28. Oncology CT

Anatomic staging by computed tomography is the operationalization of three biological behaviors of solid tumors — local invasion, lymphatic dissemination, and hematogenous metastasis — into the TNM framework codified by the AJCC/UICC and refined across successive editions. The physician must read each CT finding as a surrogate for one of these behaviors and understand the quantitative thresholds that govern category assignment. The T category encodes primary-tumor extent: size (for many parenchymal and visceral tumors), depth of mural invasion (gastrointestinal and urothelial tumors), and breach of organ boundaries into fat, adjacent organs, or serosal/pleural surfaces. CT detects the T-stage signatures of invasion as obliteration of the normal fat plane between tumor and an adjacent structure, irregular spiculation of the serosal interface, and direct contiguity with loss of an intervening cleavage plane. A central interpretive caution is that peritumoral inflammation and desmoplasia are isodense with tumor on CT and routinely overstage T-category, so fat-plane effacement is sensitive but not specific for true invasion — the dominant overcall in T-staging.

The N category rests on the recognition that lymphatic spread follows predictable, organ-specific drainage chains, and CT identifies nodal disease principally by size, with morphologic adjuncts. The conventional threshold is a short-axis diameter exceeding 10 mm for most body regions (the short axis is used because it is less affected by oblique sectioning and node orientation), but region-specific cutoffs apply — for example, retrocrural nodes are pathologic above roughly 6 mm and a gastrohepatic-ligament node above 8 mm. The fundamental limitation is that size is a poor surrogate for metastatic involvement: micrometastases reside in normal-sized nodes (false negatives) and reactive hyperplasia enlarges benign nodes (false positives), so CT nodal staging has only moderate accuracy and is the principal reason FDG-PET/CT is incorporated for many tumors (lung, lymphoma, head-and-neck, esophageal). Morphologic features that raise the posterior probability of malignant nodal involvement independent of size include rounded rather than reniform shape, loss of the fatty hilum, central necrosis (a strong sign, especially in squamous-cell and papillary thyroid carcinoma), irregular margins, and clustering.

The M category is dichotomous and decisive: the presence of distant metastasis defines stage IV and reframes intent from curative to systemic/palliative for most solid tumors, which is why M-staging carries the highest error cost. The CT search pattern is therefore comprehensive and territory-based — liver and adrenals (the commonest abdominal metastatic sites), lungs, skeleton (bone-window review is mandatory), peritoneum, and distant nodal stations — and is detailed in the Metastatic Disease section. The Bayesian discipline of staging integrates the pretest probability set by histology and primary site: a hypervascular pancreatic neuroendocrine tumor demands arterial-phase liver scrutiny for hypervascular metastases, whereas pancreatic adenocarcinoma is staged on a pancreatic-protocol (late-arterial plus portal-venous) study for the hypovascular tumor–vessel relationship (SMA/celiac/SMV–PV abutment versus encasement) that determines resectability. Failure modes are both technical and cognitive: wrong-phase or under-opacified acquisition hides hypervascular hepatic and hypodense pancreatic disease; suboptimal bowel/oral preparation simulates or masks serosal and peritoneal spread; volume averaging and respiratory misregistration obscure small lesions; and the recurring cognitive trap is satisfaction of search — completing the T- and N-stage and stopping before a comprehensive M-stage survey, the single error most likely to misdirect curative-intent therapy.

Restaging is the comparison of disease extent at a defined point in the treatment course against a prior study, and its single governing principle is that apparent change must reflect tumor biology rather than acquisition variability. The interpretive value of every restaging CT is bounded by the comparability of the two studies being compared, so the physician's first responsibility is technical: the restaging examination should replicate the contrast phase, iodine dose and injection rate, scan delay, slice thickness, reconstruction kernel, and field of view of the baseline. Protocol drift is the dominant source of spurious measurement change — a lesion measured on a late-arterial baseline and a portal-venous follow-up can appear to grow or shrink purely because conspicuity and apparent margins vary with enhancement timing, and a thicker reconstruction blurs small-lesion borders so that the same tumor yields a different caliper reading. This is why oncologic imaging programs standardize protocols and why deviations should be explicitly flagged in the report.

The timing of restaging is dictated by the therapeutic question. An interim (mid-treatment) restage assesses early response to guide continuation or escalation; an end-of-treatment restage establishes the new baseline against which subsequent surveillance is measured; and a pre-operative restage after neoadjuvant therapy assesses resectability and downstaging. Each demands a correctly-timed baseline: the post-treatment baseline is to oncologic surveillance what the early-postoperative scan is to the postsurgical brain — the anchor that gives all later comparisons meaning. A treacherous corollary is that the choice of baseline itself encodes assumptions; the nadir (smallest tumor burden achieved) becomes the reference for declaring later progression under most response frameworks, so correctly identifying the nadir study is as important as measuring the current one.

Restaging interpretation must also account for the post-therapeutic milieu, because treatment transforms the appearance of both tumor and normal tissue. Cytotoxic and radiation therapy induce necrosis, cavitation, fibrosis, and fat deposition; surgery and ablation create cavities, seromas, and reactive enhancement; and these changes can both mimic residual disease and obscure it. Radiation fibrosis, for instance, retracts and distorts tissue planes and can mask or simulate recurrent tumor within the treated field, while post-chemotherapy hepatic steatosis lowers liver attenuation and can render isoattenuating metastases conspicuous or, conversely, blunt the conspicuity of hypovascular lesions. The mature restaging report therefore separates three categories explicitly: measurable disease (compared quantitatively), non-measurable disease (assessed qualitatively for unequivocal change), and treatment-related change (named as such rather than miscounted as tumor). Failure modes crystallize here: technically, non-comparable phase/technique and slice-misregistration fabricate change; cognitively, anchoring on a single index lesion while ignoring discordant findings elsewhere, alliterative bias (uncritically propagating a prior report's measurements and impression), and failure to compare against the correct nadir are the recurring errors. Disciplined restaging is thus an exercise in controlling variability before attributing change — the prerequisite for the formal response frameworks that follow.

The Response Evaluation Criteria in Solid Tumors, version 1.1 (RECIST 1.1, Eisenhauer 2009) is the standardized, semi-quantitative language by which CT-measured tumor burden is converted into a categorical response, and physician fluency in its explicit mechanics is essential because trial endpoints, drug approvals, and many treatment decisions are expressed in its terms. RECIST operationalizes burden through unidimensional measurement: at baseline, the reader designates up to five target lesions (maximum two per organ) that are measurable — a non-nodal lesion is measurable at a longest diameter $\ge 10$ mm on contiguous $\le 5$ mm slices, while a lymph node is measurable and pathologic at a short-axis diameter $\ge 15$ mm (nodes between 10 and 15 mm are non-target, and below 10 mm are non-pathologic). Crucially, lymph nodes contribute their short axis to the sum of diameters (SoD), whereas non-nodal targets contribute their longest diameter — a frequent source of error.

Response is then defined by change in the SoD relative to two references — baseline and nadir — according to fixed thresholds, summarized below.

Category	Target-lesion criterion	Reference	Additional requirement
Complete response (CR)	Disappearance of all target lesions; any pathologic node regresses to short axis $<10$ mm	—	No new lesions; non-target CR
Partial response (PR)	$\ge 30\%$ decrease in SoD	Baseline SoD	No unequivocal progression of non-target disease; no new lesions
Progressive disease (PD)	$\ge 20\%$ increase in SoD and an absolute increase of $\ge 5$ mm	Nadir SoD	Or any new lesion, or unequivocal non-target progression
Stable disease (SD)	Neither PR nor PD	Nadir SoD	—

The $\ge 20\%$-plus-5-mm rule for progression is a deliberate guard against declaring progression on small absolute changes that fall within measurement error: a SoD rising from 12 mm to 15 mm is a 25% increase but only 3 mm absolute, and therefore is not progression. Two further RECIST 1.1 refinements matter clinically: progression can be declared on the unequivocal appearance of a new lesion or unequivocal progression of non-target disease even when target lesions are stable, and a previously normal-sized node that grows is judged against the 10-mm threshold while its short axis is summed only when it qualifies as a target. The framework explicitly defines non-measurable disease (leptomeningeal disease, ascites, pleural/pericardial effusions, lymphangitic carcinomatosis, blastic bone lesions without a soft-tissue component, and ill-defined masses) that cannot serve as targets but is followed qualitatively.

