1. Introduction to CT brain perfusion in acute stroke

CT brain perfusion (CTP) has become the indispensable fourth dimension of the modern stroke imaging triage battery. Where non-contrast CT (NCCT) confirms haemorrhage exclusion and CTA maps the occluded vessel, CT brain perfusion protocol reveals the physiological consequence at the tissue level — distinguishing irreversibly infarcted core from the hypoperfused but still viable ischaemic penumbra that endovascular thrombectomy can salvage.^[1] This single distinction now governs which patients with large-vessel occlusion (LVO) receive mechanical thrombectomy in the 6-to-24-hour extended window, directly governed by the DAWN and DEFUSE-3 trial selection criteria.^[2]

The protocol is demanding. It requires the lowest clinically permitted tube voltage (80 kVp) to maximise iodine contrast-to-noise ratio, a low-volume but ultra-rapid contrast bolus delivered at 6.0 mL/s, and a perfectly positioned shuttle or wide-area-detector acquisition that must not miss the middle cerebral artery (MCA) territory. Errors at any step — wrong table position, head-tilt, delayed trigger, suboptimal baseline — can generate maps that actively mislead the thrombectomy team. Given that every 15-minute delay in stroke reperfusion costs approximately 17 million neurones,^[3] there is no tolerance for scanning error or interpretive ambiguity.

This article provides the complete technical and interpretive framework for the CT brain perfusion protocol: every scan parameter, the physiological basis of each perfusion map, the top 10 detectable pathologies, and a structured three-tier pitfall analysis covering radiographer execution, radiologist interpretation, and non-radiology physician clinical management. Whether your department is establishing CTP for the first time or auditing an existing stroke imaging pathway, this reference gives you the evidence-based blueprint.

2. Anatomy and HU values in CT brain perfusion

Gross anatomical coverage requirements

The critical anatomical requirement in CTP is ensuring the selected slab covers the territory at risk — almost always the MCA distribution for anterior circulation LVO. On modern 256- and 320-detector-row scanners, a single shuttle or fixed wide-beam acquisition covers the entire supratentorial brain (approximately 80–160 mm z-axis), capturing the MCA, ACA, and PCA territories simultaneously. On narrower 16–64 detector scanners, coverage is restricted to a 2–4 cm slab, making table position selection the single most consequential technical decision of the acquisition. The radiographer must place the coverage zone to include: the basal ganglia (the most sensitive early MCA ischaemia indicator), the centrum semiovale, and — whenever posterior circulation stroke is suspected — the cerebellar hemispheres and brainstem.^[4]

Hounsfield unit reference for perfusion CT

In CTP, HU values at baseline (pre-contrast) and during the dynamic acquisition carry different meaning from standard CT. The arrival curve — the time-density curve at each voxel — is the raw data from which perfusion parameters are mathematically deconvolved. Key baseline and pathological HU reference values are tabulated below.

Understanding the four perfusion maps

CT perfusion generates four primary parametric maps through mathematical deconvolution of the time-density curve at each voxel against the arterial input function. Each map emphasises different physiological phenomena and must be interpreted together — no single map should be used in isolation to define the treatment-relevant infarct signature.^[5]

Cerebral blood volume (CBV) measures the total volume of blood within a given mass of brain tissue, expressed as mL/100 g. Normal cortical CBV is 4–6 mL/100 g. In acute infarct core, CBV collapses to below 2.0 mL/100 g as cell membranes fail and autoregulation is lost. This threshold forms the operational definition of irreversible infarction used in validated clinical trial protocols. CBV is the most specific but least sensitive early marker of core infarction — it requires true microvascular collapse and therefore may underestimate core volume in the hyperacute phase.

Cerebral blood flow (CBF) expresses the rate of blood movement through brain tissue per unit mass per minute (mL/100 g/min). Normal grey matter CBF is 50–80 mL/100 g/min. At flow below approximately 10–15 mL/100 g/min, irreversible neuronal death occurs. CBF maps generated by deconvolution are more sensitive than CBV to early ischaemia and form the preferred core map in some automated software platforms (e.g., RAPID, where relative CBF <30% of contralateral hemisphere defines core). However, CBF maps are more susceptible to motion artefact and errors in AIF selection than CBV maps.

