more to pass. The scanner repeats these measurements from a huge number of angles as the gantry rotates, creating projection data that describes how strongly different paths through the body attenuated the beam.

Those raw detector measurements cannot be turned directly into a clinical image. Individual detector channels have slightly different electronic offsets and sensitivities, so the scanner uses calibration data to normalize their responses. The reconstruction system then mathematically combines the corrected projection data and estimates the attenuation coefficient of the material represented by each reconstructed voxel. Those attenuation values are converted into CT numbers, commonly expressed as Hounsfield Units.

A CT image can therefore be wrong even when the detector is online and the scan completes. A drifting detector channel, stale calibration, unstable X-ray output, incorrect correction data, or reconstruction problem can change the final CT values or create artifacts. Preserve whether the problem is visible in a phantom, whether it stays fixed relative to the detector, whether water CT number or uniformity has shifted, whether recalibration changes the symptom, and whether the issue follows acquisition or reconstruction. Those clues are much more useful than treating every abnormal CT image as a bad tube or bad detector.

Start With X-Ray Attenuation

CT depends on the fact that different materials weaken an X-ray beam by different amounts.

This weakening is called:

Attenuation.

As X-rays pass through matter, photons can be:

The amount remaining in the original beam decreases.

Different Materials Attenuate Differently

Air attenuates very little.

Soft tissue attenuates more.

Bone attenuates much more.

That difference contains the information CT needs.

The Detector Measures What Survives

Imagine an X-ray beam with a known intensity entering the patient.

The detector measures the intensity that exits.

If a large fraction disappears:

The material along that path had relatively high attenuation.

If most of it survives:

Attenuation was lower.

CT Is Not Measuring Density Directly

People often describe CT as showing:

Density.

That is a useful clinical shortcut.

Technically, CT is reconstructing X-ray attenuation.

Physical density influences attenuation, but so do:

Beer-Lambert Relationship

X-ray attenuation can be described mathematically by an exponential relationship between:

A working biomed does not need to calculate this by hand during routine service.

The important concept is:

The detector measures transmitted radiation, and the scanner works backward to estimate what attenuation must have occurred.

One Measurement Is Not Enough

Suppose the detector reports:

50% of the original beam remained.

That does not tell the scanner exactly where along that path the attenuation occurred.

It only describes the total effect along the entire beam path.

Rotation Solves the Location Problem

The scanner measures the patient from many angles.

Each new angle provides another set of paths through the anatomy.

The reconstruction computer combines those intersecting measurements to estimate attenuation at individual locations inside the patient.

Detector Array

CT scanners use arrays containing many detector elements.

The exact geometry varies by scanner generation and design.

Modern scanners may have many detector rows extending along the patient's head-to-foot direction.

Each Detector Element Is a Measurement Channel

Every channel contributes information about one portion of the X-ray beam.

That means consistency between channels matters enormously.

If One Channel Reads Differently

The reconstruction system may interpret that difference as:

Real patient attenuation.

Unless calibration corrects it.

CT Detector Construction

Modern CT detectors commonly use scintillation materials coupled to photodiodes.

The basic conversion chain is:

X-Ray Photon

↓

Scintillator

↓

Visible Light

↓

Photodiode

↓

Electrical Signal

Scintillator

The scintillator absorbs X-ray energy and converts it into visible light.

Detector materials are chosen for properties such as:

Why Fast Response Matters

CT acquires data very quickly.

The detector must respond to one measurement and be ready for the next without excessive residual signal.

Afterglow

Some scintillation materials continue producing a small amount of light after the original X-ray interaction.

This residual response is called:

Afterglow.

Too much could contaminate following measurements.

Photodiode

The photodiode converts scintillator light into electrical charge.

More detected X-ray energy generally produces:

More light

and therefore:

More electrical signal.

Detector Electronics

The photodiode signal is small.

Electronics must:

with very high stability.

Data Acquisition System

The detector outputs feed into the:

Data Acquisition System, or DAS.

