Nested target volume delineation for external beam radiotherapy planning
Target volume delineation begins long before any contour is drawn. A planning CT scan — acquired with the patient immobilized in treatment position — provides the electron-density map needed for dose calculation. But CT alone often cannot distinguish tumor from adjacent soft tissue. MRI and PET are registered ("fused") onto the CT to add anatomical and biological detail that CT cannot see on its own.
CT provides the geometric and electron-density backbone of every radiotherapy plan — it is required for dose calculation and is acquired with the patient in the exact immobilization device used for treatment. But CT has poor soft-tissue contrast: a prostate tumor, a brain glioma, or a liver lesion can be nearly invisible against surrounding normal tissue.
MRI contributes superior soft-tissue discrimination (e.g. T2-weighted sequences for prostate capsule and neurovascular bundles, or contrast-enhanced T1 for brain tumor margins). PET-CT contributes functional/metabolic information — regions of high glucose uptake (FDG-PET) or other tracers can reveal viable tumor within an area of post-treatment fibrosis or atelectasis that looks identical to tumor on CT alone.
Each modality is acquired in the same or a reproducible position and then mathematically aligned ("registered") to the planning CT — rigidly for bony anatomy, or deformably for soft, mobile organs.
Fusion error propagates directly into every downstream volume. If CT-MRI registration is off by 3 mm, that error is silently baked into the GTV contour and inherited by the CTV and PTV — which is why registration accuracy is audited before contouring begins.
Rigid registration applies a single translation + rotation to align two image sets — appropriate for rigid structures like the skull or pelvis bones, where relative anatomy does not change between scans. Deformable image registration (DIR) instead computes a spatially varying displacement field, needed when organs shift, fill, or change shape between scans (bladder and rectum filling, tumor shrinkage over a treatment course, respiratory motion).
Registration quality is checked visually (bony landmarks, organ contours) and, where available, quantitatively (target registration error at fiducial points, Dice similarity of paired structures). Poor fusion is one of the most under-appreciated sources of geometric uncertainty in the entire treatment chain — it is easy to trust a smooth, plausible-looking fused image without confirming true landmark alignment.
PET tracer uptake (most commonly ¹⁸F-FDG) helps distinguish viable tumor from necrosis, atelectasis, or fibrosis — especially valuable in lung cancer, where tumor often abuts collapsed lung of identical CT density. Auto-contouring thresholds (e.g. 40% of SUVmax, or a fixed SUV of 2.5) are used as a starting point, but PET has limited spatial resolution (~4–6 mm) and is subject to partial-volume effects that can under- or over-estimate true tumor extent — so PET-derived boundaries are always interpreted alongside CT/MRI anatomy, never used as the sole determinant of the GTV.
The Gross Tumor Volume is the first and most concrete of the target volumes: the palpable, visible, or otherwise clinically demonstrable extent of malignant tumor, as defined in ICRU Report 50 (1993) and refined in ICRU 62 and 83. It is contoured directly on the fused imaging dataset, slice by slice, with no added margin.
GTV encompasses the primary tumor, and — when present and being treated — grossly involved regional lymph nodes or distant metastases, each typically contoured as separate, explicitly labeled GTV structures (GTV-T for the primary, GTV-N for nodes). It is defined by what can actually be seen or felt: a mass on CT, an enhancing lesion on MRI, an FDG-avid focus on PET, or a palpable abnormality on clinical exam.
A post-surgical patient with no residual visible disease has, by definition, no GTV — treatment planning then proceeds directly to a CTV built around the surgical bed and pattern of local recurrence risk. Where the GTV would sit is instead called the "tumor bed."
Multiple studies contouring the same patient's GTV independently across different radiation oncologists show striking disagreement — commonly 20–40% variability in volume, and substantially higher in anatomically complex or infiltrative disease such as head & neck cancer, glioblastoma, or lung tumors abutting atelectasis. This variability often exceeds the geometric uncertainty contributed by machine setup error or organ motion combined.
Mitigations include: structured contouring atlases and consensus guidelines, peer/multidisciplinary tumor-board review of contours before planning, auto-segmentation tools (atlas-based and, increasingly, deep-learning) to provide a consistent starting point, and PET/MRI fusion to reduce ambiguity at tumor-normal tissue interfaces.
