Before you build: why do radiotherapy machines look like this?

Start with How therapeutic LINACs evolved for the engineering history behind gantries, beam shaping, imaging, motion control and modern machine geometries. Then return here to learn by doing.

Radiotherapy Engineering Lab

Educational Model Build Plans

Radiotherapy can be seen in action only indirectly. Specialized cameras can capture the exceedingly faint Cherenkov light produced within irradiated tissue and emerging from the patient’s skin. Phantoms, detectors and radiographic films can then tell us what the beam does at depth and around its target. Fascinating—but not exactly the easiest live demonstration.

Fortunately, beam shaping itself can be explored cheaply and safely in the visible spectrum. Visible light does not reproduce the interactions of megavoltage photons with tissue, nor does brightness represent absorbed dose. It does, however, allow us to see and control many of the geometrical principles on which radiotherapy depends: direction, divergence, collimation, field shape, alignment, convergence, geometric penumbra and mechanical precision.

The Scientific Committee at carcinologica.com—currently represented by its sole member, who also occupies every other position in this fine emerging project—therefore dares to suggest a safer way to explore how radiotherapy works: by replacing the invisible treatment beam with visible light, which is safe and fun to play with—even for your cat, although felines stubbornly refuse to file QA reports.

The projects range from simple, inexpensive tabletop demonstrators to considerably more ambitious engineering experiments. Together, they explore beam collimation and modulation, isocenter and treatment geometry, gantry and robotic motion, stepper-motor control, image guidance, camera feedback, coordinate registration, calibration, positioning, mechanical tolerances, repeatability, quality assurance, safety interlocks, error detection and treatment-delivery verification—the many systems that must cooperate before a modern radiotherapy machine is allowed to illuminate anything at all.

These are educational optical models, not medical devices. They do not reproduce radiation dose or tissue response. What they can do is make the machinery, logic and precision of modern radiotherapy visible, approachable, safe—and rather enjoyable to watch.

PRACTICAL TRACK · ACCURACY & METROLOGY

Measure what actually moved—not only what the motor reported

This bench sequence separates resolution, accuracy, repeatability and stability. Use a fixed external reference—calibrated camera/fiducial geometry, dial indicator or another independent measuring method—and record the physical pose rather than trusting step counts or encoder readout alone. Do not claim an accuracy smaller than the uncertainty of the measuring system.

Five-test protocol

  1. Static baseline. Measure reference positions at representative gantry/head/couch states (for example 0°, 90°, 180° and 270° where applicable). Record commanded position, independently measured position and error vector.
  2. Repeatability & hysteresis. Return to the same target position repeatedly from the same direction, then from the opposite direction. Compare spread, maximum deviation and the directional offset that reveals backlash or compliance.
  3. Load & sag. Repeat the baseline unloaded and with the intended representative mass. Measure angle-dependent displacement rather than assuming a rigid structure.
  4. Thermal drift & settling. Repeat one or more reference poses from cold start through a defined operating interval. Separate immediate settling from slower thermal drift.
  5. Correct & re-measure. Change one cause at a time—mounting, bearing preload, stiffness, backlash adjustment, direct-position sensing or software compensation—then repeat the identical protocol and report the before/after delta.

Record with every run

FieldWhy it matters
Build / configuration versionMakes the result reproducible after hardware or software changes.
Reference instrument + calibrationPrevents display resolution from being mistaken for measurement accuracy.
Commanded pose + approach directionExposes backlash and direction-dependent error.
Load, temperature / warm-up stateSeparates structural and thermal effects.
Measured displacement / angleThe physical outcome, independent of the controller.
Repeat number + timestampAllows repeatability, drift and settling to be distinguished.
Intervention since baselineLinks an improvement to a specific mechanical or control change.

Improvement ladder

1. Measurement first → 2. mechanics → 3. direct sensing → 4. control compensation → 5. re-validation. Improve reference calibration and camera/fiducial geometry before chasing tiny mechanical errors; then address stiffness, bearings, backlash and cable forces; add direct position sensing where useful; only then apply software compensation maps. Every improvement returns to the same independent measurement baseline.

Educational engineering research only. Clinical LINAC commissioning and quality assurance require the appropriate commissioned instrumentation, procedures, tolerances and institutional programme.

LOCAL-ONLY MEASUREMENT NOTEBOOK

Accuracy & Metrology Notebook

Enter one measurement series at a time. Calculations happen only in this browser tab; nothing is sent to the site. Use one unit per series so the statistics remain meaningful.

Runs0
Mean signed error—
Repeatability SD—
Maximum |error|—
#ReferenceCommandedMeasuredErrorDirectionLoadTemp °CElapsed min
No measurements yet.

PRACTICAL TRACK · TREATMENT GEOMETRY

Predict the edge, build it, then measure the disagreement

Before the build cards below, use Treatment Geometry to calculate ideal finite-source penumbra and field projection. Then use the Penumbra Box to vary source size, source-to-collimator distance and collimator-to-screen gap one at a time. Record the measured transition and compare it with the ideal prediction rather than forcing agreement.

  • Small vs large source: isolate effective-source-size effect
  • Near vs far screen: test penumbra growth with gap
  • Move the final edge: test the source-to-collimator denominator
  • Replace a straight edge with shaped apertures: connect the experiment to jaws, blocks and MLC concepts

Penumbra Box

A simple visible-light experiment showing how source size and geometry create beam penumbra. Complexity: beginner. Estimated cost: €5–15. Build time: 1–2 hours.

Ethos-Style Collimator Demonstrator

A rotating light-beam model showing why different treatment angles require differently shaped apertures. Complexity: intermediate, with 3D printing. Provisional cost: €20–50. Build time: 4–8 hours, excluding printing.

Programmable LINAC Mockup

A working tabletop model demonstrating beam shaping, positioning, motion control and basic quality assurance. Complexity: advanced. Provisional cost: €150–400. Build time: several days.

Robotic Arm With Camera Feedback

A camera-guided robotic experiment exploring positioning accuracy, calibration, feedback correction and repeatability. Complexity: advanced. Provisional cost: strongly dependent on the selected robot and cameras. Build time: several days.