CTCT·Academy
Lv 1
Curriculum · Pillar 1 · Imaging Science

5. Radiation Biology and Safety

In this chapter · 3 sections
  1. Radiation Physics
  2. Biological Effects
  3. Clinical Practice

🎯 Learning objectives

  • Derive absorbed dose, kerma, and effective dose from the underlying physical quantities and compute organ-equivalent and effective dose from CTDIvol and DLP using the appropriate tissue-weighting and conversion factors with correct units.
  • Distinguish the microdosimetric and molecular mechanisms by which low- versus high-LET radiation generates DNA double-strand breaks, and relate repair-pathway fidelity (NHEJ versus homologous recombination) to mutational and carcinogenic outcome.
  • Differentiate deterministic (tissue-reaction) from stochastic effects by threshold behavior, dose-response shape, and latency, and identify the CT scenarios (perfusion, interventional, repeated surveillance) capable of approaching deterministic skin and lens thresholds.
  • Critically appraise the linear-no-threshold model against the principal epidemiologic cohorts (Life Span Study, Pearce 2012, Mathews 2013, Berrington de Gonzalez 2009) and articulate where collective-dose extrapolation is and is not defensible.
  • Apply tube-current modulation, kV optimization, and iterative or deep-learning reconstruction to reduce dose while preserving diagnostic task performance, quantifying the trade-off against image noise and contrast-to-noise ratio.
  • Construct size-specific dose estimates (SSDE) for pediatric patients and justify weight-based protocol scaling using the steep age dependence of stochastic risk.
  • Formulate an evidence-based imaging plan for the pregnant or potentially pregnant patient, estimating conceptus dose against the 100 mGy deterministic threshold and counseling on stochastic risk.
  • Recognize the dominant technical artifacts (photon starvation, beam hardening, metal streak) and cognitive biases (dose creep, satisfaction of search) that degrade the dose-quality balance.

01Radiation Physics

The dosimetric description of CT begins not with the patient but with the photon fluence emerging from the bowtie-filtered x-ray beam, and proceeds through a hierarchy of physical quantities that must be kept conceptually distinct. The most fundamental is the absorbed dose DD, the expectation value of energy dεˉ\mathrm{d}\bar{\varepsilon} imparted by ionizing radiation to matter of mass dm\mathrm{d}m:

D=dεˉdm,[Gy]=Jkg1.D = \frac{\mathrm{d}\bar{\varepsilon}}{\mathrm{d}m},\qquad [\,\mathrm{Gy}\,] = \mathrm{J\,kg^{-1}}.

Absorbed dose is a point quantity defined for any material, and it is the physically real determinant of biological injury. It is operationally derived from kerma (kinetic energy released per unit mass), the energy transferred to charged particles per unit mass of the medium, K=dEtr/dmK = \mathrm{d}E_{tr}/\mathrm{d}m. Under charged-particle equilibrium and with radiative losses negligible at diagnostic energies (40–140 keV effective), DKcolD \approx K_{col}, which is why air-kerma measurements in a phantom can be converted to tissue dose through the ratio of mass energy-absorption coefficients (μen/ρ)tissue/(μen/ρ)air(\mu_{en}/\rho)_{tissue}/(\mu_{en}/\rho)_{air}. In CT the practical surrogate is the CT dose index, specifically CTDIvol\mathrm{CTDI}_{vol}, which folds the helical pitch pp into the weighted index CTDIw=13CTDI100,center+23CTDI100,periphery\mathrm{CTDI}_{w}=\tfrac{1}{3}\mathrm{CTDI}_{100,center}+\tfrac{2}{3}\mathrm{CTDI}_{100,periphery} via CTDIvol=CTDIw/p\mathrm{CTDI}_{vol}=\mathrm{CTDI}_{w}/p. The dose-length product, DLP=CTDIvol×L\mathrm{DLP}=\mathrm{CTDI}_{vol}\times L, integrates over the scanned length LL and is the quantity most directly tied to integral energy deposition and hence to stochastic risk. It is essential to internalize that CTDIvol\mathrm{CTDI}_{vol} is a property of the scanner and a reference phantom (16 or 32 cm PMMA), not of the patient; it systematically misrepresents dose in bodies much smaller or larger than the phantom, a discrepancy formalized later by the size-specific dose estimate.

Because different tissues differ in radiosensitivity, and because radiations of differing linear energy transfer (LET) produce different biological effect per gray, absorbed dose alone cannot index detriment. The equivalent dose to an organ, HT=RwRDT,RH_T=\sum_R w_R D_{T,R}, applies the radiation-weighting factor wRw_R (unity for the photons and electrons of CT, but 20 for alpha particles), yielding units of sieverts. The effective dose, the most widely quoted and most widely misused quantity in clinical radiology, sums the equivalent doses across organs weighted by the ICRP tissue-weighting factors wTw_T:

E=TwTHT=TwTRwRDT,R.E=\sum_T w_T H_T=\sum_T w_T \sum_R w_R D_{T,R}.