Mean transit time (MTT) is the average time, in seconds, for blood to travel from the arterial input to the venous output through the capillary bed. Normal MTT is 4–6 seconds. Prolonged MTT identifies hypoperfused tissue broadly, encompassing the penumbra, benign oligaemia, and core — making it the most sensitive but least specific perfusion map. It should never be used alone to define treatment targets.

Time to maximum (Tmax) is the delay, in seconds, between the arrival of the contrast bolus in the arterial input and the peak of the tissue residue function. This parameter — generated specifically by deconvolution software — is the most clinically validated marker of the ischaemic penumbra. Tmax >6 seconds defines the hypoperfused but potentially salvageable region in validated stroke imaging trials including DAWN and DEFUSE-3. Tmax maps are the primary output of FDA-cleared automated software platforms and form the basis of the core-penumbra mismatch calculation used for patient selection.^[6]

Arterial input function — the hidden quality control point

All four maps depend entirely on an accurate AIF. If the AIF cursor is placed in a vessel that is occluded, partially calcified, or averaged with surrounding brain, the entire map set is mathematically invalid. Modern automated software auto-selects the AIF from vessels with the earliest, sharpest rise in the time-density curve. Radiographers and radiologists must verify AIF quality on every scan before releasing results to the clinical team. A broad, delayed, or low-amplitude AIF is a red flag that requires manual correction or rescan.

3. Scanning technique for CT brain perfusion

The CT brain perfusion protocol requires precise, step-by-step execution with zero tolerance for improvisation. The 7-step workflow below reflects best practice for both wide-detector platforms (256/320-row) and legacy 16/64-row scanners operating in shuttle mode.

1. Step 1 — Patient positioning and immobilisation

   Supine position with the head in the neutral anatomical position, chin neither elevated nor depressed. Use a head holder or foam wedge to eliminate head tilt. Apply a soft forehead restraint band if safe and tolerated. Head-tilt of more than 5° will cause the shuttle acquisition slab to clip the MCA territory asymmetrically, creating perfusion deficits that are purely artefactual. This is the single most preventable cause of diagnostic failure in CTP. Confirm with a rapid scout image (0.5 s low-dose topogram) before proceeding.

2. Step 2 — Intravenous access and injector priming

   Establish 18-gauge or larger antecubital vein access. Connect a high-pressure-rated line set verified for the flow rate of 6.0 mL/s (a minimum pressure rating of 300 PSI is required). Prime both the contrast syringe (40 mL of 370–400 mg I/mL iodinated contrast) and the saline syringe (100 mL). Perform a 2 mL test injection at 2 mL/s to confirm free IV access and exclude extravasation. Document the access site, needle gauge, and line pressure rating as per institutional protocol.^[7]

3. Step 3 — Protocol selection and slab positioning

   Select the CTP protocol from the scanner console. For shuttle-mode scanners (16–64 row), position the acquisition slab at the level of the basal ganglia and ensure it covers the MCA M1–M2 territory. Use the scout view to confirm: (a) gantry isocenter alignment with the slice plane; (b) basal ganglia centred in the coverage zone; (c) no evidence of head tilt. For wide-detector 256/320-row scanners, the entire supratentorial brain is covered automatically — verify the z-axis coverage extends from the skull base to the vertex on the topogram.

4. Step 4 — Acquisition parameters

   Set kVp to 80 kVp (mandatory for maximum iodine conspicuity at lowest dose). Tube current: 150–200 mA (auto-mAs may be applied if validated for CTP on your platform). Rotation time: 0.5 seconds. Pitch: 0 (shuttle or axial). Scan interval: typically every 1–2 seconds for the first 60–70 seconds of acquisition (dynamic phase), beginning 5 seconds after injection start (immediate dynamic toggle). Total acquisition time: 60–70 seconds to capture both the arterial first-pass and the venous washout phases required for accurate deconvolution. Kernel: smooth/standard (B20–B30 equivalent) — do not use a sharpening kernel for CTP raw data.

