Imaging & Diagnosis

1. Imaging in Acute Stroke — The Goal

From the moment a suspected stroke patient enters the emergency department, imaging must answer three operational questions in succession. Every downstream therapeutic decision — thrombolysis, thrombectomy, hemicraniectomy, BP target, ICU admission — flows from these answers.

The same triad governs decision-making in every modern stroke pathway. Note the ordering: a haemorrhage on NCCT halts the LVO-and-perfusion track entirely. The guiding principle (Powers et al., AHA 2019; ESO 2021): imaging should never delay treatment — door-to-needle < 45 min, door-to-puncture < 90 min — but it must be detailed enough to identify the ~10–15% of patients in the late window who can still benefit from reperfusion.

2. Non-Contrast CT (NCCT)

The first scan in nearly every stroke pathway. NCCT is fast (< 1 min acquisition), ubiquitous, and exquisitely sensitive for acute blood. Its weakness — the limitation that drives the rest of this chapter — is poor sensitivity for early ischaemic change.

Early ischaemic signs on NCCT

Tissue density follows the classical Hounsfield relation \(HU = 1000 \cdot (\mu_{tissue} - \mu_{water})/\mu_{water}\). Early ischaemic oedema raises water content by ~1% per hour, dropping density ~2.5 HU/h — right at the limit of human visual perception. This is why experienced readers consistently outperform novices on NCCT and why automated tools (e-ASPECTS, Brainomix) are increasingly deployed at the front door.

3. ASPECTS — Alberta Stroke Program Early CT Score

Barber et al. (Lancet 2000) introduced ASPECTS to quantify early ischaemic change on NCCT in anterior circulation stroke. The MCA territory is divided into 10 standardised regions — one point is subtracted for each region showing hypodensity or grey-white loss. A normal scan scores 10; a scan with diffuse MCA hypodensity scores 0.

The 10 regions, scored on two axial slices:

Inter-rater reliability is moderate (κ ~0.5–0.7) and improves with automated software (e-ASPECTS, RAPID ASPECTS). The score remains the most widely used quantifier of early ischaemic change on NCCT and is built into virtually every modern thrombectomy trial inclusion criterion.

4. CT Angiography (CTA)

CTA is the principal LVO-detection tool. A timed iodinated contrast bolus (~50–80 mL at 4–5 mL/s) is acquired during peak arterial enhancement, from aortic arch through circle of Willis. It answers four questions essential for thrombectomy planning.

Multiphase CTA (Menon, Radiology 2015) acquires three sequential phases (peak arterial, mid-venous, late-venous) and dramatically improves collateral assessment. It is now standard at most comprehensive stroke centres and approximates a poor-man’s CT perfusion when full CTP is unavailable.

Contrast safety: contrast-induced nephropathy in modern protocols is ~2–5% even at moderately reduced eGFR; ESO and AHA guidelines do not require eGFR before emergent stroke CTA.

5. CT Perfusion (CTP) — the flagship

CT perfusion images brain haemodynamics, not anatomy. A continuous cine acquisition is performed during a contrast bolus; voxel-wise time-density curves are generated, and a deconvolution against an arterial input function (AIF) yields quantitative perfusion maps. CTP is the modality that most powerfully extends the therapeutic window beyond 6 hours.

The four standard perfusion maps

The central indicator dilution relation: \(CBF = CBV / MTT\) (the central volume principle, Meier & Zierler 1954). The voxel time-density curve \(C(t)\) is the convolution of the AIF with the tissue residue function \(R(t)\):

Deconvolution recovers \(R(t)\), from which Tmax (the time at which \(R\) peaks) and CBF (its peak value) are read directly. CBV is the integral \(\int C(t)\,dt / \int AIF(t)\,dt\). The apparent diffusion coefficient on DWI MRI obeys an analogous Stejskal–Tanner relation \(S = S_0 e^{-b \cdot ADC}\) (see section 6).

Core vs penumbra — the operational definitions

Automated processing — RAPID, Olea, Viz.ai

The 2018 thrombectomy trials would have been impossible without automated CTP post-processing. Three platforms dominate clinical practice:

6. MRI in Acute Stroke

MRI is more sensitive than CT for ischaemia (DWI sensitivity ~95% within minutes), detects posterior-fossa stroke far better, and is the only modality that distinguishes haemorrhage age. Its weaknesses are availability, scan time (~10–15 min for a full stroke protocol), and patient screening (pacemaker, claustrophobia).

The acute stroke MRI protocol

DWI — the diffusion sequence

DWI exploits Brownian motion of water. Stejskal–Tanner gradients dephase moving spins; restricted water (in cytotoxic oedema) retains signal while freely diffusing water loses it. The signal obeys \(S(b) = S_0 \, e^{-b \cdot ADC}\), where \(b\) is the diffusion weighting (typically 1000 s/mm²) and ADC the apparent diffusion coefficient. Acute infarct ADC drops to ~400–600 × 10⁻⁶ mm²/s (normal ~800–900). The mechanism: failure of Na⁺/K⁺-ATPase (see Part III) shifts water from extracellular to intracellular space — intracellular water diffuses less freely.

DWI sensitivity for ischaemic stroke at presentation: ~95% (vs ~50% for early NCCT); specificity ~95% as well. DWI lesion volume is the imaging gold standard for ischaemic core.

DWI–FLAIR mismatch — dating the unwitnessed stroke

FLAIR signal change in an ischaemic lesion lags DWI by 3–6 h. So a positive DWI lesion without matching FLAIR change indicates ischaemia onset less than ~4.5 h ago. This is the imaging trick that allowed WAKE-UP trial (Thomalla, NEJM 2018) to treat “wake-up strokes” with alteplase: among 503 patients with unknown onset and a DWI–FLAIR mismatch, alteplase improved 90-day mRS (53% vs 42% with mRS 0–1, p=0.02), at the cost of more intracranial haemorrhages.

GRE / SWI — the haemorrhage detector

Susceptibility-weighted imaging exploits the magnetic susceptibility of paramagnetic haemoglobin breakdown products. Acute deoxy-haemoglobin and met-haemoglobin produce dramatic signal loss (“blooming”). SWI detects:

• Acute haemorrhage — comparable sensitivity to NCCT
• Cerebral microbleeds — predictor of post-thrombolysis haemorrhage and amyloid angiopathy
• Haemorrhagic transformation of infarct
• Susceptibility vessel sign — thrombus visible as a blooming linear focus in M1

When MRI first? Posterior-fossa syndromes (vertigo, ataxia, diplopia) where NCCT is blind to small infarcts; suspected stroke mimic (multiple sclerosis, encephalitis); subacute presentations > 24 h where the perfusion question is moot. Otherwise CT-first remains standard because of speed and availability.

7. The 4.5 h Window vs 6 h vs 24 h — the flagship

Modern acute stroke care is organised around three time windows, each defined by a landmark clinical trial. Imaging is the gatekeeper that determines which window a given patient occupies.

The trial-defined windows

Trial readouts — the numbers that built modern practice

Selecting from time vs selecting from tissue

The conceptual revolution of the late 2010s: stop selecting patients by the clock and start selecting them by their brain. A patient at 22 h with a 25 mL core and 90 mL penumbra has more salvageable tissue than a patient at 4 h with a 60 mL core. The first benefits from thrombectomy; the second probably does not.

The unifying generalisation of DAWN, DEFUSE-3, EXTEND and WAKE-UP: any imaging proxy for “significant living tissue downstream of an occlusion” — whether perfusion mismatch, clinical–core mismatch, or DWI–FLAIR mismatch — identifies a population that benefits from late reperfusion. The 2022–2023 large-core trials (RESCUE-Japan LIMIT, SELECT2, ANGEL-ASPECT, TENSION) extended this further by showing benefit even at ASPECTS 3–5.

8. Imaging Haemorrhagic Stroke

Haemorrhagic stroke imaging answers a different question: not “is there salvageable tissue?” but “will this haematoma expand, and what is the underlying lesion?” See Part IV for the clinical/pathological framework.

CT spot sign — predicting expansion

The spot sign on CTA (Wada, Stroke 2007) is a focus of contrast extravasation within an acute parenchymal haematoma. It identifies active bleeding into the haematoma and predicts subsequent expansion.

Quantitative spot-sign scores (number of spots, maximum dimension, density) refine the prediction. PREDICT (Lancet Neurol 2012) validated the score across centres. Practical use: triage to neurosurgical observation; trigger BP control to systolic < 140 mmHg (INTERACT-2, ATACH-2); consider haemostatic therapy in selected cases (although SPOTLIGHT and STOP-IT showed no benefit of factor VIIa).