In several published head & neck contouring studies, GTV volumes drawn by different experienced radiation oncologists for the same patient varied by more than a factor of two — a reminder that observer variability, not machine precision, is often the dominant source of geometric uncertainty in radiotherapy.
On the simulation display, the GTV is drawn as a closed, irregular contour that hugs the true tumor boundary as closely as the imaging allows — it is never smoothed into a simple circle or ellipse, because real tumors are irregular, and every millimeter of the GTV boundary determines where the CTV expansion begins. The contour is typically reviewed and edited slice-by-slice in axial, then verified in sagittal and coronal reconstructions to catch discontinuities between slices.
Tumors do not end cleanly where they stop being visible. Microscopic tumor cells routinely extend beyond the imageable GTV edge along tissue planes, lymphatics, and perineural spaces. The Clinical Target Volume adds a margin — often anatomically informed and non-uniform — to statistically capture this subclinical extension while respecting natural barriers to tumor spread.
The CTV margin is not a measured biological boundary — it is a population-derived estimate of how far microscopic disease is likely to extend, based on historical whole-mount pathology specimens, patterns of local recurrence, and surgical/autopsy series for each tumor type. Typical clinical CTV margins:
• Prostate: the CTV for the primary is often the entire gland (± seminal vesicles by risk group), since microscopic extraprostatic extension is diffuse rather than directional; margin beyond the visible MRI lesion is not meaningfully described in millimeters the way it is for other sites. • Head & neck: 5–10 mm around the primary GTV to cover submucosal and perineural spread, expanded further and specifically along named nerves or fascial planes at high risk, plus separate elective nodal CTVs covering lymphatic drainage basins at risk even with no visible nodal disease. • Lung: CTV margin around the primary is often modest (comparatively little microscopic extension beyond a well-defined solid tumor), but the larger geometric challenge for lung is respiratory motion, which is handled by the related Internal Target Volume (ITV) concept at the next stage.
Because microscopic spread does not respect a uniform radius, an experienced contourer expands the CTV asymmetrically: further along fascial planes, vascular sheaths, or directions of known lymphatic drainage; less in directions of low biological risk. Critically, CTV expansion is cropped at anatomical barriers that genuinely impede tumor spread — cortical bone, air cavities, and uninvolved fascial planes are not included in the CTV even where a naive uniform-radius expansion would cross into them, since tumor cells cannot realistically infiltrate several millimeters of dense cortical bone or an air-filled cavity in the time frame relevant to treatment.
A uniform-radius expansion is a computational convenience, not a biological one. Real CTV contouring is an anatomically reasoned edit: grow generously where disease plausibly extends, and stop hard at bone, air, or an uninvolved compartment boundary.
Of all target volumes, the CTV is the one most directly tied to tumor control probability — if the CTV under-covers true microscopic disease, that disease can recur regardless of how precisely the plan later delivers dose to the volumes built around it. This is why CTV delineation guidelines are among the most extensively studied and consensus-driven aspects of radiation oncology, with site-specific atlases published for head & neck, gynecologic, gastrointestinal, and other disease sites to reduce the inter-observer variability discussed at the GTV stage.
Even a perfectly contoured CTV will not receive its intended dose if the patient is not positioned identically every day, or if the target itself moves between and during fractions. The Planning Target Volume adds a geometric, largely uniform margin around the CTV so that the treated (and imaged) region reliably covers the CTV despite these uncertainties.
Setup and motion uncertainty separate into two statistically distinct components:
• Systematic error (Σ): a consistent, patient-specific offset that recurs across the whole treatment course — e.g. a mis-calibrated laser alignment, a planning CT taken with the bladder in a non-representative fill state, or a consistent immobilization error. This shifts the entire dose distribution relative to the CTV in the same direction every day. • Random error (σ): day-to-day variation around that systematic offset — small differences in patient positioning, breathing pattern, or organ filling from fraction to fraction, which tend to average out over a multi-fraction course but still blur the effective dose distribution.
Van Herk et al. (2000) derived the now-standard margin recipe M = 2.5Σ + 0.7σ, calibrated so that the CTV receives at least 95% of the prescribed dose in 90% of patients — systematic error is weighted far more heavily than random error because it does not average out over a fractionated course.
