An X-ray image is a measurement translated into a visible display. Radiation passes through an object, interacts with its materials and reaches a detector in a pattern. Geometry, energy, scattering, detector response and processing all help determine what the final image can show.
The gold standard is to connect the displayed feature with the measurement that supports it. This is an educational physics guide covering projection and reconstructed images, not clinical interpretation, advice about whether to have an examination, or instructions for operating X-ray equipment.
eduKateSG · X-ray Imaging
Find your next learning step
Follow the path from transmitted radiation to a displayed image, then use simple models to recognise both the information and the ambiguity.
Full chapter index · Practice and transfer · Education root · How X Works
Full chapter index
Understand image formation · Chapters 1–2
Calculate a projection · Chapters 3–4
Examine image quality · Chapters 5–7
Question and practise · Chapters 8–10
Find the next explanation · Chapters 11–12
CHAPTER 1 OF 12 · Understand image formation
1. The detector does not see an ordinary photograph
In transmission imaging, an object changes the radiation reaching a detector. The detector’s measurements are then recorded and processed into an image. This differs from a visible-light photograph made chiefly from light reflected off a surface.
The displayed brightness is a convention linked to the acquisition and processing system. A brighter region should not be interpreted without knowing how the image was produced and displayed. Start by identifying the modality, orientation, scale and relevant processing information.
FDA’s description of medical X-ray imaging explains that absorption and scattering affect the transmitted pattern. The same physical distinction is useful when reasoning about non-medical inspection images.
CHAPTER 2 OF 12 · Understand image formation
2. Attenuation depends on more than material name
Attenuation describes reduction of the primary beam along a path. Both absorption and scattering can remove photons from that primary beam. The effect depends on radiation energy, material composition, density and path length.
NIST provides mass attenuation data whose energy dependence matters. A coefficient at one energy is not a universal constant for every beam. Also distinguish a mass attenuation coefficient from a linear attenuation coefficient; their units and use differ.
A familiar material can appear different when its thickness or the acquisition changes. Conversely, different combinations of material and thickness can produce similar transmission. That is an ambiguity to investigate, not a contradiction.
CHAPTER 3 OF 12 · Calculate a projection
3. Worked example: a simple transmission model
For a uniform material and a monochromatic narrow-beam model without added scattered radiation, let T = I/I₀ = exp(−μx). Here I₀ is the incident intensity, I the transmitted primary intensity, μ a linear attenuation coefficient and x the path length.
Choose μ = 0.20 cm⁻¹ and x = 5 cm for a fictional example. The exponent is −1, so T ≈ 0.3679. About 36.79% of the modelled primary intensity remains along that path.
Doubling the path length to 10 cm gives exp(−2) ≈ 0.1353, or 13.53%. Transmission is not reduced by the same number of percentage points for every extra thickness. Actual broad-spectrum imaging needs more than this ideal model.
CHAPTER 4 OF 12 · Calculate a projection
4. A projection combines contributions along a ray
For multiple uniform layers in the same ideal model, transmission is exp(−Σμᵢxᵢ). Taking the negative natural logarithm gives a sum of attenuation–length products. That sum describes the ray, not the unique arrangement of everything along it.
Layer A with μ = 0.10 cm⁻¹ and x = 2 cm contributes 0.20. Layer B with μ = 0.30 cm⁻¹ and x = 1 cm contributes 0.30. Total is 0.50, so transmission is exp(−0.50) ≈ 0.6065.
Reversing the order of those layers leaves this simplified transmission unchanged. One projection cannot recover the order merely from this sum. Real additional physics can complicate the model, but does not justify pretending the simple sum contains unique depth information.
| Image or quantity | What it represents | Inference limit |
|---|---|---|
| Projection image | Overlapping ray-path contributions | Depth is not uniquely separated |
| Transmission ratio | Signal relative to a stated reference | Material and thickness may be confounded |
| CT reconstruction | An estimated spatial attenuation distribution | Acquisition and reconstruction assumptions matter |
With a diverging beam, objects nearer the source can project at a larger scale. In a simple point-source geometry, magnification is source-to-detector distance divided by source-to-object distance.
If those distances are 100 cm and 80 cm, magnification is 1.25. A 20 mm object ideally projects to 25 mm. A measured 25 mm image feature therefore does not necessarily imply a 25 mm object.
A finite source size, motion and detector response can blur boundaries. A smaller displayed pixel is not proof of sharper real information. Separate sampling spacing from the system’s ability to distinguish nearby features.