The wTw_T values of ICRP Publication 103 (2007) — 0.12 for lung, stomach, colon, breast, red marrow; 0.08 for gonads; 0.04 for bladder, esophagus, liver, thyroid; 0.01 for skin, bone surface, brain, salivary glands — were derived from sex- and age-averaged stochastic detriment in a reference population. This origin dictates the central caveat of expert practice: effective dose is a protection quantity designed for comparing modalities and managing populations, not a patient-specific risk metric. Applying a 70 kg reference wTw_T schema to a neonate, or quoting EE to three significant figures for an individual, is a category error. In practice EE is estimated from DLP through region-specific conversion coefficients kk (mSv·mGy⁻¹·cm⁻¹): roughly 0.0021 for the head, 0.014 for the chest, and 0.015 for the abdomen-pelvis in adults, with pediatric kk values several-fold higher because the same DLP irradiates a smaller, more radiosensitive volume. The failure mode to guard against is treating these conversions as exact; they carry uncertainties of 20–40 percent and are invalid outside the body region and age for which they were tabulated.

02Biological Effects

The biological consequences of CT exposure are entirely downstream of a stochastic physical event: the deposition of energy in or near the DNA double helix. At diagnostic photon energies the dominant initiating interactions are the photoelectric effect and Compton scattering, both of which liberate secondary electrons whose track structure determines the spatial pattern of ionization. This is the crux of microdosimetry: although CT delivers low-LET radiation in the macroscopic average, the secondary electrons produce dense clusters of ionization at the ends of their tracks, and it is these clusters that generate the biologically critical DNA double-strand break (DSB) and, more importantly, the clustered or complex lesion in which two or more breaks and base damages occur within one to two helical turns. A single gray deposits on the order of 10510^5 ionizations per cell nucleus, producing roughly 1000 single-strand breaks, about 40 DSBs, and a few thousand base lesions. The single-strand breaks and isolated base damages are repaired with high fidelity and are biologically near-silent; the clustered DSB is the lesion that matters, because its complexity defeats faithful repair.

The cell disposes of DSBs through two principal pathways whose competition governs outcome. Non-homologous end joining (NHEJ), active throughout the cell cycle and dominant in G1, directly ligates broken ends through the Ku70/80–DNA-PKcs–LIG4 axis and is fast but error-prone, frequently producing small insertions, deletions, and chromosomal translocations. Homologous recombination (HR), restricted to S and G2 where a sister chromatid is available as template, is high-fidelity but slow. The mutagenic and carcinogenic potential of a given exposure therefore depends not only on dose but on the proliferative state of the tissue — a mechanistic basis for the empirical radiosensitivity of marrow, gut crypt, and gonad. Misrepair generates point mutations and, through illegitimate joining of breaks on different chromosomes, the reciprocal translocations that activate oncogenes or disrupt tumor suppressors. The dicentric chromosome, the cytogenetic signature exploited in biological dosimetry, is a direct readout of this misrepair and rises essentially linearly with dose in the diagnostic-to-low-therapeutic range.

From these mechanisms the two classes of clinical effect diverge sharply. Deterministic effects (now termed tissue reactions) arise when cell killing exceeds the tissue's reserve capacity for repopulation; they exhibit a practical threshold below which the reaction does not clinically manifest, and a severity that escalates with dose above it. Representative thresholds — approximately 0.5 Gy for detectable lens opacity (revised downward by ICRP from the historical 2 Gy as evidence for a much lower or absent cataract threshold accumulated), 2 Gy for transient erythema, 3–6 Gy for temporary epilation, and 0.1 Gy for measurable but reversible suppression of spermatogenesis — are essentially never reached by a single diagnostic CT but are clinically pertinent to CT perfusion, repeated interventional CT, and overlapping multiphase protocols, where fixed-table skin dose can accumulate toward erythema thresholds. Stochastic effects — carcinogenesis and heritable mutation — have, in the prevailing regulatory model, no threshold: the probability (not severity) of the effect scales with dose. The linear-no-threshold (LNT) model, anchored in the atomic-bomb Life Span Study and supported in the diagnostic range by the Pearce (2012, Lancet) and Mathews (2013, BMJ) pediatric CT cohorts, posits excess relative risk proportional to organ dose. The expert must hold this model critically: at the per-examination doses of CT (typically 1–20 mSv effective), the predicted individual lifetime risk increment is small (order 10410^{-4} to 10310^{-3}) and statistically unverifiable against a baseline lifetime cancer incidence near 40 percent, while reverse causation and confounding by indication threaten the pediatric cohorts. Multiplying such per-capita risks across large populations to compute 'collective dose' attributable deaths is, in the ICRP's own current position, inappropriate; the defensible use of LNT is prospective optimization, not retrospective body-counting.