Systematic error is weighted roughly 3.5× more heavily than random error in the van Herk formula (2.5Σ vs 0.7σ) precisely because a consistent daily offset erodes CTV coverage every single fraction, while random error partially cancels out over the course of treatment.
ICRU 62 introduced an intermediate volume, the Internal Target Volume (ITV), for sites where organ or tumor motion is large and can be directly imaged and quantified — most notably lung tumors, which move with respiration. A 4D-CT scan captures the tumor position across the breathing cycle; the ITV is the union of the CTV contoured at every phase, directly encoding the true physical excursion rather than an assumed statistical margin. The PTV setup margin is then added on top of the ITV, so the final expansion is: CTV → ITV (motion, from 4D imaging) → PTV (setup uncertainty on top of ITV).
PTV margins are not fixed constants — they are a direct, quantifiable function of how well setup error is measured and corrected. Weekly portal imaging alone leaves large uncorrected systematic error, requiring generous (often 10 mm+) margins. Daily image guidance — cone-beam CT, kV/MV portal matching, surface-guided tracking, or implanted fiducial markers — measures and corrects the actual daily setup error before each fraction, collapsing the systematic component (Σ) toward zero and allowing PTV margins to shrink substantially, commonly by roughly half. This is the direct clinical payoff of IGRT: tighter margins mean less normal tissue irradiated for the same CTV coverage, which is why margin reduction is a primary justification for IGRT infrastructure investment.
A radiotherapy plan is never judged on target coverage alone. Every normal structure near the treatment volume — bladder, rectum, spinal cord, lungs, parotids, and others depending on site — must be contoured and, where it also moves or shifts, given its own Planning organ at Risk Volume (PRV) margin, mirroring the CTV→PTV logic for critical normal tissue.
Organs at risk (OARs) are just as susceptible to daily setup variation and physiological motion as the tumor target — a rectum or bladder that shifts between simulation and treatment, or a spinal cord whose position relative to the treatment isocenter carries the same setup uncertainty as the target. ICRU Report 83 formalized the Planning organ at Risk Volume (PRV): an OAR expanded by a margin analogous to the CTV→PTV expansion, ensuring that dose-volume constraints are evaluated against a volume that realistically accounts for where the organ could be on any given treatment day — not just where it was on the planning scan.
PRV margins are typically smaller than PTV margins for serial, sharply dose-limited structures like the spinal cord (where even a small underestimate of position risks a serious toxicity), and are applied more selectively for large, less motion-sensitive parallel organs.
Once OARs and PRVs are contoured, the treatment planning process becomes a negotiation between two competing goals: deliver a tumoricidal dose across the entire PTV, while keeping dose to nearby OARs below thresholds associated with acceptable toxicity rates. Published normal-tissue dose constraint compendia (such as the QUANTEC effort) summarize dose-volume relationships from decades of clinical data — for example, limiting spinal cord maximum dose to reduce myelopathy risk, or limiting the rectal volume receiving high dose to reduce late rectal bleeding risk after prostate radiotherapy. These constraints are evaluated on the PRV, not the raw OAR contour, whenever a PRV has been defined — a stricter, more conservative standard than checking the OAR alone.
The entire GTV → CTV → PTV chain exists to answer one question with a number: "how much normal tissue must be irradiated to reliably cover the tumor?" Every added margin trades geometric safety against normal-tissue dose — which is precisely why reducing uncertainty (better imaging, better immobilization, daily IGRT) is clinically valuable, not just technically elegant.
Before a plan is approved for treatment, the full structure set — GTV, CTV, PTV, all relevant OARs and PRVs — is reviewed together in three dimensions, typically by both the treating radiation oncologist and a second independent reviewer, often within a peer-review/tumor-board setting. Reviewers check: contour continuity slice-to-slice, appropriate cropping at anatomical barriers, plausible margin symmetry/asymmetry, correct structure naming per institutional or national convention, and that no OAR is inadvertently included inside the PTV. This final review is the last opportunity to catch a contouring or margin error before it becomes a delivered — and irreversible — dose to the patient.