17. Infection

Pyogenic bacterial infection is, to a first approximation, an exercise in neutrophil-mediated liquefaction, and computed tomography images the consequences of that process rather than the organism itself. Once a critical inoculum of pyogenic bacteria — staphylococci, streptococci, Gram-negative enterics — overwhelms local defenses, complement activation (C3a, C5a) and macrophage-derived cytokines (TNF-$\alpha$, IL-1, IL-8) establish a chemotactic gradient that floods the focus with neutrophils. The respiratory burst and degranulation that kill bacteria also autodigest host tissue, and the released lysosomal hydrolases convert solid parenchyma into a semifluid slurry of dead neutrophils, liquefied cellular debris, and protein-rich exudate: pus. This is the central event the radiologist must keep in mind, because liquefactive necrosis is what produces the two cardinal CT properties of established bacterial infection — a fluid-attenuation centre and a vascularized, enhancing margin.

The prototypical evolution is best understood in the brain, where the histopathology has been correlated with CT in classic stereotactic series. In the earliest phase (roughly days 1–3, early cerebritis) there is a poorly marginated zone of edema and microvascular congestion that appears as ill-defined low attenuation — hypodense relative to normal white matter, typically in the vasogenic-edema range of roughly $10$–$22\,\mathrm{HU}$ — with little or only patchy enhancement because the blood–brain barrier breakdown is incomplete. By late cerebritis (days 4–9) a necrotic centre develops and a rim of neovascular granulation tissue with leaky, immature capillaries begins to enhance, but the rim is thick, irregular, and shows centripetal filling-in on delayed images. The defining transition is capsule formation (early capsule days 10–14, late capsule beyond two weeks): fibroblasts recruited from the adventitia of cortical vessels lay down a collagenous wall, and the mature abscess displays the textbook thin, smooth, uniform ring of intense enhancement surrounding a near-fluid centre and girdled by vasogenic edema. The wall is characteristically thinner on its deep (ventricular) aspect because relative hypoperfusion of white matter retards fibroblast delivery, an asymmetry that both explains the propensity for daughter-abscess rupture into the ventricle and gives a useful directional clue on CT.

Elsewhere in the body the same biology produces homologous signatures: a renal or hepatic pyogenic abscess shows a low-attenuation core, often containing locules of gas where gas-forming organisms are present, surrounded by an enhancing rim and a perilesional rim of hyperemia, while bacterial lobar pneumonia produces dependent consolidation with air bronchograms as alveoli fill with the same neutrophilic exudate. The expert diagnostic move is quantitative and kinetic rather than purely morphologic. Pus is non-enhancing centrally (a post-contrast rise of less than $\sim10\,\mathrm{HU}$ in the core) and, where dual-energy or perfusion data exist, the necrotic centre shows no measurable iodine uptake and near-zero cerebral blood volume, in contrast to the elevated relative CBV of a hypervascular neoplastic rim. Bayesian reasoning sharpens this: in a febrile patient with leukocytosis, recent dental work, or right-to-left shunt, a thin smoothly ring-enhancing lesion is far more likely abscess than glioblastoma, whereas in an older patient with weight loss the prior shifts toward malignancy. The principal mimics — necrotic glioblastoma and metastasis (thicker, nodular, irregular rims), resolving hematoma, and demyelinating tumefactive plaques (open-ring enhancement) — are separated by rim morphology, the presence or absence of central restricted diffusion on correlative MRI, and clinical priors. The dominant failure modes are technical (partial-volume averaging of a small rim mimicking a solid nodule; beam-hardening from adjacent bone obscuring a juxtacortical collection) and cognitive (anchoring on tumour in a patient with known cancer, missing the subtle early-cerebritis hypodensity). Prognostically, mature encapsulation is what makes percutaneous or surgical drainage feasible and is the imaging prerequisite for source control; failure to recognize gas, rim breakdown, or extension toward an ependymal surface portends ventriculitis, the most lethal complication.

Viral infection injures CT-visible tissue by a fundamentally different mechanism from pyogenic bacteria: rather than recruiting a neutrophil flood that liquefies parenchyma, viruses inflict cell-level cytopathic damage and provoke a lymphocyte- and interferon-driven response that alters the alveolar-capillary interface, the airway epithelium, or the neuron, producing diffuse, often bilateral attenuation changes without the discrete fluid collections of bacterial disease. In the lung — the dominant compartment for clinically important viral CT — the target is the type I and type II pneumocyte and the adjacent capillary endothelium. Viral replication and the host cytokine response (type I and III interferons, IL-6) increase alveolar-capillary permeability, allowing proteinaceous fluid and a few cells to seep into the interstitium and partially fill alveoli. Because the alveoli are only partially filled and their walls thickened, X-ray transmission is reduced but vessels and bronchial walls remain visible through the haze: this is the cellular basis of ground-glass opacity (GGO), the signature of early viral pneumonitis, typically peripheral, subpleural, and basal in distribution.

