18. Neoplasia

The diagnostic radiologist who reads an oncologic CT is, whether or not the act is made explicit, inferring a dysregulated cellular program from its macroscopic shadow. The conceptual scaffold for that inference is the hallmarks framework articulated by Hanahan and Weinberg in 2000, expanded in 2011, and broadened again in Hanahan's 2022 synthesis to include phenotypic plasticity, non-mutational epigenetic reprogramming, polymorphic microbiomes, and senescent cells. For the purposes of imaging pathobiology the operative insight is that each hallmark is a tissue-level behavior, and tissue-level behaviors are precisely what CT measures: the geometry of growth, the density of cellularity, the kinetics of perfusion, and the topology of spread. Sustaining proliferative signaling and evading growth suppressors together produce the most elementary CT correlate of malignancy — a mass that enlarges over serial studies with a measurable doubling time, which for solid tumors typically falls between roughly 40 and 400 days and is itself a coarse surrogate for grade, since high-grade lesions outrun their stroma and their blood supply. Resisting cell death, mechanistically the loss of $p53$-mediated apoptosis and the engagement of autophagic and necroptotic escape, manifests as the survival of cells at distances from a feeding vessel that normal tissue could not tolerate, so that the tumor accretes a mantle of viable cells around a core that has nonetheless outstripped diffusion and undergone coagulative necrosis. That core is the source of the low-attenuation center, frequently $0$ to $25\,\mathrm{HU}$, that the reader interprets almost reflexively as 'necrotic tumor.'

Enabling replicative immortality through telomerase reactivation removes the Hayflick brake and underwrites the relentless serial enlargement; genome instability and tumor-promoting inflammation supply the mutational fuel and the cytokine-rich, edematous, often hemorrhagic microenvironment that blurs the tumor margin and recruits the host stroma. The remaining hallmarks — inducing angiogenesis, activating invasion and metastasis, reprogramming energy metabolism, and evading immune destruction — are sufficiently central to CT interpretation that they constitute the subsequent sections of this chapter. The unifying epistemology is Bayesian. No single CT feature is pathognomonic; rather, each feature is a likelihood ratio that updates the pre-test probability set by age, risk factors, organ, and clinical context. A $2\,\mathrm{cm}$ enhancing renal mass in a $70$-year-old smoker carries a high prior for renal cell carcinoma, and avid early enhancement with washout sharply raises the posterior; the identical attenuation profile in a febrile patient with bacteremia must be weighed against an abscess, whose rim enhances but whose center is liquefactive pus rather than coagulative necrosis. The characteristic failure mode at this level of reasoning is the conflation of growth with malignancy: benign processes (organizing hematoma, inflammatory pseudotumor, progressive fibrosis) also enlarge, and indolent malignancies (well-differentiated thyroid carcinoma, some neuroendocrine tumors) may be radiologically stable for years. The expert therefore reads the hallmarks not as a checklist of signs but as competing generative hypotheses, each predicting a constellation of attenuation, enhancement, and temporal behavior, and selects the hypothesis whose predictions the image best satisfies.

Of all the hallmarks, angiogenesis most directly authors the CT image, because it is the determinant of how iodinated contrast distributes in space and time. A solid tumor that has reached roughly $1\text{–}2\,\mathrm{mm}$ exhausts the capacity of passive diffusion to supply oxygen and nutrients; beyond this threshold proliferation and necrosis reach equilibrium unless the tumor flips the angiogenic switch. The switch is a shift in the local balance between pro-angiogenic ligands — chiefly vascular endothelial growth factor A (VEGF-A), released in response to hypoxia-inducible factor $1\alpha$ (HIF-$1\alpha$) stabilization under low $pO_2$ — and endogenous inhibitors such as thrombospondin-1. The resulting neovasculature is structurally aberrant in ways that are mechanistically legible on CT. Tumor vessels are tortuous, irregularly caliber, and chaotically branched; their endothelium is loosely apposed with wide interendothelial gaps; the basement membrane is discontinuous; and pericyte coverage, which normally stabilizes and regulates capillaries, is sparse and abnormally attached. These defects make the tumor microvasculature both hyperpermeable and hemodynamically inefficient.