🖐️ Energy deposition, attenuation, and the metal-artifact signature

Connect the measurable HU map to the underlying attenuation physics and recognize photon-starvation/beam-hardening artifact at high-density interfaces.

real CT · interactive
Preparing interactive viewer…

Window across brain, subdural, and bone presets on a real head CT. The Hounsfield scale is a direct map of the linear attenuation coefficient μ\mu that governs where photoelectric and Compton interactions deposit the energy responsible for DNA injury. Note the streak artifact radiating from the metal electrodes: locally extreme attenuation causes photon starvation and beam hardening, the same physics that drives the dose-quality trade-off discussed in this chapter.

03Clinical Practice

Translating radiobiology into practice is an exercise in optimization under the ALARA (as low as reasonably achievable) principle, constrained always by the overriding requirement that the study answer its clinical question — an underexposed, non-diagnostic scan delivers dose for zero benefit and is the most common failure of mis-applied dose reduction. The governing relationship is that image noise (the pixel standard deviation σ\sigma) scales inversely with the square root of the photons reaching the detector, so that for fixed reconstruction σ1/D\sigma \propto 1/\sqrt{D}; halving noise costs a quadrupling of dose. Every dose-reduction lever is a negotiation against this relation. Tube-current modulation (automatic exposure control) varies mA angularly and along the z-axis to hold a target noise index constant as patient attenuation changes, typically saving 20–40 percent. Tube-potential (kV) optimization is more physically subtle and more powerful for iodinated studies: lowering kV from 120 to 100 or 80 shifts the spectrum toward the iodine K-edge (33.2 keV), sharply increasing iodine attenuation through the photoelectric effect (whose cross-section scales roughly as Z3/E3Z^3/E^3) so that contrast-to-noise ratio per unit dose rises even as raw noise increases — the basis of low-kV CTA. Iterative and deep-learning reconstruction break the historical filtered-back-projection coupling between dose and noise by modeling photon statistics and system optics, permitting 30–70 percent dose reduction at matched noise, though the expert must watch for the characteristic 'plastic' or 'waxy' texture and the potential loss of low-contrast detectability that can mask subtle lesions. Beyond these, restricting scan length to the indicated anatomy (recall DLP=CTDIvol×L\mathrm{DLP}=\mathrm{CTDI}_{vol}\times L), eliminating reflexive multiphase acquisitions, and applying organ-based current reduction or bismuth shielding for superficial radiosensitive organs complete the toolkit. The pervasive cognitive failure mode here is dose creep: the gradual upward drift of technique to obtain ever-smoother images, untethered from diagnostic necessity, which dose-monitoring registries and diagnostic reference levels exist to police.

Pediatric considerations elevate optimization from good practice to imperative, for two compounding reasons grounded in the preceding biology. First, children possess more dividing, radiosensitive stem cells and a longer post-exposure lifespan over which a stochastic clone can express, so age-averaged risk coefficients understate their true per-mGy detriment; the Pearce and Mathews cohorts quantify excess leukemia and brain-tumor and solid-cancer incidence after childhood CT, with cumulative doses of 50–60 mGy associated with roughly two- to three-fold relative risks. Second, a small body irradiated to the same CTDIvol\mathrm{CTDI}_{vol} absorbs a far higher organ dose than the 32 cm reference phantom implies — the rationale for the size-specific dose estimate (SSDE), SSDE=fsize×CTDIvol\mathrm{SSDE}=f_{size}\times\mathrm{CTDI}_{vol}, where the conversion factor fsizef_{size} derived from effective diameter can exceed 2 in neonates. Practice therefore demands weight- or diameter-based protocol scaling of mA and kV, scrupulous justification under the Image Gently framework, and aggressive collimation, with the explicit recognition that conversion coefficients kk for effective dose are several-fold higher in infants. Pregnancy considerations require the physician to reason on two distinct axes. Deterministically, the conceptus is at risk of malformation, growth restriction, and reduced IQ chiefly during organogenesis and early fetal life, but the relevant ICRP and AAPM threshold for these tissue reactions is approximately 100 mGy of conceptus dose — far above the typical fetal dose of a single body CT (commonly 1.5–25 mGy for abdominopelvic acquisitions, and sub-mGy for chest CT or pulmonary embolism studies with appropriate technique). The corollary, endorsed in the AAPM position statement, is that no individual diagnostic CT delivers a fetal dose warranting therapeutic abortion. Stochastically, however, the LNT model still applies, with a per-mGy childhood-cancer risk to the conceptus on the order of 0.006%0.006\% per mGy; this small absolute increment justifies dose minimization (low-kV, reduced length, shielding where it does not compromise the study) and, where an equivalent non-ionizing test such as ultrasound or MRI answers the question, substitution — but it must never be allowed to deny a pregnant patient an indicated, potentially life-saving CT, as in suspected pulmonary embolism or major trauma. Defensible counseling pairs an explicit conceptus-dose estimate against the 100 mGy threshold with a candid statement of the small stochastic risk relative to the much larger maternal and fetal hazard of a missed diagnosis.