The DAS converts detector signals into digital values the reconstruction system can use.

The DAS Has a Huge Job

During a scan it must process:

with precise timing.

Channel Offset

Even with no X-rays hitting a detector channel, the electronics may not produce a perfect zero.

There can be a small baseline signal.

This is:

Offset.

Why Offset Needs Correction

Suppose two channels receive no radiation.

Channel A reports:

10 counts.

Channel B reports:

14 counts.

If those baselines are not corrected, the system may interpret the four-count difference as real signal.

Gain

Different detector channels can also respond differently to the same amount of radiation.

That difference is often described as:

Gain variation.

Example

Expose two channels equally.

Channel A produces:

10,000 counts.

Channel B:

9,700 counts.

Neither channel must necessarily be defective.

Manufacturing and electronic differences exist.

The scanner calibrates them.

Detector Normalization

Calibration allows the scanner to normalize individual detector responses so a uniform X-ray field produces a uniform interpreted response.

Air Calibration

CT systems commonly perform calibration with no patient or phantom in the beam.

The X-ray beam passes through:

Air.

The scanner measures how each detector channel responds.

Why Use Air?

Air provides a predictable low-attenuation reference condition.

It lets the scanner characterize:

depending on calibration design.

The System Builds Correction Data

When future scans occur, the scanner can compensate for known channel differences.

Calibration Does Not Repair a Bad Detector

This is important.

Calibration can correct stable differences.

It cannot necessarily compensate for a detector channel that is:

Stable Error vs Unstable Error

A detector that is always 2% low may be correctable.

A detector that randomly changes between:

Normal

and:

20% low

is much harder to correct.

Drift

Detector response can change over time because of:

That is why periodic calibration and QC matter.

Detector Temperature

Detector output can be temperature-dependent.

Modern CT systems control or compensate for detector temperature.

Warm-Up and Stability

A scanner that has been off for an extended time may require:

before stable quantitative imaging.

Follow OEM procedures.

Projection Data

After correction, the detector readings become part of the raw projection dataset.

Each projection describes X-ray attenuation along many paths through the patient at one tube angle.

Sinogram

CT projection data can be represented in a form called a:

Sinogram.

This is not a clinical image.

It is a representation of projection measurements arranged by detector position and angle.

Why Raw Data Can Be Useful

Certain detector defects become easier to recognize before reconstruction.

A bad detector channel may create a recognizable pattern in raw projection or sinogram data.

Reconstruction

The computer mathematically reconstructs attenuation values from the projection measurements.

Historically, methods such as:

Filtered back projection

were widely used.

Modern CT also uses various forms of:

depending on scanner.

Reconstruction Produces Voxels

A CT image is made of pixels, but each reconstructed pixel represents a volume element with a finite slice thickness.

That three-dimensional element is called a:

Voxel.

Each Voxel Gets an Attenuation Estimate

The reconstruction system estimates the linear attenuation coefficient for the material represented by that voxel.

But Raw Attenuation Coefficients Are Not Convenient Clinically

The numbers would be less intuitive and vary with measurement units.

CT therefore converts them to:

Hounsfield Units.

Hounsfield Units

Hounsfield Units, or:

HU

place reconstructed attenuation on a standardized scale relative to water.

Water

Water is defined approximately as:

0 HU.

Air

Air is approximately:

-1000 HU.

Bone

Dense bone has strongly positive HU values.

Exact values vary substantially based on:

Fat

Fat commonly has negative CT numbers.

Soft Tissue

Many soft tissues cluster around modest positive values.

The exact numbers vary.

Why Water Is So Important

Water provides a convenient reproducible reference material.

That is why CT QC often pays close attention to reconstructed water value.

Water CT Number Shift

Suppose a water phantom should reconstruct near:

0 HU

but now measures:

+20 HU.

The image may still look like a uniform circle of water.

But quantitatively something has shifted.

That Matters

Possible areas include:

depending on the system.

Hounsfield Unit Formula

Conceptually, CT number compares the attenuation of a material with the attenuation of water.

You do not need to calculate it manually in routine service.

The important point is:

HU is derived from measured attenuation.

It is not simply an arbitrary grayscale number.

Pixel Value vs Display Brightness

A common source of confusion:

The underlying CT number is not the same as how bright the pixel appears on the monitor.

Display brightness depends on:

Window and level.

Window Level

The level determines the center of the displayed HU range.

Window Width

The width determines how large an HU range is displayed across the grayscale.

Example

A narrow window can emphasize small differences in soft tissue.

A wide window can display a broader range useful for structures such as lung or bone.

Same CT Data, Different Appearance

The same reconstructed scan can look dramatically different with different window settings.

Nothing about the actual detector measurement changed.

This Is Why “Image Looks Too Bright” Needs Clarification

Is the problem:

Those are different systems.

CT Number Accuracy

QC may use known materials to verify reconstructed CT numbers.

If known material consistently reconstructs incorrectly:

That indicates a quantitative imaging problem.

CT Number Uniformity

A uniform water phantom should produce reasonably consistent CT numbers across the image.

Center vs Edge

Measurements may be taken at:

Why Uniformity Matters

If the same material appears:

0 HU in the center

but:

+15 HU at one edge,

the system is not treating identical material uniformly.

Possible Causes of Nonuniformity

Depending on pattern:

may be considered.

Noise

Even a perfectly uniform phantom does not produce exactly the same CT value in every pixel.

Random statistical variation exists.

This appears as:

Image noise.

Quantum Noise

X-rays arrive as discrete photons.

The number detected has natural statistical variation.

Fewer photons generally mean:

More relative quantum noise.

Higher Dose

More detected photons can reduce quantum noise.

That is why image quality and dose are linked.

But Noise Has Other Sources

Noise can also involve:

Measuring Noise

QC commonly measures the standard deviation of CT numbers in a uniform region.

A larger standard deviation means:

More variability.

Noise Trend

If the same phantom and technique suddenly produce substantially higher noise:

Something changed.

Do Not Blame the Detector Immediately

Possible changes include:

Ring Artifacts

Ring artifacts are one of the most classic CT detector-related patterns.

They appear as:

Circular or concentric rings.

Why Rings Form

Imagine one detector channel reads slightly wrong during every projection.

As the gantry rotates, that same detector channel traces a circular path around the reconstructed image.

Its repeated error becomes a ring.

This Is Geometry Giving You a Clue

The shape of the artifact reflects how the acquisition system moved.

Ring Artifact Does Not Automatically Mean Replace Detector

Possible causes include:

Run the appropriate OEM diagnostics.

Calibration May Correct a Stable Ring

If the problem is stable channel-response drift:

Approved recalibration may correct it.

Ring Returns Quickly

If the same ring repeatedly comes back after calibration:

A detector channel may be unstable rather than simply miscalibrated.

Bad Detector Channel

A severely defective channel may:

Detector Redundancy and Correction

Modern systems may have sophisticated correction methods.

A small detector defect does not always produce a dramatic clinical artifact.

Hidden Degradation

This is why QC can find problems before clinicians report them.

Beam Hardening

X-ray attenuation depends on photon energy.

Lower-energy photons are absorbed more strongly than higher-energy photons.

As a polychromatic beam passes through the patient:

Its average energy increases.

This is:

Beam hardening.

Beam-Hardening Correction

The CT reconstruction system applies corrections based on expected behavior.

Calibration and Beam Spectrum Are Linked

If tube output or filtration differs substantially from expected conditions:

Corrections may not work as intended.

Bowtie Filter

CT uses shaped filtration to alter X-ray intensity across the fan beam.

The filter reduces unnecessary peripheral exposure and helps manage detector dynamic range.

Wrong Filter Position

A filter-position or identification problem can therefore influence:

Scatter

Scatter also changes detector measurements.

CT geometry and reconstruction corrections reduce its impact, but it remains part of the physical system.

Detector Saturation

Detector and DAS electronics have a finite measurable range.

Too much signal can exceed that range.

Photon Starvation

The opposite can occur through extremely attenuating anatomy or metal.

Too few photons reach some detector channels.

Result

The projection data becomes noisy or unreliable.

This can cause:

Metal Artifact

Metal can produce a combination of:

Modern correction methods help, but not all artifact means equipment failure.

Patient vs Equipment Artifact

This distinction matters.

An artifact caused by:

may vary with the patient.

A detector-related ring may repeat on:

in the same detector coordinates.

Phantom Testing Removes Patient Variables

That is why QC phantoms are so valuable.

Water Phantom

A simple uniform phantom can tell you a surprising amount.

It can help evaluate:

A Scanner Can Pass Startup and Still Fail the Water Phantom

Startup proves the system can initialize.

It does not prove quantitative image performance.

Detector Communication Fault

The system may directly report a detector or DAS communication error.

That is different from subtle image degradation.

Hard Failure

A detector module does not communicate.

Scan may be:

Soft Failure

All modules communicate.

One channel is drifting.

Scan completes.

Image quality slowly deteriorates.

Soft failures are often more interesting.

DAS Failure

Do not assume every detector-related symptom originates in the scintillator itself.

Possible areas include:

Rotating-Side Power

The detector system rides on the rotating gantry.

It depends on stable power.

A power problem can therefore affect:

Data Transfer

Digital projection data must move from the rotating gantry to the stationary reconstruction system.

Communication failure can produce:

Raw Data vs Image Data

If raw acquisition completes correctly but images reconstruct incorrectly:

Look farther downstream.

If raw data itself is abnormal:

Look farther upstream.

Reconstruction Kernel

The selected reconstruction kernel can strongly change image:

Sharp Kernel

May improve edge detail but increase visible noise.

Smooth Kernel

Can reduce noise but soften detail.

Do Not Compare QC Images Using Different Reconstruction Settings

The test conditions need to match.

Iterative Reconstruction

Modern algorithms can significantly change noise appearance.

If software or protocol changes:

Historical image appearance may change without hardware degradation.

Software Update

After a reconstruction-software update:

Do not immediately assume hardware failure if image texture changes.

Verify:

CT Number Can Be Energy-Dependent

Materials other than water can produce different HU values at different kVp because attenuation characteristics change with photon energy.

This Is Normal Physics

It is another reason QC should use the specified:

Dual-Energy and Spectral CT

Some modern scanners intentionally acquire or separate information at different X-ray energies.

These systems can derive information beyond conventional single-energy CT numbers.

Detector Technology Is Evolving

Some systems use:

Energy-integrating detectors.

Newer systems may use:

Photon-counting detector technology.

Energy-Integrating Detector

Traditional CT detectors measure the total deposited energy over an acquisition interval.

Photon-Counting Detector

A photon-counting detector can detect individual photon events and may separate them by energy range.

Why That Matters

It changes detector architecture and allows additional spectral information.

But the core principle remains:

Measure transmitted X-rays accurately enough to reconstruct anatomy.

Calibration Is Still Fundamental

Different detector technology does not eliminate:

calibration needs.

Real-World Example: Ring Artifact

Technologists report a faint circular artifact.

Water phantom reproduces it at the same radius.

Air calibration reduces the ring temporarily.

It returns the next day.

One detector channel shows unstable response.

Now the issue is no longer simply:

CT artifact.

You have a repeatable detector-channel pattern.

Real-World Example: Water CT Number Shift

Image looks generally normal.

QC water measurement changes from:

0 HU

to:

+16 HU.

Noise remains normal.

Detector calibration or beam/reconstruction correction now deserves investigation.

Clinical appearance alone would not have caught it.

Real-World Example: Increased Noise

Routine phantom historically measures:

5 HU standard deviation.

Now:

9 HU.

Technique appears unchanged.

Further testing shows X-ray tube output is reduced.

The detector was accurately reporting fewer photons.

Real-World Example: Edge Nonuniformity

Water center measures close to 0 HU.

Peripheral measurements shift progressively positive.

Detector communication is normal.

This pattern directs troubleshooting toward:

rather than one dead detector channel.

Real-World Example: Scan Completes but Reconstruction Fails

Detector acquisition completes.

Raw data exists.

No clinical images are produced.

Now the detector may have done its job.

The failure lies farther downstream in:

Common Mistakes

Assuming every CT image artifact is a detector failure. Patient motion, metal, beam hardening, tube output, calibration, and reconstruction can all create artifacts.

Assuming a detector that communicates is a good detector. Communication does not prove calibration, stability, or channel accuracy.

Recalibrating repeatedly without asking why calibration will not hold. A drifting detector channel may need repair rather than another normalization.

Looking only at image appearance. CT-number accuracy, noise, and uniformity can fail before the image looks obviously bad.

Comparing images reconstructed with different kernels. Reconstruction settings strongly influence noise and sharpness.

Treating Hounsfield Units as display brightness. HU is reconstructed attenuation information. Window and level determine how those values are displayed.

A Useful CT Detector Framework

Think:

X-Ray Output

↓

Patient / Phantom Attenuation

↓

Scintillator

↓

Photodiode

↓

Detector Electronics

↓

DAS

↓

Offset / Gain Correction

↓

Projection Data

↓

Reconstruction

↓

Hounsfield Units

↓

Window / Level Display

That chain helps separate acquisition, calibration, reconstruction, and display.

Another Useful Troubleshooting Question

When a CT image is abnormal, ask:

Is the measurement wrong, or is the measurement being displayed differently?

Then ask:

Is the problem present in raw acquisition, reconstructed CT numbers, or only the displayed image?

Those questions can save a lot of random parts replacement.

What Did You Actually Prove?

If all detector modules report:

Online,

you proved:

The scanner can currently communicate with those detector modules.

You did not prove:

If a water phantom measures near:

0 HU

at the center:

You proved:

The reconstructed water value at that location met the expected condition under that technique.

You did not necessarily prove:

If the full applicable QC procedure passes:

You have much stronger evidence that the acquisition and reconstruction chain is functioning properly under those controlled conditions.

Final Thoughts for Biomeds

CT detectors are not cameras.

They are precision measurement systems.

Each detector element measures X-ray energy.

The DAS turns those signals into numbers.

Calibration corrects channel differences.

Reconstruction turns thousands of measurements from different angles into estimated attenuation values.

Then those values become Hounsfield Units.

Only after all of that does the scanner turn them into the grayscale image the clinician sees.

That is why:

It makes an image

is such a weak test of CT performance.

The better questions are:

Does known material reconstruct to the correct CT number?

Is uniform material actually uniform?

Has image noise changed?

Does an artifact remain fixed relative to detector geometry?

Does calibration correct the problem and stay corrected?

Once you understand that chain, a ring artifact is not just a weird circle.

It is evidence.

A water-value shift is not just:

QC being picky.

It tells you the scanner's relationship between physical attenuation and reconstructed number has changed.

That is where CT troubleshooting becomes less about staring at pictures and more about understanding the measurements that created them.

And as always:

What did you actually prove?

— Jake

Important Note

CT detector materials, detector geometry, DAS architecture, calibration routines, acceptable Hounsfield Unit values, uniformity limits, image-noise requirements, reconstruction algorithms, QC procedures, and service boundaries vary significantly by manufacturer, scanner model, protocol, and jurisdiction. Follow current OEM documentation, approved CT quality-control procedures, radiation-safety requirements, and qualified medical-physics guidance when evaluating quantitative CT image performance.

Related Biomed Basics

Back to Biomed Basics