A deceptively mild stroke can conceal a rapidly worsening threat, according to a new systematic review that identifies brain-tissue blood-flow patterns capable of revealing which patients with a major cerebral artery blockage are most likely to deteriorate. The analysis focuses on people who arrive at hospital with a low score on the National Institutes of Health Stroke Scale, or NIHSS, even though imaging shows a large-vessel occlusion (LVO). These patients may have only slight weakness, subtle speech problems or limited sensory changes, yet the blocked artery can leave large regions of the brain dependent on fragile collateral blood vessels. Liu and colleagues report that perfusion and hemodynamic imaging can distinguish a high-risk “tissue-at-risk” phenotype associated with early neurological deterioration and poor functional recovery. Their findings could intensify a central debate in emergency stroke medicine: whether patients who appear too well for an invasive procedure should undergo mechanical thrombectomy, in which a catheter is used to remove the clot. But the review also delivers a warning. Although blood-flow imaging appears useful for predicting danger, there is not yet reliable evidence that using it to select patients for thrombectomy improves outcomes.

The clinical problem arises because the NIHSS is a measure of observable neurological impairment, not a direct measure of how much brain tissue is threatened. The scale emphasizes findings such as limb weakness, language disturbance, gaze deviation and neglect, but it can underrepresent deficits that are highly disabling for a particular patient or that have not yet emerged. A person with an occlusion in the internal carotid artery or proximal middle cerebral artery may initially compensate through collateral circulation, with alternative vessels temporarily supplying blood around the blockage. That compensation can fail as blood pressure changes, collateral channels become exhausted or the clot propagates. When it does, a previously stable patient may experience early neurological deterioration, sometimes within hours. The review examined this low-NIHSS LVO population because clinicians must balance two dangers: withholding thrombectomy from someone whose brain is about to become irreversibly injured, and exposing a person with a mild, potentially stable stroke to an invasive procedure that carries risks including vessel injury, bleeding and symptomatic intracranial hemorrhage.

To investigate whether advanced imaging can improve that decision, the researchers searched PubMed, Embase, Scopus, Web of Science and the Cochrane Library from their beginnings through June 2026. The search produced 1,698 records; after duplicate studies were removed, 1,372 remained for screening. The evidence base ultimately incorporated studies assessing perfusion or hemodynamic markers, investigations linking imaging to early deterioration or later disability, observational studies of imaging-informed endovascular treatment, previous reviews and meta-analyses, and protocols for randomized trials still under way. The team used random-effects models for exploratory pooled analyses, an approach that allows the estimated effect to vary between studies rather than assuming that all research was measuring precisely the same phenomenon. Risk of bias was examined using domains informed by QUIPS and ROBINS-I, while the certainty of evidence was graded using GRADE. This methodology matters because most available evidence in this narrowly defined patient group is observational: doctors chose treatment, imaging thresholds differed and patient populations were not identical.

The imaging markers highlighted by the review are designed to measure not merely whether blood reaches the brain, but how delayed or inadequate that delivery is. Computed tomography perfusion and perfusion-weighted magnetic resonance imaging track the passage of contrast through cerebral tissue and generate maps of blood flow, blood volume and transit time. One widely used metric is Tmax, the time to the maximum of the residue function, which estimates the delay between contrast entering an unaffected arterial territory and reaching tissue downstream of an obstruction. Larger regions with prolonged Tmax indicate more extensive or severe hypoperfusion. Researchers also examine the total perfusion-lesion volume, relative cerebral blood flow and relative cerebral blood volume, as well as the hypoperfusion intensity ratio, which compares severely delayed tissue with the broader area of delayed perfusion. A further concept is hemodynamic reserve, the brain’s ability to increase blood flow when demand rises. If that reserve is impaired, tissue may look clinically stable at one moment but have little capacity to withstand even a modest reduction in perfusion.

Across the studies, larger Tmax-defined hypoperfusion volumes and larger perfusion lesions were repeatedly associated with early neurological deterioration or an unfavorable functional outcome. Impaired hemodynamic reserve showed a similar relationship, suggesting that the most informative scans may be those that reveal exhausted compensation rather than simply the presence of a blocked vessel. In three studies suitable for exploratory pooling, patients classified as having a high-risk perfusion phenotype had markedly greater odds of early deterioration or deterioration judged to be ischemic in origin. The pooled odds ratio was 6.55, with a 95 percent confidence interval of 1.65 to 25.99. In practical terms, the estimate suggests a several-fold elevation in risk, but the wide interval indicates considerable uncertainty. Statistical heterogeneity was also substantial, with an I² value of approximately 78 percent, meaning that differences between studies accounted for much of the variation in their results. Two studies examining functional outcome produced a pooled odds ratio of 9.24, with a 95 percent confidence interval of 1.99 to 42.99 and I² of about 66 percent.

Those numbers are compelling enough to make perfusion imaging an attractive early-warning system, but they do not prove that imaging-guided intervention is beneficial. A scan can identify tissue that is vulnerable without establishing that removing the clot will rescue it, or that the benefits of thrombectomy outweigh the procedure’s hazards in a patient with limited symptoms. Observational treatment studies are especially vulnerable to confounding: patients sent for thrombectomy may be younger, treated faster, cared for at specialized centers or selected because clinicians already suspected a dangerous course. Conversely, patients managed medically may have anatomy, comorbidities or delays that make them less suitable for intervention. Differences in scanner technology, software, Tmax thresholds, definitions of early deterioration and follow-up timing can further alter the apparent strength of associations. The review therefore separates prognostic evidence from therapeutic evidence. Perfusion patterns may help forecast what happens next, but the analysis found that evidence for using those patterns as a decisive trigger for endovascular therapy remains heterogeneous and observational.

For emergency physicians and neurologists, the findings support a more nuanced approach than treating a low NIHSS score as reassurance. Initial clinical examination, noncontrast computed tomography and CT angiography remain central: the first helps assess established injury and bleeding, while angiography identifies the occluded artery and its location. