Why Science? | X-Ray Computed Tomography, Attenuation Projections and 3D Internal Evidence

Section 1 of 36

1. Begin with transmitted X-rays

X-ray computed tomography begins by sending X-rays through an object and measuring what reaches a detector. Materials attenuate the beam through absorption and scattering, so each detector value combines contributions along one path. The first measurement is a projection, not a slice. CT earns three-dimensional meaning only after many projections, calibration and a reconstruction model connect those line measurements.

Contents · Next section

Section 2 of 36

2. Translate intensity into attenuation

A flat-field image estimates the beam without the object, while a dark image characterises detector offset. Comparing transmitted intensity with the open beam supports a logarithmic line-integral model. That model assumes stable source and detector behaviour and treats complex polychromatic interactions approximately. Incorrect dark or flat correction can create structure before reconstruction has even begun.

Contents · Previous section · Next section

Section 3 of 36

3. Rotate the specimen through many angles

In many industrial systems, the object rotates between source and detector. Each angle supplies a different set of paths through the interior. Too few views leave angular gaps that appear as streaks or blurred edges. Motion between views breaks the assumption that all projections describe one unchanged object. Acquisition geometry belongs to the evidence, not merely to instrument setup.

Contents · Previous section · Next section

Section 4 of 36

4. Distinguish cone beam and fan beam

A fan beam spans a slice, while a cone beam covers a volume on a two-dimensional detector. Cone-beam geometry enables efficient 3D acquisition but increases sensitivity to alignment, scatter and reconstruction assumptions. Source-to-object and object-to-detector distances control magnification. The same voxel size does not guarantee the same spatial resolution when focal spot, motion and detector blur differ.

Contents · Previous section · Next section

Section 5 of 36

5. Build a voxel volume

Reconstruction estimates attenuation within small volume elements called voxels. A voxel is a computational cell, not a miniature physical cube removed from the object. Its value averages material and artefact influences over a finite neighbourhood. Visualisation software may colour, smooth or threshold these values. A beautifully rendered surface is therefore one interpretation of a reconstructed field.

Contents · Previous section · Next section

Section 6 of 36

6. Relate attenuation to material cautiously

Attenuation depends on composition, density and X-ray energy. Dense or high-atomic-number materials often attenuate more strongly, but different combinations can overlap, especially with a broad source spectrum. A grey value is not automatically a material identity or mass density. Calibration phantoms, energy knowledge and complementary chemistry are needed when composition, rather than geometry, is the claim.

Contents · Previous section · Next section

Section 7 of 36

7. Treat resolution as a system property

Resolution depends on focal-spot size, magnification, detector sampling, contrast, noise, reconstruction and motion. Voxel spacing may be smaller than the smallest reliably separable feature. Report how resolution was tested, perhaps with a calibrated feature or modulation method, instead of equating one voxel with one detectable crack. The relevant resolution can also vary across the field.

Contents · Previous section · Next section

Section 8 of 36

8. Read orthogonal slices before the rendering

Axial, sagittal and coronal views expose internal grey values without hiding them behind a surface. Scroll through neighbouring slices to test whether a feature persists and follows plausible geometry. A single screenshot may exaggerate a threshold choice or partial-volume edge. Slices, projections and 3D rendering answer different questions; they should support each other rather than compete for visual drama.

Contents · Previous section · Next section

Section 9 of 36

9. Use window and level honestly

Display windows map a chosen attenuation range into visible brightness. Narrow windows reveal subtle contrast but may clip dense regions; wide windows preserve range but hide small differences. Changing the window does not change reconstructed data, yet it changes perception. State display settings for critical comparisons and inspect histograms or quantitative values instead of trusting one optimised screenshot.

Contents · Previous section · Next section

Section 10 of 36

10. Segment with declared rules

Segmentation assigns voxels to regions such as pore, metal or polymer. A global threshold may work for well-separated values, while gradients, noise and partial-volume effects require more careful models. Manual editing introduces judgement. Record thresholds, software, preprocessing and reviewer rules. A segmented volume is an analytical result, not the untouched scan, and alternative segmentations can change measured dimensions.

Contents · Previous section · Next section

Section 11 of 36

11. Measure internal dimensions with uncertainty

CT can access internal features that touch probes cannot reach, but edge location depends on contrast, blur and segmentation. Dimensional metrology therefore needs calibrated artefacts, geometry checks, environmental control and uncertainty. NIST’s traceability work emphasises a chain from instrument response to recognised standards. Reporting four decimal places does not make an uncalibrated boundary traceable.

Contents · Previous section · Next section

Section 12 of 36

12. Count defects without inventing completeness

A pore or crack must be large and contrasted enough to be detected. Threshold choice can merge nearby pores, split one irregular pore or miss features beside dense material. State detection limits and count rules. CT may establish that a defect exists without proving that no smaller defects are present. “No defect detected” is bounded by the validated performance window.

Contents · Previous section · Next section

Section 13 of 36

13. Recognise beam hardening

Lower-energy X-rays are attenuated preferentially as a polychromatic beam crosses material, so the surviving spectrum becomes harder. Reconstructions may show cupping, dark bands or bright edges, especially near dense regions. Filtration and correction models help, but residual effects can mimic material variation. A radial grey-value trend deserves an artefact test before it becomes a composition map.

Contents · Previous section · Next section

Section 14 of 36

14. Recognise scatter and rings

Scattered photons add signal that did not follow the assumed straight path, reducing contrast and biasing values. Detector elements with imperfect calibration can create circular ring artefacts around the rotation axis. Corrections, shielding and detector calibration help. Features that align with the instrument coordinate system, rather than the specimen, should be treated as suspicious until independently tested.

Contents · Previous section · Next section

Section 15 of 36

15. Control motion and stage geometry

Rotation-axis wobble, stage drift, thermal expansion or sample motion can blur or duplicate edges. NIST studies show that detector and rotation-stage geometry affect dimensional results, especially across magnifications. Use stable mounting, warm-up, geometric calibration and repeat scans. A feature that shifts with scan setup may belong to the instrument model rather than the object.

Contents · Previous section · Next section

Section 16 of 36

16. Practise with an invented CT table

These fictional values support classroom reasoning only; they are not inspection limits.

The measurement window changes what “defect-free” can mean.

Contents · Previous section · Next section

Section 17 of 36

17. Understand partial-volume effects

When a voxel contains more than one material, its value becomes a mixture influenced by blur and reconstruction. Thin walls and oblique surfaces are especially vulnerable. Thresholding may then move the apparent boundary as voxel size or orientation changes. Scan calibrated thicknesses and repeat orientations when small-dimensional claims approach the system resolution. Mixed voxels are physics, not merely bad software.

Contents · Previous section · Next section

Section 18 of 36

18. Use phantoms as performance questions

A phantom contains known or calibrated features designed to test resolution, dimensional accuracy, contrast or defect detection. NIST’s 2026 review highlights fabricated artefacts and the need for standardised metrology. A useful phantom resembles the material, geometry and feature scale of the real task. Passing one easy phantom does not validate every object, magnification and reconstruction.

Contents · Previous section · Next section

Section 19 of 36

19. Trace measurements to standards

Traceability is an unbroken, documented calibration chain linking a result to recognised references, with uncertainty at each step. For CT, that may include length standards, calibrated artefacts, environmental conditions, geometry and analysis software. Traceability does not mean perfect truth. It means another laboratory can understand how the number relates to standards and where uncertainty enters.

Contents · Previous section · Next section

Section 20 of 36

20. Compare repeated scans and orientations

Repeatability tests expose noise, stage and segmentation variation under the same setup. Rotating or repositioning the specimen can reveal directional artefacts and hidden surfaces. Reproducibility across operators or instruments tests a larger chain. Agreement strengthens a claim when analysis rules are fixed before comparison. Quietly choosing the best-looking scan after several attempts introduces selection bias.

Contents · Previous section · Next section

Section 21 of 36

21. Connect local detail to the whole object

High magnification improves sampling of a small region but may exclude the full object or require stitching. Whole-object scans provide context at lower detail. A defensible workflow can use a coarse survey to select documented regions for finer scans. The final claim must state whether it covers one region, all scanned volume or the manufactured population—not slide between scales.

Contents · Previous section · Next section

Section 22 of 36

22. Challenge the claim “the 3D model is the object”

The model is reconstructed from finite, noisy projections through algorithms and displayed through segmentation. Surfaces can shift with thresholds and corrections. Validate critical features in raw projections, neighbouring slices, repeats and, where appropriate, an independent destructive or optical method. A rendering is scientifically valuable precisely when its construction and limits remain visible.

Contents · Previous section · Next section

Section 23 of 36

23. Challenge the claim “no pore means no pore”

A scan can only detect pores within its contrast, size, orientation and artefact limits. Dense inclusions, edge blur and partial-volume mixing can conceal voids. Phrase conclusions as “no pores above the validated detection threshold were observed in the scanned volume.” That longer sentence is stronger because it identifies both the result and the boundary of the method.

Contents · Previous section · Next section

Section 24 of 36

24. Compare CT with radiography and microscopy

Radiography compresses a 3D object into a projection and can be fast; CT reconstructs internal location from many projections; microscopy may provide finer surface or local detail on a prepared section. CT is non-destructive for many objects but not automatically highest resolution or chemically specific. Combining methods can test whether a suspected feature persists across different signal pathways.

Contents · Previous section · Next section

Section 25 of 36

25. Did You Know? Every slice depends on all-around views

A displayed cross-section can feel like a photograph taken inside the object, yet its values were calculated from many external projections. Mathematics distributes measured attenuation back through the volume according to a model. That is the cheerful magic of CT—and the reason missing angles, motion or geometry errors can place artefacts far from the original measurement path.

