Core–penumbra mismatch on CT perfusion imaging — quantifying salvageable brain tissue to select stroke patients for reperfusion therapy
CT perfusion (CTP) turns a routine CT scanner into a dynamic physiology tool. By injecting a bolus of iodinated contrast and imaging the brain repeatedly over roughly 60–90 seconds, the scanner captures how quickly and how much contrast reaches every cubic millimeter of tissue. From these time-density curves, dedicated software (RAPID, Viz.ai, syngo.via, Olea) derives voxel-wise maps of cerebral blood volume (CBV), cerebral blood flow (CBF), mean transit time (MTT), and Tmax — the physiological fingerprint of a stroke.
Each CT voxel is scanned repeatedly as contrast passes through it. Iodinated contrast is radio-opaque, so tissue Hounsfield units rise transiently as the bolus arrives, peak, and wash out. Plotting attenuation versus time for every voxel produces a time-density curve (TDC).
Deconvolution of each tissue TDC against an arterial input function (measured from a large feeding artery, typically the MCA or ACA) yields the tissue residue function — an estimate of how contrast would behave if delivered as an instantaneous bolus. From the residue function, four canonical parameters are derived:
• CBV (cerebral blood volume): area under the residue function — total blood volume per 100g tissue (mL/100g) • CBF (cerebral blood flow): peak height of the residue function — flow rate per 100g tissue (mL/100g/min) • MTT (mean transit time): CBV/CBF — average time blood takes to cross the capillary bed (seconds) • Tmax: time from bolus arrival to the peak of the residue function — a delay/dispersion metric, the most sensitive marker of hemodynamically significant hypoperfusion
Tmax is calculated relative to the arterial input function, so it is far less sensitive to patient-specific circulatory delay than raw arrival time — this is why Tmax, not simple time-to-peak, became the standard penumbra marker in DEFUSE 3 and DAWN.
Manual interpretation of four parametric maps under time pressure is impractical, so FDA-cleared software packages automate the pipeline:
1. Motion correction and skull stripping of the raw 4D CT series 2. Arterial input function selection (automated, from a contralateral MCA branch) 3. Deconvolution (often delay-insensitive, block-circulant SVD) to compute CBF, CBV, MTT, Tmax per voxel 4. Thresholding: rCBF <30% of normal-appearing tissue flags core voxels; Tmax >6s flags hypoperfused voxels 5. Volumetric summation with visually intuitive color overlays (red = core, green/yellow = penumbra) delivered to the stroke team in under five minutes
This automation is what makes CTP-based mismatch selection feasible in the emergency setting — a neurointerventionalist can review a single color map and a one-line summary ("Core 12 mL, Mismatch volume 96 mL, Ratio 8.0") rather than raw physiological curves.
The ischemic core is the volume of brain tissue that has already sustained flow reduction severe enough to be considered infarcted — dead or dying tissue that will not recover meaningful function even with immediate reperfusion. Accurately sizing the core is arguably the single most important measurement in acute stroke triage, because it anchors both the numerator-independent futility threshold and the denominator of the mismatch ratio.
Core segmentation is deliberately performed using relative CBF (rCBF) — the ratio of ischemic-side CBF to a mirrored region of normal contralateral tissue — rather than an absolute CBF number in mL/100g/min.
This matters because absolute CBF measurements from CT perfusion deconvolution are sensitive to arterial input function selection, cardiac output, scanner calibration, and processing algorithm. Relative thresholds normalize out much of this technical noise by using the patient's own healthy tissue as an internal reference.
The rCBF <30% threshold was empirically validated against diffusion-weighted MRI (DWI) infarct volumes and PET-defined irreversible ischemia in landmark validation studies, and is now the threshold embedded in RAPID and other FDA-cleared automated software used in DEFUSE 3 and DAWN.
It is important to understand that the CTP-defined "core" is a statistical estimate of tissue that is very likely infarcted at the time of the scan — not a deterministic biological boundary. Several caveats apply:
• Core can be overestimated early after symptom onset (before infarction is complete), especially with delayed or motion-degraded scans • Core can be underestimated in patients with excellent collateral circulation who tolerate very low flow for longer • A small fraction of core-positive tissue does recover if reperfusion is extremely rapid and collaterals are robust — this is why core is a probabilistic triage variable, not an absolute contraindication in every case • Core growth is not linear: patients with poor collaterals ("fast progressors") can convert penumbra to core within 1–2 hours, while "slow progressors" may preserve substantial penumbra for 24 hours or more — this variability is precisely why imaging-based (not clock-based) selection outperforms fixed time windows
The DEFUSE 3 and DAWN trials revolutionized stroke care by proving that patient selection based on the tissue clock (mismatch) — not the wall clock (hours since last known well) — safely extends the treatment window for thrombectomy out to 24 hours in appropriately selected patients.
Coined by Astrup, Symon, and colleagues in 1981, the "ischemic penumbra" describes brain tissue that is functionally silent but structurally intact — electrically failed but metabolically alive. On modern CT perfusion, the penumbra is operationalized as tissue with Tmax greater than 6 seconds that falls outside the core: severely delayed perfusion, but not yet below the flow threshold that predicts irreversible infarction. This is the tissue reperfusion therapy is designed to rescue.
Cerebral tissue tolerates flow reduction along a graded continuum, not a binary switch:
• Normal CBF: ~50–60 mL/100g/min • ~20–25 mL/100g/min: electrical (EEG) silence begins — the neuron stops firing but remains viable ("functional penumbra") • ~10–12 mL/100g/min: membrane ion pump failure begins — anoxic depolarization, cytotoxic edema, the threshold of irreversible injury if sustained • <10 mL/100g/min: rapid, often irreversible cell death within minutes
The penumbra occupies the zone between the electrical-failure threshold and the membrane-failure threshold — tissue that has stopped functioning but has not yet crossed into structural death. Its fate is time- and collateral-dependent: the longer flow stays this low, the more penumbra converts to core.