5. Step 5 — Injection synchronisation

   The CTP injection must be synchronised precisely with scan start. At most institutions, the contrast injection (40 mL at 6.0 mL/s, immediately followed by 100 mL saline at 6.0 mL/s) is triggered simultaneously with scan initiation, with a pre-set 5-second delay built into the acquisition — this is the “immediate dynamic toggle” trigger. Unlike bolus-tracked protocols, CTP does not use a fixed HU threshold trigger at a vessel; instead the scan begins automatically and captures the entire bolus transit from arterial input through parenchymal peak to venous outflow. Confirm with the injector technologist that the injection and scan triggers are synchronized.

6. Step 6 — Motion management during acquisition

   The 60–70 second dynamic acquisition is the most vulnerable phase for motion artefact. Agitated or confused stroke patients present a significant challenge. Measures include: verbal reassurance at regular intervals; minimal table vibration; confirmed absence of gantry tilt; and in selected patients, cautious sedation in consultation with the stroke physician. Post-processing software can apply rigid motion correction to shift each dynamic frame back into registration, but severe motion — particularly sudden head movement — cannot be corrected retrospectively and requires rescan if clinically feasible.

7. Step 7 — Post-processing and map generation

   Transfer raw dynamic data immediately to the post-processing workstation (or via automated DICOM push to the AI perfusion software platform). Generate CBF, CBV, MTT, and Tmax maps using validated deconvolution software. Confirm AIF auto-selection quality — visually inspect the AIF time-density curve for: sharp early rise, clear peak, and appropriate amplitude (>150 HU rise above baseline). Export maps to PACS, flag the study as STAT, and notify the reporting radiologist immediately. Total door-to-CTP-report time target: ≤20 minutes from patient arrival to completed map review.

Scanner comparison: coverage and protocol adaptation

Deep learning reconstruction (DLR) in CTP

CT brain perfusion at 80 kVp with 150–200 mA produces inherently noisy raw data. Traditional filtered back-projection (FBP) amplifies this noise into the dynamic dataset, degrading the accuracy of AIF extraction and time-density curve fitting. Iterative reconstruction (ASIR-V, AIDR 3D, iDose, SAFIRE) improved image quality but introduced well-documented temporal blurring artefacts that degraded perfusion map accuracy. Deep learning reconstruction (DLR) — such as TrueFidelity (GE), AiCE (Canon), and Precise Image (Siemens Healthineers) — now offers significant advantages for CTP specifically: noise reduction without temporal blurring, improved CNR at low-dose 80 kVp settings, and more accurate voxel-level time-density curve fitting.^[8] Where DLR is available on the CTP scanner, it should be applied as the default reconstruction algorithm for all perfusion series.

4. Contrast media protocol for CT brain perfusion

The contrast injection in CTP is fundamentally different from any other CT protocol. The goal is not to achieve steady-state vascular enhancement but to engineer a compact, high-concentration bolus that transits through the cerebral circulation in a single, traceable pass. Every parameter — volume, rate, concentration, and chaser — is engineered to serve the deconvolution algorithm.^[9]

Full injection protocol

Special populations and contrast adaptations

Renal impairment: In acute stroke, the marginal additional contrast risk from 40 mL of contrast must be weighed against the certainty of death or severe disability without optimal triage. Current AHA/ASA and ESR guidance confirms that acute ischaemic stroke is an emergency indication where contrast-induced nephropathy (CIN) risk should not delay or prevent CTP imaging. Ensure post-procedure IV hydration is initiated as soon as the patient reaches the ward or angiography suite.

Known allergy history: For patients with prior severe contrast reactions, rapid pre-medication (intravenous methylprednisolone and diphenhydramine) should be administered per institutional protocol if time allows. In extremis — patient actively deteriorating — the decision to proceed without pre-medication must be made by the attending physician with consent documented in the medical record.

Paediatric CTP: Weight-based dosing applies (1.5–2.0 mL/kg, maximum 40 mL). Flow rate should be adjusted downward proportionally and IV access assessed carefully. Dedicated paediatric perfusion post-processing thresholds differ from adult values and require specialist neuroradiological review.