Contents · Previous section · Next section

Section 26 of 36

26. Did You Know? Smaller voxels can add noise

Reducing voxel spacing samples the reconstruction more finely, but fewer photons per detector region and stronger magnification demands may reduce signal quality. Tiny voxels do not automatically add resolvable information. The best setup balances feature contrast, field of view, dose, time and stability for the claim. Evidence quality is an optimisation problem, not a smallest-number contest.

Contents · Previous section · Next section

Section 27 of 36

27. Preserve the complete CT audit trail

Record specimen identity and mounting, source voltage and current, filtration, distances, magnification, detector and binning, projection count, exposure, rotation path, dark and flat corrections, reconstruction algorithm and parameters, voxel spacing, artefact corrections, phantom results, segmentation rules, software versions, repeats, environmental conditions, raw projections, volume data and uncertainty.

Contents · Previous section · Next section

Section 28 of 36

28. Write a claim–evidence–limit paragraph

Try: “Three scans detected an internal void whose segmented equivalent diameter was 0.84 mm with 0.03 mm repeat spread. A calibrated phantom confirmed detection and dimensional performance near this scale, and the feature persisted after specimen reorientation. The result applies to the scanned volume; smaller or low-contrast voids may remain below the validated threshold.”

Contents · Previous section · Next section

Section 29 of 36

29. Connect Physics, Chemistry and Mathematics

Physics supplies X-ray production, attenuation, scattering and detectors. Chemistry explains composition-dependent interactions. Mathematics supplies logarithms, projections, reconstruction, geometry and uncertainty. Singapore’s 2026 O-Level Physics and Chemistry syllabuses build related models of radiation and matter. CT shows how those ideas cooperate to answer a question that no single projection could solve.

Contents · Previous section · Next section

Section 30 of 36

30. Learn safely with open datasets

Students can inspect supplied radiographs and slices, vary display windows, compare thresholds, identify artefacts and write detection-limited claims. Real CT equipment uses ionising radiation and controlled enclosures, with regulatory and facility procedures. Never improvise X-ray sources or defeat interlocks. Classroom learning should use verified public datasets, simulations or supervised facility demonstrations.

Contents · Previous section · Next section

Section 31 of 36

31. Make science tuition earn its place

Good science tuition asks what the detector measured, which assumptions created the slice and how a threshold changes the conclusion. Learners can progress from Primary Science shadows and PSLE Science fair tests to Secondary Science, O-Level Science and STEM reasoning about waves, materials, graphs, models and uncertainty. The dramatic 3D image becomes a transparent argument.

Contents · Previous section · Next section

Section 32 of 36

32. Use the topic for school choices

Verify official school or programme descriptions of imaging, engineering design, data science and safety. A school need not own a CT scanner to teach excellent reconstruction reasoning; open volumes, models and coding exercises can build foundations. Do not infer guaranteed instrument access, admission advantage, scholarships or career outcomes from a facility photograph or promotional rendering.

Contents · Previous section · Next section

Section 33 of 36

33. See the career ecosystem

X-ray CT work spans medical physics, radiography, materials science, manufacturing, aerospace, geology, palaeontology, conservation, metrology, software and detector engineering. Roles range from acquisition and radiation safety to reconstruction, calibration, segmentation and quality assurance. Qualifications and legal authorisations vary. Current official course, facility and employer sources should guide pathway decisions.

Contents · Previous section · Next section

Section 34 of 36

34. Use a shadow-carousel analogy—with limits

Imagine recording many shadows while an object turns, then calculating where the hidden material must be to explain them all. The analogy captures projections and reconstruction. It misses energy-dependent attenuation, scatter, cone-beam geometry, partial-volume effects and regularisation. Use the carousel to orient yourself, then return to logarithms, calibration and uncertainty.

Contents · Previous section · Next section

Section 35 of 36

35. Ask what CT cannot tell you alone

CT can reveal internal attenuation structure, dimensions and detectable defects under validated conditions. It may not identify chemical composition uniquely, show features below resolution, explain why a defect formed or predict performance. Pair it with chemistry, microscopy, mechanical tests and process records when the larger question demands them. Non-destructive does not mean interpretation-free.

Contents · Previous section · Next section

Section 36 of 36

36. Keep the photon-to-projection-to-volume chain visible

Begin with a defined internal question, select energy and geometry for the object, stabilise and calibrate the system, collect sufficient projections, correct detector response, reconstruct with declared parameters, inspect slices and projections, validate resolution with relevant phantoms, segment by recorded rules, repeat orientations and report uncertainty. Then distinguish measured attenuation from calculated voxels and interpreted material boundaries. CT becomes trustworthy when the smooth 3D view never hides the photons, assumptions and detection limits that made it possible. A strong report also identifies which projections were rejected, whether corrections were chosen before inspecting the result, how repeated scans changed the dimensions and which independent observation could falsify the preferred interpretation. Those details make a reconstruction reusable evidence rather than a persuasive animation. They also let a future analyst reproduce the volume, test a different threshold and discover whether the conclusion survives a better-calibrated instrument.

Contents · Previous section · Continue to the Science Learning Hub