MRI haemorrhage age — the haemoglobin lifecycle

MRI dating of intraparenchymal blood depends on the magnetic state of haemoglobin and its breakdown products. Each species has distinct T1 and T2 properties because of its electron configuration (oxy = diamagnetic; deoxy/met = paramagnetic; haemosiderin = superparamagnetic in macrophages).

Mnemonic (the “IB-ID-BD-BB-DD” sequence on T1 / T2 across the five stages): the magnetic susceptibility of iron drives the contrast change. SWI/GRE remain positive for years and reveal old microbleeds even when other sequences have normalised — an essential clue for cerebral amyloid angiopathy (lobar microbleeds) versus hypertensive vasculopathy (deep microbleeds).

9. Vascular Imaging Beyond the Acute Window

Beyond the first hours, vascular imaging answers the secondary-prevention questions: what was the source, is the parent vessel diseased, will it bleed again, will it clot again?

10. Diagnostic Workup of Cryptogenic Stroke

About 25% of ischaemic strokes remain “cryptogenic” after standard evaluation — no large-artery stenosis, no AF on standard ECG, no obvious cardioembolic source, no small-vessel disease pattern. The contemporary subcategory ESUS (Embolic Stroke of Undetermined Source) defines a non-lacunar infarct without identified cause that nevertheless looks embolic. The workup chases six classes of source.

1. Cardiac structural sources — TTE and TEE

Transthoracic echocardiography (TTE) screens for LV thrombus (post-MI, low EF), valvular vegetation, and intracardiac mass. Bubble study during TTE detects PFO with sensitivity ~70%. Transoesophageal echocardiography (TEE) is the higher-yield study for left atrial appendage thrombus, aortic arch atheroma (> 4 mm or mobile), and PFO with atrial septal aneurysm.

2. Occult atrial fibrillation — prolonged monitoring

Two trial families established that AF is the dominant occult source in cryptogenic stroke and that detection scales with monitoring duration:

Detection of clinically silent AF mandates anticoagulation (DOAC preferred) per AF guidelines, with stroke-recurrence reduction of ~64%. NAVIGATE-ESUS and RE-SPECT ESUS both showed no benefit of empirical DOACs in undifferentiated ESUS — monitoring to find AF, then anticoagulating, is the winning strategy.

3. Hypercoagulable workup

Targeted in young (< 50 y), recurrent, multi-territory, family history, or concomitant venous thrombosis. Standard panel:

• Antiphospholipid antibodies (lupus anticoagulant, anti-cardiolipin, anti-β2-glycoprotein-I) × 2 readings 12 wk apart
• Factor V Leiden, prothrombin G20210A — controversial in arterial stroke
• Protein C, protein S, antithrombin — check off-anticoagulation, off-acute-phase
• Homocysteine (modest effect)
• JAK2 V617F if polycythaemia / thrombocytosis
• Sickle-cell screening in appropriate populations

4. Vasculitis serologies

In young stroke, multifocal infarcts, or evidence of meningitis: ESR, CRP, ANA, ANCA, complement, RPR, HIV. CSF analysis and MR vessel-wall imaging if primary CNS vasculitis suspected (concentric vessel-wall enhancement). DSA may show characteristic alternating stenoses-and-dilatations. RCVS (reversible cerebral vasoconstriction syndrome) mimics PACNS but resolves in 1–3 months and lacks CSF inflammation.

5. Aortic and large-artery sources

Aortic arch atheroma (TEE plaque > 4 mm or mobile) carries ~12% annual recurrent stroke risk untreated. Cervical artery dissection (carotid or vertebral) accounts for ~20% of strokes in young adults — CTA or fat-suppressed T1 MRI of the neck shows mural haematoma.

6. PFO closure — the modern decision

Three landmark 2017–18 trials (CLOSE, REDUCE, RESPECT) established that percutaneous PFO closure plus antiplatelet beats antiplatelet alone in selected patients (< 60 y, embolic-appearing infarct, large shunt or septal aneurysm, no other source). NNT ~17 over 5 years for stroke prevention. The RoPE score (0–10) estimates the probability that the PFO is causal, guiding closure decisions.