5. Radiation dose in CT brain perfusion

CT brain perfusion is the highest-radiation CT brain protocol in clinical use. The multi-phase dynamic acquisition at 80 kVp produces effective doses substantially exceeding single-phase brain CT, reflecting the sequential repeat exposures required to track bolus transit. This dose must be contextualised against the clinical benefit: in acute LVO stroke, the information from CTP directly determines whether a patient receives a thrombectomy that can prevent permanent disability — a benefit that overwhelmingly outweighs the stochastic radiation risk from any single examination. Nonetheless, as Low As Reasonably Achievable (ALARA) principles apply, and dose optimisation is an ongoing obligation.^[10]

Diagnostic reference levels (DRLs)

5 evidence-based dose reduction strategies for CTP

The following strategies reduce CTP dose without compromising diagnostic adequacy, as validated in peer-reviewed literature:^[11]

1. Reduce temporal sampling frequency

   Acquiring CTP frames every 2–3 seconds rather than every 1 second reduces total dose by 40–60% with minimal impact on perfusion map accuracy, as validated in large datasets processed with modern deconvolution software. This strategy is particularly appropriate on 256/320-row scanners where improved CNR from wider coverage partially compensates for reduced temporal density.

2. Apply deep learning reconstruction (DLR)

   DLR enables a 30–50% reduction in tube current (mA) at 80 kVp while maintaining sufficient CNR for accurate AIF extraction and time-density curve fitting. TrueFidelity, AiCE, and ADMIRE are validated DLR platforms for CTP where dose reduction has been demonstrated without loss of perfusion map accuracy.

3. Optimise the pre-contrast baseline phase

   Acquire only 3–5 baseline frames before injection rather than extending baseline acquisition unnecessarily. Many older protocols acquired 10–15 baseline frames, contributing cumulative dose without improving deconvolution accuracy. Modern software requires as few as 3 pre-contrast frames to establish a stable baseline.

4. Truncate the acquisition at confirmed venous washout

   The dynamic acquisition can be terminated once the venous output function (VOF) in the superior sagittal sinus (SSS) has returned to near-baseline. Extending acquisition beyond this point — typically 60–65 seconds — adds dose without additional perfusion information. Programme the acquisition to stop automatically at 65 seconds maximum.

5. Implement tube current modulation within the dynamic series

   Some modern scanner platforms permit limited angular tube current modulation even within the constrained geometry of a fixed axial or short-shuttle CTP acquisition. Where available, this reduces dose to radiosensitive orbital structures by 10–20% without affecting the perfusion-relevant brain tissue sampling in the scan field.

6. Top 10 pathologies detectable on CT brain perfusion

CT brain perfusion is not a single-diagnosis protocol. While it is primarily deployed in acute ischaemic stroke triage, it detects a broad spectrum of conditions characterised by abnormal cerebrovascular haemodynamics. Each condition has a characteristic perfusion signature, and the radiologist must be familiar with all 10 to avoid anchoring on the stroke diagnosis when an alternative pathology is present.

7. Pitfalls for radiographers in CT brain perfusion

The CTP scanning pitfalls below are structured by category, with detailed descriptions and evidence-based mitigations. They reflect the highest-frequency errors encountered in clinical practice audits and reported in the peer-reviewed literature on CTP quality metrics.^[12]

8. Pitfalls for radiologists interpreting CT brain perfusion

CTP interpretation requires simultaneous integration of four parametric maps, the underlying vascular anatomy from the CTA component of the stroke triage battery, and the patient’s clinical history. Isolated map review without clinical and vascular context is a major source of diagnostic error.^[13]

9. Pitfalls for non-radiology physicians reviewing CT brain perfusion

Stroke neurologists, emergency physicians, and interventional neuroradiologists increasingly review CTP maps directly in the stroke suite prior to formal radiological reporting. This clinical empowerment is valuable — but it carries the risk of systematic misinterpretation when the clinician lacks specialist perfusion imaging training. The pitfalls below represent the most clinically consequential errors in non-radiologist CTP review.^[14]

10. Pitfall comparison summary

The following three-column overview maps the primary pitfall category for each professional role to enable rapid recognition and cross-discipline awareness. The most effective stroke imaging teams are those where all three groups understand not only their own failure modes but those of their colleagues.