The quantitative discipline of RECIST presupposes the reproducibility addressed in Restaging: the same lesions, measured on the same plane and phase, with caliper placement at the lesion's true longest dimension. The failure modes are characteristic and were specifically clarified by the RECIST committee (Schwartz 2016): selecting non-reproducible or cavitating/necrotic lesions as targets, measuring nodes on long rather than short axis, inconsistent lesion selection between readers, and overcalling progression on measurement noise. RECIST also has acknowledged biological blind spots — it does not capture density/necrosis (relevant for GIST and anti-angiogenic therapy), cavitation, or metabolic change — which is precisely why the alternative frameworks in the next section exist. Despite these limitations, RECIST 1.1 remains the lingua franca of solid-tumor response, and the physician must apply its numeric thresholds exactly rather than impressionistically.

Response assessment must be matched to the mechanism of action of the therapy, because the relationship between tumor biology and CT-measured size differs fundamentally across drug classes — and applying size-based RECIST indiscriminately produces systematic errors. Cytotoxic chemotherapy and radiation kill proliferating cells and shrink tumors, so size reduction (RECIST) is a reasonable surrogate, though even here CT captures secondary signatures: necrosis and central hypoattenuation, cavitation of pulmonary metastases (classic after anti-angiogenic and some cytotoxic agents, where a shrinking cavitating lesion can paradoxically increase in overall diameter), calcification of treated lesions (osteosarcoma, mucinous tumors), and hepatic pseudocirrhosis — a nodular, retracted, capsular-scalloped liver morphology that develops after chemotherapy for diffuse hepatic metastases (notably breast cancer) and which, with portal hypertension and ascites, can be misread as progression rather than treatment effect.

Anti-angiogenic and molecularly targeted agents frequently devascularize and necrose tumor without immediately shrinking it, decoupling viability from size. The paradigm is gastrointestinal stromal tumor (GIST) treated with imatinib: responding lesions lose attenuation (becoming hypodense and cystic) and lose enhancement long before they shrink, and may transiently enlarge as they necrose. RECIST therefore underestimates response, and the Choi criteria were developed specifically for GIST, defining partial response as a $\ge 10\%$ decrease in unidimensional size or a $\ge 15\%$ decrease in tumor attenuation (Hounsfield units) on contrast-enhanced CT — an explicit, quantitative incorporation of density that better predicts outcome than size alone. A clinically critical mimic in this setting is the hyperattenuating nodule-within-a-cyst (or 'nodule in a mass'), an enhancing focus reappearing within a previously hypodense treated GIST that signals acquired resistance and focal progression even when overall size is stable or decreasing.

Immune-checkpoint inhibitors introduce the most profound divergence from conventional response. Because antitumor immune activation causes lymphocytic infiltration, tumors may transiently enlarge or new lesions may appear that subsequently regress — pseudoprogression — so a single CT showing increased burden cannot reliably distinguish true progression from immune-mediated inflammatory enlargement. This drove development of immune-specific criteria: the original immune-related response criteria (irRC, Wolchok 2009), later harmonized as irRECIST/iRECIST (Seymour 2017). The defining mechanic of iRECIST is the concept of unconfirmed progressive disease (iUPD), which must be confirmed (iCPD) on a follow-up scan, typically 4–8 weeks later, before progression is declared in a clinically stable patient — new lesions are recorded but reset the bar rather than automatically ending assessment. Additional immunotherapy-specific patterns the physician must recognize include dissociated (mixed) response (some lesions regress while others grow), durable response persisting after therapy cessation, and the ominous hyperprogression (paradoxical acceleration of growth on immunotherapy associated with poor prognosis). Metabolic frameworks complement these: PERCIST (Wahl 2009) standardizes FDG-PET response by SUL (lean-body-mass-corrected SUV), capturing metabolic response that precedes size change. Finally, immunotherapy generates immune-related adverse events that are CT diagnoses in their own right — colitis (diffuse or segmental bowel-wall thickening, mucosal hyperenhancement, mesenteric stranding), pneumonitis (ground-glass and organizing-pneumonia patterns), hypophysitis, thyroiditis, hepatitis, and sarcoid-like mediastinal/hilar lymphadenopathy that can be mistaken for nodal progression. Failure modes: technically, applying the wrong framework (size-only RECIST to a devascularizing GIST or an immunotherapy-treated melanoma) systematically misclassifies response; cognitively, anchoring on size, failing to confirm immune-related progression per iRECIST, and overlooking immune-related adverse events as alternative explanations for new CT findings are the dominant errors.

Surveillance for recurrence is the post-treatment search for re-emergent tumor against a correctly-timed baseline, and its design is tumor- and stage-specific because the probability, timing, and pattern of recurrence vary by histology and the curative modality employed. The unifying interpretive problem is that the treated field is biologically transformed — surgery leaves scar, anastomoses, reconstructed planes, and cavities; radiation leaves fibrosis and volume loss; ablation leaves a non-enhancing coagulation cavity — and recurrent tumor must be distinguished from these expected post-therapeutic changes, which both mimic and conceal it. The decisive discriminators are enhancement, growth over serial studies, and morphology: mature fibrosis and scar are typically non-enhancing or only faintly/uniformly enhancing, retractile, and stable or shrinking over time, whereas recurrence tends to be a nodular, mass-like, enhancing soft-tissue focus that grows on serial imaging. Because a single time point is often indeterminate, temporal behavior on serial studies is frequently the most reliable sign, and a stable post-treatment appearance over successive intervals argues strongly against active recurrence.

The patterns of recurrence are predictable and should structure the search. Local recurrence arises at the surgical margin, anastomosis, tumor bed, or resection-cavity wall — for example, presacral soft-tissue nodularity after rectal-cancer resection (where post-radiation presacral fibrosis is the principal mimic, and increasing bulk or new enhancement over time, sometimes adjudicated by PET, favors recurrence), or enhancing nodularity at a hepatic ablation-zone margin signalling incomplete treatment. Regional nodal recurrence appears in the expected drainage stations and is judged by the size and morphologic criteria of Staging. Distant (metastatic) recurrence follows the hematogenous patterns of Metastatic Disease. Peritoneal recurrence — critical after ovarian, gastric, appendiceal, and colorectal cancer — manifests as omental caking, peritoneal and serosal nodularity, and new or increasing ascites, and is among the most easily overlooked patterns on CT.

The Bayesian framing of surveillance integrates the risk and timing distribution specific to the tumor: aggressive histologies recur early and prompt intensive early surveillance, whereas some carcinomas (e.g., certain breast and renal cancers) recur late, mandating prolonged follow-up. The pretest probability of any new finding being recurrence rises steeply with shorter interval from treatment, higher initial stage, and positive margins, and falls in the long disease-free survivor — context that should temper the overcall of an indeterminate nodule. Adjunctive data sharpen the posterior: rising tumor markers (CEA in colorectal cancer, CA-125 in ovarian cancer, PSA in prostate cancer, thyroglobulin in thyroid cancer) raise suspicion and often trigger or direct imaging, and FDG-PET/CT helps adjudicate scar versus viable tumor when CT is equivocal. Failure modes: technically, fibrosis isodense with tumor and beam-hardening or susceptibility from surgical clips/hardware obscure marginal recurrence, while suboptimal phase/contrast hides hypervascular or hypovascular recurrent deposits; cognitively, the dominant errors are misattributing recurrence to post-treatment change (or vice versa) on a single study, satisfaction of search after identifying one site while neglecting the peritoneum or skeleton, alliterative bias from prior reports, and failure to weight the tumor-specific recurrence-risk prior. Mature surveillance is therefore the disciplined comparison of serial studies against a correct baseline, with explicit naming of expected post-therapeutic change and deferral to temporal behavior and metabolic adjuncts where anatomy alone is indeterminate.

Hematogenous and lymphatic dissemination produce organ-specific metastatic signatures, and a systematic, compartment-by-compartment search pattern combined with Bayesian use of the known or suspected primary is the core competence of oncologic CT. The liver is the most common visceral metastatic site and the canonical illustration of how vascular physiology dictates appearance: metastases derive their blood supply from the hepatic artery while normal parenchyma is predominantly portal, so hypovascular metastases (colorectal, lung, most pancreatic and gastric adenocarcinomas, transitional-cell carcinoma) are most conspicuous on the portal-venous phase as hypodense lesions relative to enhanced liver, whereas hypervascular metastases (neuroendocrine tumors, renal cell, thyroid, melanoma, choriocarcinoma, some breast) enhance transiently on the late-arterial phase and may become isodense and invisible on portal-venous imaging — making phase selection the determinant of detection. A peripheral continuous-rim or target appearance (an enhancing viable rim around a hypodense necrotic center) is characteristic of metastatic deposits.

The lungs are seeded hematogenously, producing well-circumscribed nodules concentrated in the basal and peripheral (subpleural) zones where pulmonary arterial flow is greatest; a miliary micronodular pattern suggests highly vascular primaries (thyroid, renal, melanoma, choriocarcinoma), cavitation favors squamous-cell metastases, calcified metastatic nodules suggest osteosarcoma or mucinous adenocarcinoma, and lymphangitic carcinomatosis (smooth and nodular interlobular septal and peribronchovascular thickening) is a distinct, non-measurable lymphatic pattern most common with breast, lung, and gastric primaries. Nodal metastases follow the size and morphologic criteria detailed in Staging. The skeleton is the third great metastatic compartment and demands dedicated bone-window review: osteolytic deposits (renal, thyroid, lung, melanoma, multiple myeloma) appear as lucent destructive foci, osteoblastic deposits (prostate, treated breast, carcinoid, some lymphoma) as dense sclerotic foci, and mixed patterns (breast) as both — with the recognition that lytic disease and pathologic fracture risk drive management. The peritoneum is seeded transcoelomically (ovarian, gastric, appendiceal, colorectal, pancreatic), producing omental caking, serosal and mesenteric nodularity, scalloping of visceral surfaces, and malignant ascites — a compartment requiring deliberate scrutiny of the omentum, paracolic gutters, pelvic cul-de-sac, and subdiaphragmatic spaces.

The adrenal gland deserves special treatment because the incidentally or intentionally discovered adrenal nodule in a cancer patient is a recurring diagnostic crossroads, and CT provides a quantitative, validated tool to separate the common benign lipid-rich adenoma from metastasis. On unenhanced CT, an attenuation $\le 10$ HU is diagnostic of a lipid-rich adenoma with high specificity. For indeterminate (>10 HU) lesions, a dedicated adrenal washout protocol (unenhanced, 60–75 s enhanced, and 15-minute delayed phases) exploits the rapid contrast washout of adenomas relative to the slow washout of metastases and other non-adenomatous lesions. The absolute percentage washout (APW) and relative percentage washout (RPW) are computed as $\text{APW} = \frac{E - D}{E - U}\times 100\%, \qquad \text{RPW} = \frac{E - D}{E}\times 100\%,$ where $U$, $E$, and $D$ are the unenhanced, enhanced (portal-venous), and 15-minute delayed attenuations. An APW $\ge 60\%$ or RPW $\ge 40\%$ indicates a lipid-poor adenoma, whereas lower washout values are indeterminate and raise concern for metastasis (or pheochromocytoma), warranting further work-up. A critical caveat is that washout thresholds are not validated for hyperenhancing lesions such as pheochromocytoma, which can mimic adenoma washout. The Bayesian synthesis of metastatic interpretation is explicit: the pretest probability of a finding representing metastasis depends on the primary's known dissemination behavior and the patient's stage, and multiplicity strongly favors metastatic over a second primary or benign process. The oligometastatic state — historically a limited burden on the order of $\le 3$–$5$ lesions — has acquired direct therapeutic weight, since the randomized SABR-COMET trial (Palma 2019) demonstrated a survival benefit from stereotactic ablative radiotherapy added to standard care in oligometastatic disease; accurate enumeration of lesion number and total burden on CT therefore changes management, not merely prognosis. Failure modes: technically, single-phase or wrong-phase acquisition hides hypervascular hepatic and adrenal disease, omission of bone-window review misses skeletal metastases, and inadequate bowel/peritoneal evaluation misses serosal spread; cognitively, satisfaction of search (stopping after the first metastasis), anchoring an adrenal nodule as benign without applying the HU and washout criteria, and ignoring multiplicity as a discriminator are the recurring, high-cost errors.