The macroscopic consequences are the bread and butter of multiphase contrast CT. Hypervascular tumors — hepatocellular carcinoma, renal cell carcinoma, neuroendocrine tumors, and many metastases from these primaries — recruit arterial supply and therefore enhance briskly in the late arterial phase (roughly $35\text{–}40\,\mathrm{s}$ after injection), often rising $50\,\mathrm{HU}$ or more above baseline. The same leaky, pericyte-poor vessels cannot retain contrast, and the elevated interstitial fluid pressure — a direct mechanical consequence of fluid extravasation through permeable walls combined with absent functional lymphatics — drives contrast back out, producing the portal-venous and delayed-phase 'washout' that, for hepatocellular carcinoma in an at-risk liver, is specific enough to permit a noninvasive diagnosis under the LI-RADS and AASLD frameworks. Quantification deepens this reasoning. CT perfusion models the passage of a contrast bolus to derive blood flow ($\mathrm{BF}$, $\mathrm{mL}/100\,\mathrm{g}/\mathrm{min}$), blood volume ($\mathrm{BV}$, $\mathrm{mL}/100\,\mathrm{g}$), mean transit time, and, critically, a permeability term. The modified Tofts model expresses tissue contrast concentration as

$C_t(t) = K^{trans}\!\int_0^t C_p(\tau)\,e^{-k_{ep}(t-\tau)}\,d\tau + v_p\,C_p(t),$

where $K^{trans}$ is the volume transfer constant between plasma and the extravascular extracellular space, $k_{ep}$ the efflux rate constant, and $v_p$ the plasma volume fraction. Elevated $K^{trans}$ and blood volume correlate with histologic grade, microvessel density, and VEGF expression, and their decline is one of the earliest signals of effective antiangiogenic therapy — often preceding any change in size. The principal interpretive pitfalls are technical and temporal. Mistimed acquisition is the dominant error: a hypervascular metastasis imaged only in the portal-venous phase may become isodense to enhancing liver and vanish, the classic 'flash-filling' miss, while a hypovascular pancreatic adenocarcinoma is conspicuous precisely because its desmoplastic, vessel-poor stroma enhances less than the surrounding gland and is therefore best seen in the pancreatic-parenchymal phase. Pseudo-enhancement adjacent to high-attenuation structures and beam-hardening from contrast-dense vessels can spuriously raise measured HU, and the expert confirms true enhancement only against a properly co-registered non-contrast baseline using a meaningful threshold, conventionally a rise exceeding $10\text{–}20\,\mathrm{HU}$.

The metabolic rewiring of cancer cells, the hallmark Warburg first described in the 1920s, is the biochemical bridge between the molecular phenotype and both the anatomic CT image and its functional counterpart, FDG-PET. Proliferating tumor cells preferentially metabolize glucose to lactate by aerobic glycolysis even when oxygen is abundant — a thermodynamically wasteful strategy in ATP yield per glucose but advantageous because it supplies carbon skeletons and reducing equivalents (via the pentose phosphate shunt) for the biosynthesis of nucleotides, amino acids, and lipids that a dividing cell requires. This is enforced by upregulated glucose transporters (GLUT1) and hexokinase, the same machinery that traps the glucose analogue $^{18}$F-fluorodeoxyglucose intracellularly as FDG-6-phosphate, rendering metabolic avidity directly imageable. On CT alone, metabolism is read indirectly through its structural consequences. The high nuclear-to-cytoplasmic ratio and dense cellular packing of an aggressive, glycolytic tumor reduce extracellular water and raise soft-tissue attenuation, while the same metabolic appetite, when it outpaces a deranged blood supply, produces the regional hypoxia and coagulative necrosis seen as central low attenuation. Reprogrammed lipid metabolism and de novo lipogenesis underlie the intracytoplasmic fat that, when detected as macroscopic negative-HU foci (below approximately $-10\,\mathrm{HU}$, and unambiguously below $-30\,\mathrm{HU}$), is diagnostically powerful: macroscopic fat within a renal mass is essentially diagnostic of angiomyolipoma, fat within a hepatic lesion suggests hepatocellular carcinoma or adenoma, and fat within a retroperitoneal mass points toward a well-differentiated liposarcoma.

Dystrophic calcification, the deposition of calcium phosphate in metabolically failing and necrotic tissue, is another metabolic footprint with discriminating value, and its morphology and distribution refine the differential: psammomatous (fine, punctate) calcification suggests papillary thyroid or serous ovarian carcinoma, chondroid (rings-and-arcs) calcification a cartilaginous tumor, and coarse calcification within a treated mass frequently signals response. Spectral and dual-energy CT now sharpen these metabolic readouts: virtual non-contrast images separate iodine from intrinsic calcium, iodine maps quantify the contrast that the neovasculature delivered, and material decomposition can characterize lipid and mineral content with a specificity unavailable to single-energy attenuation. The deepest interpretive synthesis is the anatomic–metabolic correlation, because CT and FDG-PET answer complementary questions. A lesion may be morphologically stable yet metabolically extinguished after therapy, or anatomically present yet representing only fibrotic scar with no viable glycolytic cells; conversely, low FDG avidity in a well-differentiated hepatocellular carcinoma, a mucinous adenocarcinoma, or a low-grade neuroendocrine tumor reflects genuinely low glycolytic flux rather than a normal finding, and is a recognized cause of false-negative PET. The reader must also recognize the metabolic mimics: brown fat, inflammation, granulomatous disease, and post-procedural healing are intensely glycolytic and FDG-avid without being neoplastic, so metabolic avidity, like enhancement, is a likelihood ratio to be integrated with morphology and clinical context rather than a categorical verdict.

Local invasion is the behavior that converts a contained neoplasm into a surgical and prognostic problem, and its CT signature is written almost entirely at the tumor–host interface. Mechanistically, invasion proceeds through a coordinated dismantling of tissue architecture. Carcinoma cells, or a subset of them, undergo epithelial–mesenchymal transition (EMT): they downregulate E-cadherin and the junctions that bind them into an epithelium, acquire a motile mesenchymal phenotype with front–rear polarity, and secrete or induce matrix metalloproteinases (notably MMP-2 and MMP-9) and other proteases that degrade type IV collagen of the basement membrane and remodel the surrounding interstitial matrix. The breach of the basement membrane is the histologic definition that separates invasive carcinoma from carcinoma in situ, and although that membrane is far below CT resolution, its breach is inferred from the macroscopic loss of the tissue boundary it once defined. The host does not remain passive: invasive epithelial tumors, pancreatic ductal adenocarcinoma being the archetype, provoke an intense desmoplastic stromal reaction — activated fibroblasts laying down dense, hypocellular, hypovascular collagen — that both stiffens the tumor and, because it enhances poorly, accounts for the characteristically hypoattenuating, ill-defined appearance of such cancers.

These mechanisms translate into a vocabulary of interface findings the radiologist uses to stage and to judge resectability. A preserved fat plane between a mass and an adjacent structure is reassuring evidence of non-invasion, whereas effacement of that fat plane, soft-tissue stranding, and a spiculated rather than smooth margin signal transgression into surrounding tissue — though here the central pitfall arises, because peritumoral inflammation and edema produce identical stranding and routinely cause over-staging. The single most consequential application is vascular involvement in pancreatic adenocarcinoma, where the circumferential contact between tumor and the celiac axis, superior mesenteric artery, and superior mesenteric vein is quantified: contact of $180^{\circ}$ or less, more than $180^{\circ}$, vessel contour deformity, and thrombosis stratify lesions along the resectable–borderline–unresectable axis that determines whether a patient is offered surgery or neoadjuvant therapy. Analogous interface reasoning governs the T-category across organs — the depth of mural invasion and breach of the serosa or perirectal/perigastric fat in gastrointestinal cancers, capsular and extracapsular extension in renal and prostate tumors, and chest-wall or mediastinal invasion in lung cancer, the last judged by the obtuse-versus-acute angle of pleural contact, the length of contact, and frank rib destruction. A capsule or pseudocapsule, conversely, is evidence of expansile rather than infiltrative growth and shifts the differential toward more indolent or benign entities; hepatocellular carcinoma's enhancing pseudocapsule is itself a recognized diagnostic feature. The expert reader, aware that desmoplasia and inflammation are indistinguishable from microscopic tumor on attenuation alone, calibrates confidence accordingly — describing vascular contact in degrees rather than asserting invasion, and reserving the term for unambiguous structural destruction.

Metastasis is the cause of the great majority of cancer deaths, and its CT patterns are not random but the predictable output of tumor cell mechanics and host anatomy. The invasion–metastasis cascade requires cells to detach, intravasate into vessels, survive anoikis and shear in the circulation as single cells or emboli, arrest in a distant capillary bed, extravasate, and — the rate-limiting step — successfully colonize the foreign microenvironment. The non-randomness of the destination has two complementary explanations. The mechanical, or hemodynamic, hypothesis holds that cells lodge in the first capillary bed they encounter, which explains the dominant venous-drainage patterns of CT: colorectal carcinoma drains through the portal vein and seeds the liver first; sarcomas and other tumors draining systemically seed the lungs through the pulmonary arterial bed; and tumors with access to the vertebral venous plexus of Batson, a valveless network in communication with the systemic veins, seed the spine and axial skeleton, as prostate and breast carcinoma characteristically do. The biological seed-and-soil hypothesis, originally Paget's, holds that successful colonization requires a compatible microenvironment, accounting for tropisms that hemodynamics alone cannot — the propensity of breast cancer for bone, brain, liver, and lung, and the role of the pre-metastatic niche prepared in advance by tumor-derived factors and exosomes.

Each dissemination route inscribes a distinct CT signature. Hematogenous spread produces multiple, typically rounded, often peripheral nodules and masses that respect the vascular distribution of the target organ — randomly distributed pulmonary nodules, and hepatic deposits whose enhancement recapitulates the vascularity of the primary, hypervascular from renal, neuroendocrine, melanoma, and thyroid primaries and hypovascular from gastrointestinal adenocarcinomas. Lymphatic spread enlarges nodes along predictable drainage chains, and the diagnostic difficulty is that nodal size is an imperfect biomarker: the conventional short-axis threshold of $1\,\mathrm{cm}$ both misses micrometastatic disease in normal-sized nodes and over-calls reactive nodal hyperplasia, so morphology (rounded shape, loss of the fatty hilum, central necrosis, irregular enhancement) and clinical context must supplement size. Lymphangitic carcinomatosis, tumor permeating the pulmonary lymphatics, produces nodular interlobular septal thickening with preserved lung architecture. Transcoelomic spread across the peritoneum, the dominant route for ovarian, gastric, and appendiceal mucinous tumors, follows the gravitational and ascitic-flow-determined sites — the pouch of Douglas, the right paracolic gutter, the subphrenic spaces, and the greater omentum, where confluent deposits form an 'omental cake' — and may be subtle, signaled only by trace ascites, fine peritoneal nodularity, or stranding of the mesenteric fat. Perineural spread along nerve sheaths, characteristic of adenoid cystic carcinoma, pancreatic adenocarcinoma, and some head-and-neck cancers, produces nerve thickening and enhancement and foraminal widening that are easily overlooked. The governing cognitive failure across all routes is satisfaction of search: having identified the dominant lesion, the reader's vigilance for additional, smaller deposits falls, and a systematic survey of liver, adrenals, peritoneum, nodal stations, and skeleton at the appropriate window is the discipline that counters it.

The assessment of treatment response is where imaging pathobiology meets therapeutics most consequentially, and where naive size-based interpretation fails most often, because modern therapies act on the biology of the tumor rather than merely shrinking it. The historical standard, RECIST 1.1, reduces response to the sum of longest diameters of target lesions: a partial response is a $30\%$ decrease and progressive disease a $20\%$ increase (with an absolute minimum increase of $5\,\mathrm{mm}$) or the appearance of new lesions. This anatomic, unidimensional framework is well suited to conventional cytotoxic chemotherapy, whose effect is cell kill and volume loss, but it systematically misreads three increasingly common biologies. Antiangiogenic and targeted agents frequently produce profound devascularization without prompt shrinkage: a gastrointestinal stromal tumor responding to imatinib drops markedly in attenuation as it becomes hypovascular and myxoid or cystic, sometimes while enlarging, and the Choi criteria — which incorporate a $15\%$ decrease in HU as a response marker alongside a $10\%$ size decrease — were devised precisely to capture this, correlating with outcome far better than size alone. Hepatocellular carcinoma treated with locoregional or systemic therapy is assessed by mRECIST and the EASL approach, which measure only the viable, arterially enhancing component and disregard the treated, non-enhancing necrotic volume. The mechanistic lesson is that loss of enhancement, cavitation in a responding pulmonary metastasis, and intratumoral hemorrhage or low attenuation can all signal response rather than progression.

The most important contemporary phenomena are the response patterns of immune checkpoint inhibitors, which unmask the limits of conventional timing and morphology. Because these agents work by activating a T-cell response, an effective response can be preceded by a transient increase in lesion size or even the appearance of new lesions as cytotoxic lymphocytes infiltrate the tumor — pseudoprogression — which on CT appears as enlargement that subsequently regresses and which, scored as progression by RECIST, would prematurely terminate effective therapy. The immune-related response frameworks (irRC, iRECIST) address this by requiring confirmation of progression on a follow-up scan after a defined interval before declaring true progression. The converse and more sinister pattern is hyperprogression, a paradoxical acceleration of tumor growth kinetics on immunotherapy seen in a subset of patients, which demands genuine progression be distinguished from the benign pseudoprogression it superficially resembles. Immunotherapy also produces a distinctive spectrum of immune-related adverse events with CT correlates the radiologist must not mistake for tumor or infection — pneumonitis (ground-glass and organizing-pneumonia patterns), colitis (bowel-wall thickening and mucosal hyperenhancement), hypophysitis, thyroiditis, and sarcoid-like reactive lymphadenopathy that is FDG-avid and mimics nodal metastasis. The expert synthesis is to read response as a change in biology, not merely in caliber: to compare against the correct prior study and on matched phases, to quantify enhancement and not only diameter, to weight new low attenuation and cavitation as potential response in the right drug context, and to demand confirmation of equivocal progression on immunotherapy before acting. The cognitive traps are anchoring on a prior 'progression' read, mismatched contrast timing that simulates enhancement change, and the failure to recognize that the same morphologic change carries opposite meanings under cytotoxic, antiangiogenic, and immune mechanisms.