Multiple perfusion parameters were tested as penumbra markers historically (MTT, time-to-peak, CBF-based mismatch), but Tmax emerged as the best-validated threshold because:
• Tmax is derived from a delay-corrected deconvolution, making it robust to variability in bolus timing and arterial input function selection between patients and scanners • Tmax >6s has been extensively validated against PET oxygen extraction fraction (the gold-standard physiological definition of penumbra) and against follow-up infarct growth in randomized trial cohorts • Longer Tmax thresholds (>8s, >10s) identify progressively more severely hypoperfused — and less likely to survive — subregions of the penumbra, which is used in some centers for more granular risk stratification • Collateral flow directly determines how much Tmax>6s tissue exists: robust leptomeningeal collaterals can sustain large penumbral volumes for many hours, which is the biological basis for extended-window thrombectomy eligibility
Once core and penumbra volumes are segmented, two simple arithmetic quantities translate a color-coded map into an actionable triage number: the mismatch ratio (penumbra volume divided by core volume) and the absolute mismatch volume (penumbra volume minus core volume). Both were prospectively validated as inclusion criteria in the DEFUSE 3 trial, and both are now embedded directly into the automated reports radiologists and neurointerventionalists review in the hyperacute setting.
Mismatch ratio = Penumbra volume / Core volume Absolute mismatch = Penumbra volume − Core volume
The ratio captures relative salvageable burden — useful when core is very small, since even modest penumbra volumes produce large ratios that flag a strongly favorable profile. The absolute mismatch captures whether there is enough salvageable tissue in real terms to justify procedural risk — a huge ratio built on a tiny total volume (e.g., 2 mL core, 10 mL penumbra → ratio 5.0, absolute mismatch only 8 mL) may not be clinically meaningful.
DEFUSE 3 therefore required both criteria simultaneously: ratio ≥1.8 AND absolute mismatch ≥15 mL AND core volume <70 mL. This combination avoids being fooled by either metric alone and is why modern automated CTP software reports all three numbers together rather than any single one in isolation.
An alternative selection strategy used in the DAWN trial pairs imaging core volume against clinical stroke severity (NIHSS) rather than a second imaging volume — "clinical-core mismatch." A patient with a small CTP/DWI core but a severe neurological deficit (high NIHSS) implies a large amount of functionally impaired but structurally salvageable tissue, even without directly measuring penumbra volume.
DAWN's three age- and NIHSS-stratified core thresholds (a 80-year-old with NIHSS ≥10 and core <21 mL; NIHSS ≥10 and core 31–50 mL for younger patients; NIHSS ≥20 and core 31–50 mL) achieved similarly dramatic outcome benefits to DEFUSE 3's perfusion-mismatch approach, extending the eligible window to 24 hours from last known well in both trials.
DEFUSE 3 (perfusion mismatch, 6–16h) and DAWN (clinical-core mismatch, 6–24h) were stopped early for overwhelming efficacy — both showed roughly triple the rate of functional independence at 90 days for thrombectomy versus medical management alone in imaging-selected patients.
The final step converts a mismatch profile into a concrete treatment recommendation. A favorable mismatch — adequate penumbra relative to a limited core — supports proceeding with mechanical thrombectomy (and, in appropriate candidates, IV thrombolysis) even well outside the traditional 4.5–6 hour windows. An unfavorable profile, dominated by a large, already-completed core, signals that reperfusion is unlikely to improve outcome and may increase hemorrhage risk — the essence of imaging-guided futility screening.
A pragmatic bedside decision tree, mirroring how automated CTP software presents its summary:
1. Is core volume ≥70 mL? → Core too large: even successful recanalization is unlikely to meaningfully improve outcome and hemorrhage risk rises; treatment is often deferred or reserved for exceptional cases (young patient, very short time from onset, borderline core).
2. Is core <70 mL AND mismatch ratio ≥1.8 AND absolute mismatch ≥15 mL? → Favorable mismatch: proceed with thrombectomy (large-vessel occlusion) and/or thrombolysis per eligibility, regardless of how many hours have elapsed within the validated window.
3. Is core small but mismatch criteria only partially met (borderline ratio or absolute volume)? → Consider extended-window treatment on a case-by-case basis, weighing clinical severity, collateral status, and patient-specific factors; multidisciplinary discussion is often warranted.
"Time is brain" remains true, but CT perfusion mismatch reframes it: it is not the clock since symptom onset that determines eligibility, it is the tissue clock — how much penumbra a patient's individual collateral circulation has managed to preserve. Two patients at the same elapsed time can have completely different treatment eligibility.
Mismatch-based selection is powerful but imperfect:
• Software packages differ slightly in exact thresholding algorithms, causing modest inter-vendor variability in reported core/penumbra volumes for the same patient • Core can be systematically overestimated very early after onset or with poor-quality/motion-degraded acquisitions, potentially excluding patients who would actually benefit • Mismatch criteria were validated primarily for anterior-circulation large-vessel occlusion; applicability to posterior circulation and medium-vessel occlusion strokes is less well established • Newer trials and registries continue to explore whether even patients with unfavorable core profiles by current thresholds (e.g., large core 70–150 mL) may still benefit from thrombectomy (e.g., ANGEL-ASPECT, SELECT2, TENSION), suggesting the futility boundary may shift further as evidence accumulates