🖐️ Functional CT and the dose-efficiency frontier

Appreciate that repeated same-volume acquisition (perfusion, interventional CT) is the diagnostic context most likely to approach deterministic skin-dose thresholds, motivating dose-area limits alongside ALARA.

real CT · interactive
Preparing interactive viewer…

A real CT perfusion colour map. Perfusion protocols repeatedly scan one slab over time, so fixed-position skin dose accumulates and can approach deterministic erythema thresholds — the clearest routine example of where stochastic optimization and deterministic safety limits converge. The diagnostic yield of functional maps must be weighed against this integral dose.

Check your understanding

8 questions
  1. 1.

    A 32 cm reference-phantom CTDIvol of 8 mGy is reported for an abdominopelvic CT in a small adult whose effective diameter is well below 32 cm. Compared with the true average dose absorbed by this patient, the reported CTDIvol most likely:

    med
  2. 2.

    Effective dose (E) is frequently quoted to counsel an individual patient about the risk of a specific CT. The most rigorous objection to this practice is that effective dose:

    hard
  3. 3.

    Among the lesions produced when 1 Gy is deposited in a cell nucleus, which is the principal determinant of carcinogenic and lethal outcome, and why?

    med
  4. 4.

    A patient undergoes a brain CT perfusion study. Which radiation effect is the most relevant safety concern specific to this protocol, and what governs its appearance?

    hard
  5. 5.

    Lowering tube potential from 120 kV to 80 kV for a CT angiogram improves iodine contrast-to-noise ratio per unit dose despite increasing image noise. The dominant physical reason is that:

    hard
  6. 6.

    A pregnant patient with suspected pulmonary embolism is being counseled. The estimated conceptus dose from a modern CT pulmonary angiogram is well under 1 mGy. Which statement reflects current ICRP/AAPM guidance?

    med
  7. 7.

    Why are children assigned higher per-mGy stochastic risk coefficients than adults for an identical organ dose?

    med
  8. 8.

    Deep-learning and iterative reconstruction permit substantial dose reduction at matched noise compared with filtered back projection. What is the principal expert caution when relying on them?

    hard
Answer all questions to submit.

🌐 Keep exploring — Radiopaedia & more

Hand-picked, free external references to deepen this topic.

References & primary literature

  1. 1.The 2007 Recommendations of the International Commission on Radiological Protection. ICRP Publication 103. Ann ICRP. 2007;37(2-4). (Defines effective dose, tissue-weighting factors w_T, and the basis for deterministic/stochastic distinction and the LNT model.)
  2. 2.Pearce MS, Salotti JA, Little MP, et al. Radiation exposure from CT scans in childhood and subsequent risk of leukaemia and brain tumours: a retrospective cohort study. Lancet. 2012;380(9840):499-505.
  3. 3.Mathews JD, Forsythe AV, Brady Z, et al. Cancer risk in 680,000 people exposed to computed tomography scans in childhood or adolescence: data linkage study of 11 million Australians. BMJ. 2013;346:f2360.
  4. 4.Berrington de Gonzalez A, Mahesh M, Kim KP, et al. Projected cancer risks from computed tomographic scans performed in the United States in 2007. Arch Intern Med. 2009;169(22):2071-2077.
  5. 5.AAPM Position Statement on Radiation Risks from Medical Imaging Procedures (PS 4-B). American Association of Physicists in Medicine, 2023. (States that risks from doses below ~50-100 mGy are too low to be detectable and that individual diagnostic CT doses do not justify pregnancy termination.)
  6. 6.Image Gently Alliance. Society for Pediatric Radiology, ACR, ASRT, AAPM. (Framework and protocols for pediatric CT dose optimization, size-based technique scaling, and justification.)
  7. 7.Boone JM, Strauss KJ, Cody DD, et al. Size-Specific Dose Estimates (SSDE) in Pediatric and Adult Body CT Examinations. AAPM Report No. 204. College Park, MD: AAPM; 2011.
  8. 8.National Research Council. Health Risks from Exposure to Low Levels of Ionizing Radiation: BEIR VII Phase 2. Washington, DC: National Academies Press; 2006. (Authoritative derivation of low-dose stochastic risk and the LNT framework underpinning CT risk estimation.)

Tip: use ← / → to move between chapters.