11. AI and automation in CT brain perfusion

CT brain perfusion is one of the most mature areas of FDA- and CE-cleared AI deployment in diagnostic radiology. Automated perfusion post-processing — once the exclusive domain of specialist neuroradiology workstations — is now performed by cloud-based and on-premise AI platforms that deliver validated perfusion maps in minutes, directly from the scanner. This represents a transformative shift in stroke care: hospitals without 24/7 specialist neuroradiology cover can now have automated RAPID maps available for the stroke neurologist within 5–7 minutes of scan completion.^[15]

FDA-cleared and CE-marked CTP AI platforms (as of 2026)

What AI does not replace in CTP

Despite the maturity of automated CTP platforms, critical human oversight remains irreplaceable. AI platforms are validated for specific scanner types, contrast protocols, and patient populations. They assume adequate scan quality — map quality verification (AIF confirmation, motion assessment, correct protocol parameters) must be performed by a trained radiographer or radiologist on every study before automated results are acted upon. Automated CTP output should always be accompanied by a flag to the reporting radiologist confirming map quality. The thrombectomy decision must remain a physician responsibility — AI provides the substrate for that decision, not the decision itself.^[16]

Emerging technologies in CTP

Photon-counting CT (PCCT) is beginning to reshape CTP capability. PCCT’s superior energy resolution enables iodine quantification at lower dose — producing perfusion maps with improved spatial resolution and reduced radiation compared to energy-integrating detector (EID) scanners. Early clinical data from Siemens NAEOTOM Alpha and GE Revolution Apex platforms confirm improved AIF quality and reduced image noise in CTP at 80 kVp.^[17] As PCCT platforms become more widespread, existing CTP thresholds (rCBF <30%, Tmax >6s) will require re-validation against PCCT-specific data before routine clinical application.

Dual-energy CT (DECT) perfusion offers virtual monoenergetic image (VMI) post-processing that can enhance AIF conspicuity and reduce beam-hardening artefacts in the skull base — a persistent challenge in standard EID CTP. DECT-derived iodine maps also allow direct quantification of iodine concentration as a surrogate perfusion marker. These capabilities remain research-stage for perfusion purposes and are not yet part of validated clinical trial protocols.

13. Conclusion

CT brain perfusion is the most technically demanding, physiologically rich, and clinically decisive protocol in the neuroradiology suite. Executed precisely, it transforms the stroke imaging triage from an anatomical exercise into a tissue-level diagnosis — identifying exactly how much brain is irreversibly infarcted, how much is salvageable, and whether the patient will benefit from mechanical thrombectomy. Executed poorly, it can actively mislead a team into catastrophic treatment errors, either denying intervention to a salvageable patient or proceeding with thrombectomy in a patient whose core infarct excludes benefit.

The protocol requirements are absolute: 80 kVp for maximum iodine conspicuity, a compact 40 mL bolus at 6.0 mL/s with a 100 mL saline chaser, a dynamic shuttle or wide-detector acquisition launched within 5 seconds of injection start, and meticulous slice positioning to ensure the MCA territory is fully within the coverage slab. Any deviation from these parameters introduces systematic error into the perfusion maps — and systematic error in CTP maps translates directly to patient harm.

The physiological maps — CBV, CBF, MTT, and Tmax — must be interpreted together, not in isolation. Core is defined by rCBF <30% or CBV <2 mL/100g; penumbra by Tmax >6 seconds. The mismatch ratio governs the treatment decision in the extended 6–24-hour window. Automated AI platforms including RAPID, Brainomix, and Viz.ai have been validated for this specific application and deliver maps within minutes — but they require radiologist quality verification on every study and cannot replace clinical judgement or specialist map review.

The pitfall framework across three professional roles — the radiographer’s positioning and injection responsibilities, the radiologist’s AIF quality review and bilateral collateral recognition, and the clinician’s disciplined use of relative rather than absolute map values — forms a multi-layered safety net that every stroke imaging team should adopt as institutional protocol. When all three layers function correctly, CT brain perfusion performs its extraordinary function: giving a patient the possibility of leaving the hospital walking, when without it they might not leave at all.