The temporal trajectory follows the underlying pathology of diffuse alveolar damage. In the exudative phase (roughly the first week) bilateral, multifocal, peripheral GGO predominates. As protein-rich fluid coats alveolar walls and interlobular septa become edematous and infiltrated, the superimposition of smooth septal thickening on a background of GGO yields the crazy-paving pattern. With progression to the organizing phase (second week and beyond), intra-alveolar fibroblastic plugs (Masson bodies) and consolidation appear, often peripheral and band-like, frequently with the architecture of organizing pneumonia, and may evolve into the reversed halo (atoll) sign — a central GGO ring rimmed by denser consolidation — which reflects organizing inflammation peripherally with sparing or fibrinous filling centrally. When alveolar collapse and early fibrosis supervene, traction bronchiectasis and architectural distortion signal a worse prognostic class. This stereotyped exudative-to-organizing sequence, validated extensively during the SARS-CoV-2 pandemic, is why a single time point on CT must always be interpreted against the day of symptom onset; identical GGO carries very different meaning on day 2 versus day 18.

In the central nervous system, neurotropic viruses produce a parallel but distinct signature. Herpes simplex encephalitis, the most consequential, has a tropism for the limbic system: viral cytolysis and an intense lymphocytic, often hemorrhagic, necrotizing inflammation in the medial temporal lobes, insula, and cingulate gyrus produce gyral swelling and low attenuation that, crucially, is frequently subtle or even occult on early CT — CT sensitivity in the first 48 hours is poor, and a normal CT must never be allowed to lower suspicion when the clinical picture (fever, altered mentation, seizures) is consistent. The Bayesian and management implications are sharp: because empiric aciclovir is low-risk and untreated HSV encephalitis is devastating, the appropriate response to a normal or equivocal CT is MRI and lumbar puncture, not reassurance — a deliberate refusal to let a negative test reduce posterior probability. The principal pulmonary mimics of viral GGO are non-infective: hydrostatic pulmonary edema (more central, with effusions and cardiomegaly), hypersensitivity pneumonitis, drug toxicity, and alveolar hemorrhage, separated by distribution, clinical context, and the company the GGO keeps. The dominant failure modes are technical — respiratory motion and dependent atelectasis mimicking or obscuring GGO, and overly wide window settings that visually erase faint ground-glass — and cognitive: satisfaction of search after identifying one lobe of GGO, and premature anchoring on a single pathogen during an outbreak when co-infection or an alternative diagnosis coexists.

Fungal infection occupies a mechanistic niche between the liquefactive pus of bacteria and the diffuse permeability changes of viruses, and its CT signatures are best understood as the imaging of vascular occlusion and infarction. The clinically dominant moulds — Aspergillus species and the Mucorales — are angioinvasive: their hyphae penetrate vessel walls, incite local thrombosis, and produce a wedge of ischemic, then hemorrhagic, coagulative necrosis distal to the occluded artery. This single vascular mechanism generates the most specific fungal CT sign, the halo sign: a central dense nodule of infarcted, hyphae-laden lung surrounded by a rim of ground-glass that represents alveolar hemorrhage seeping from the damaged microvasculature into the perilesional alveoli. The halo is a fleeting, high-yield early finding, most prevalent in the first days of invasive pulmonary aspergillosis in a profoundly neutropenic host, and its presence in the right clinical context should prompt pre-emptive antifungal therapy rather than await tissue confirmation.

The temporal evolution of angioinvasive fungal disease is tied intimately to the neutrophil count, and this coupling is one of the most elegant mechanism-to-image relationships in thoracic radiology. During profound neutropenia the host cannot mount a cavitating inflammatory response, so the lesion is a solid hemorrhagic infarct with a halo. As neutrophils recover — typically with marrow engraftment or count recovery around days 10–21 — the restored inflammatory and proteolytic machinery begins to liquefy and resorb the necrotic core, and the infarcted centre retracts away from the surrounding viable lung. Air dissects into the cleft, producing the air-crescent sign, a crescent of gas separating the sequestered necrotic ball from the cavity wall. Thus the air-crescent is paradoxically a sign of immune recovery and generally favorable prognosis, not of worsening infection — a counterintuitive point that must be reasoned from the underlying biology. The reversed halo sign, a focus of central ground-glass surrounded by a denser ring, is comparatively more characteristic of mucormycosis in the appropriate host, reflecting central infarction with a peripheral rim of organizing hemorrhage, though it is not specific and also occurs in organizing pneumonia.

A second mechanistic pattern is airway-invasive disease, in which fungi colonize and invade the bronchial and bronchiolar walls rather than vessels, producing centrilobular nodules and the tree-in-bud pattern as the small airways fill with inflammatory and fungal material, alongside peribronchial consolidation in invasive forms. Endemic dimorphic fungi (Histoplasma, Coccidioides, Blastomyces) and the yeast Cryptococcus produce yet other patterns — nodules that may calcify, mediastinal and hilar adenopathy, and chronic cavitary disease in those with underlying lung architecture distortion — overlapping substantially with tuberculosis and malignancy. The expert separates these by integrating host immune status, geography and exposure, tempo, and morphology. The crucial Bayesian principle is that the halo and air-crescent are powerful only when conditioned on the correct pretest probability: in a neutropenic leukaemic patient the halo is near-diagnostic of invasive fungal disease, whereas an identical halo around a haemorrhagic metastasis or vasculitic nodule in an immunocompetent patient means something entirely different. The dominant mimics are haemorrhagic metastases, granulomatosis with polyangiitis, septic emboli, and tuberculosis; the failure modes are cognitive (failing to weight neutropenia and thereby missing the diagnostic window for the transient halo) and technical (thick slices and motion blurring small centrilobular nodules and the subtle halo into invisibility). Because angioinvasive fungal disease can erode into pulmonary vessels and cause fatal haemoptysis, and because mucormycosis demands urgent surgical debridement and reversal of immunosuppression, the CT recognition of these vascular signs is directly load-bearing for survival.

Opportunistic infection is defined less by a particular organism than by a particular host, and its central radiological lesson is that the immune deficit rewrites the imaging rules by attenuating, distorting, or abolishing the inflammatory response that CT normally depends upon to render disease visible. Because the CT findings of infection are, in essence, a map of the host reaction — edema, neutrophil influx, capsule formation, granuloma — a patient who cannot mount that reaction produces a fundamentally muted or atypical image. The expert therefore reasons backward from the specific defect: the type of immunosuppression predicts both the spectrum of likely organisms and the morphology they will produce.

The most instructive coupling is between neutropenia and the suppression of suppuration. A profoundly neutropenic patient cannot wall off bacteria into a classic ring-enhancing abscess, cannot generate dense pyogenic consolidation, and cannot cavitate; bacterial and fungal foci therefore appear as ill-defined hypodense lesions with thin or absent enhancing rims and minimal surrounding inflammatory stranding, so that the very features the radiologist relies upon for confident diagnosis are precisely the ones that are missing. This is the mechanistic reason invasive fungal disease in neutropenia presents as the solid halo nodule described earlier, with cavitation deferred until count recovery. In defects of cell-mediated immunity — advanced HIV with low CD4 counts, solid-organ and allogeneic stem-cell transplantation, and biologic therapies that block TNF or T-cell signaling — the predicted pathogens are intracellular and granuloma-controlled: Pneumocystis jirovecii, reactivated tuberculosis and non-tuberculous mycobacteria, cytomegalovirus, Toxoplasma, Cryptococcus, and Nocardia. Pneumocystis pneumonia exemplifies the mechanism-to-image link: organisms fill alveoli with foamy proteinaceous exudate while the impaired immune system limits dense consolidation, producing bilateral, perihilar, geographic ground-glass opacity that classically spares the periphery and may evolve thin-walled cysts (pneumatoceles) prone to pneumothorax. CD4-stratified Bayesian reasoning is explicit and clinically codified: oesophageal candidiasis, PCP, and disseminated fungal disease cluster below $200\,\mathrm{cells/\mu L}$, while CMV retinitis, disseminated Mycobacterium avium complex, and CNS lymphoma and toxoplasmosis cluster below $50\,\mathrm{cells/\mu L}$, so the CD4 count functions as a prior that reorders the entire differential.

Intracranially, the toxoplasmosis-versus-lymphoma problem in advanced HIV is a paradigm of conditioned probability. Both produce ring-enhancing lesions with edema; toxoplasmosis tends toward multiple lesions favoring the basal ganglia and grey-white junction (an eccentric target sign on MRI), whereas primary CNS lymphoma is more often solitary, periventricular, and subependymal, and — diagnostically decisive — is hypermetabolic on FDG-PET and shows increased perfusion, in contrast to the hypometabolic, hypoperfused toxoplasma abscess. A defining clinical-radiological phenomenon unique to this population is the immune reconstitution inflammatory syndrome (IRIS): when antiretroviral therapy or marrow engraftment restores immune competence, the newly functional immune system reacts violently against residual antigen, producing paradoxical worsening — enlarging nodes, new or enlarging ring-enhancing lesions, worsening consolidation — that mimics treatment failure but is in fact recovering immunity. Recognizing IRIS prevents the catastrophic error of escalating antimicrobials or abandoning effective therapy. The governing failure modes are cognitive above all: applying immunocompetent expectations to an immunocompromised patient and dismissing a subtle hypodensity because it lacks the expected rim or edema, anchoring on a single opportunist when co-infection is common, and interpreting IRIS as progression. Technically, low-attenuation lesions in muted hosts demand narrow windows and careful contrast timing, and the temptation to limit contrast in renally impaired transplant patients can erase the only marker — faint rim enhancement — that betrays an active focus. The management stakes are high because these infections disseminate rapidly and because the correct response (antimicrobial escalation, reduction of immunosuppression, or, in IRIS, corticosteroids while continuing therapy) depends entirely on correctly reading the host context behind the image.

Abscess formation is the culmination and the unifying theme of pyogenic infection, and it deserves treatment as a process in its own right because the radiologist's task is frequently not to ask whether infection is present but whether a collection has matured into a drainable abscess. An abscess is the end state of liquefactive necrosis successfully contained by the host: a central cavity of pus — dead neutrophils, digested tissue, and proteinaceous fluid — walled off by a pyogenic membrane of granulation tissue and, with time, fibrous capsule. The biology of containment is the engine of the imaging appearance. Activated macrophages and the cytokine milieu drive angiogenesis (VEGF-mediated) and fibroblast proliferation at the periphery of the necrotic focus, generating a richly vascularized, leaky-capillaried wall. On CT this wall is the source of the cardinal sign — a rim of intense, persistent enhancement — while the avascular, liquefied centre neither enhances nor perfuses.

The quantitative signatures are reproducible across organs and worth committing to memory. The central pus typically measures higher than simple fluid, commonly in the range of $\sim20$–$45\,\mathrm{HU}$ because of its high protein and cellular content, in contrast to bile, urine, or a simple cyst near water density ($0$–$15\,\mathrm{HU}$); after contrast the centre rises by less than $\sim10\,\mathrm{HU}$ (true non-enhancement), whereas the rim enhances briskly and remains enhanced on delayed phases. The single most specific finding is locules of gas within the collection, frequently with air-fluid levels, which in the absence of recent instrumentation or a fistula essentially establishes the diagnosis and implicates gas-forming organisms. On correlative MRI the central pus shows marked diffusion restriction — low apparent diffusion coefficient — because the viscous, hypercellular, macromolecule-rich content impedes water motion; this is the property that most reliably separates an abscess from the diffusion-facilitated necrotic core of a tumour and is the highest-yield discriminator when enhancement morphology is ambiguous. Where dual-energy CT is available, iodine maps confirm an avascular centre and a hyperaemic rim, and perfusion confirms near-zero central blood volume.

The table below summarizes the staged correlation that underlies confident interpretation.

Stage	Dominant cellular event	CT attenuation and enhancement	Approx. timing (CNS model)
Phlegmon / early cerebritis	Edema, microvascular congestion, early neutrophil influx	Ill-defined low attenuation (hypodense to white matter, $\sim10$–$22\,\mathrm{HU}$), little/patchy enhancement	Days 1–3
Late cerebritis / suppuration	Central liquefactive necrosis, immature leaky neovessels	Thick, irregular rim with centripetal fill-in; low-attenuation centre	Days 4–9
Early capsule	Fibroblast-laid collagen wall forming	Thin, smooth, intensely enhancing rim; non-enhancing core	Days 10–14
Mature abscess	Established collagenous capsule, contained pus	Thin uniform ring, true non-enhancing centre, surrounding edema; $\pm$ gas	>2 weeks

The central diagnostic reasoning is the separation of drainable pus from solid disease and from sterile fluid. The two great mimics are necrotic neoplasm — distinguished by a thick, nodular, irregular, often mural-enhancing rim, no diffusion restriction, and a clinical prior favoring malignancy — and a resolving haematoma, distinguished by its evolving high-attenuation blood products and lack of a true enhancing capsule. Sterile post-operative or pancreatic collections overlap heavily, and here the radiologist must weight context: enhancement, gas, wall thickness, and the clinical trajectory together, never any single sign. Bayesian prioritization is decisive — a thin-rimmed, gas-containing, restricting collection in a febrile patient with leukocytosis is an abscess until proven otherwise, while the same morphology in an afebrile patient one day after surgery may be an expected seroma or postoperative gas. The principal failure modes are technical (partial-volume averaging falsely solidifying a small abscess or falsely thinning a tumour rim; mistiming the contrast bolus so the rim is underenhanced; mistaking physiologic postoperative gas for infective gas) and cognitive (premature closure on tumour in an oncology patient, and satisfaction of search after finding one collection while a second, multiloculated component is missed). The prognostic and management payoff is direct and immediate: the recognition of a mature, walled, drainable collection is the imaging trigger for percutaneous catheter or surgical drainage and is the definition of achievable source control, while signs of rim breakdown, rapid enlargement, gas extension along fascial planes (heralding necrotizing infection), or rupture into a body cavity or ventricle convert a contained, treatable process into a surgical and sometimes life-threatening emergency.