CHAPTER 6 OF 12 · Examine image quality
6. Noise, contrast and processing are different questions
Two regions can differ physically yet be difficult to distinguish when the signal difference is small relative to noise. Image quality therefore concerns the task, not simply whether an image looks smooth or visually striking.
In a simplified photon-counting model with independent Poisson counts, standard deviation is √N and relative counting fluctuation is 1/√N. At an expected 100 counts this is 10%; at 400 it is 5%. This is a statistical illustration, not a recommendation to change an exposure.
Processing may make a feature easier to see but cannot automatically restore missing or unreliable measurement information. Smoothing, sharpening and display-window changes can alter appearance, so record them when making comparisons.
CHAPTER 7 OF 12 · Examine image quality
7. CT uses multiple projections and reconstruction
Computed tomography combines measurements from multiple directions with a reconstruction method to estimate cross-sectional structure. FDA distinguishes these reconstructed slices from conventional two-dimensional projections.
A reconstructed slice is not an independent photograph taken inside the object. Its appearance depends on the measurements, geometry, calibration and reconstruction choices. Missing information, motion, scattering and other departures from assumptions can create artefacts.
Reconstruction helps resolve overlap, but it does not remove every ambiguity or make every feature an identified substance. Quantitative interpretation still requires appropriate calibration, modelling and specialist knowledge.
CHAPTER 8 OF 12 · Question and practise
8. Quality and radiation protection must be considered together
Medical X-ray imaging uses ionizing radiation. FDA describes justification and optimisation as radiation-protection principles: the examination should answer an appropriate question, and the procedure should be managed for that purpose.
This educational explanation is not a way to decide an individual examination or select acquisition settings. Those decisions belong with the responsible qualified professionals. The paper exercises here require no radiation source, equipment modification or exposure.
In any application, a valid image must be suitable for its task and produced within the relevant controlled process. A visually pleasing display alone does not establish that requirement.
CHAPTER 9 OF 12 · Question and practise
9. Misconceptions that an image cannot settle
A dark region is not automatically a hole, and a bright region is not automatically a specific material. Display conventions, thickness, overlapping structures and processing affect such claims. Clinical images require qualified interpretation; this guide does not provide diagnostic rules.
Equal transmission does not establish equal composition. A single projection does not generally reveal depth order. More pixels do not guarantee more resolved detail. A reconstruction with a sharp boundary does not prove that every visible boundary corresponds to a genuine physical feature.
Replace ‘the image proves’ with a statement naming the feature, acquisition and model that support the inference.
A homogeneous teaching model has μ = 0.10 cm⁻¹ and x = 5 cm. Its attenuation–length product is 0.50, giving the same 0.6065 transmission as the two-layer example. Explain why those equal transmissions do not imply the same internal structure.
Next, a feature measures 30 mm in an ideal projection at magnification 1.5. The inferred object size is 30/1.5 = 20 mm. State the point-source geometry assumption and one source of real blur.
For transfer, compare an inspection radiograph with a reconstructed CT slice. Write what each can help locate, what acquisition information is required and what additional evidence would be needed to identify a material. Do not infer medical findings from these exercises.
Radiation and image evidence
- X-ray Science — understand radiation–matter interactions
- Radiation — separate radiation quantities and effects
- Transmission and material identity — extend the non-uniqueness example
- Projection magnification — extend the geometry calculation
- Pooling and logarithms — examine processing order
Related methods and professional practice
- Medical Imaging Engineering — connect image trust with controlled systems
- Radiographers — understand professional responsibility
- X-ray Spectroscopy — distinguish energy-resolved measurements
- X-ray Diffraction — separate structural scattering patterns from projections
- X-ray Astronomy — compare detected cosmic emission with transmission imaging
CHAPTER 12 OF 12 · Find the next explanation
12. Questions, sources and the next image
Frequently asked questions
Is an X-ray image an ordinary photograph? No; its formation and display require different interpretation. Can one transmission identify a material? Not uniquely without further constraints. Does CT eliminate every artefact? No; reconstruction remains dependent on measurement and modelling.
Choose a published non-clinical example with clear acquisition information. Identify projection versus reconstruction, one geometrical effect and one inference limit. Use simulated or supplied data for the numerical work.
Primary source reading
- FDA: computed tomography — projection and cross-sectional imaging
- FDA: medical X-ray imaging — image formation and radiation-protection context
- NIST: attenuation coefficients — material and energy-dependent data
- NIST: attenuation model and coefficient definitions — distinguish mass thickness from geometrical path length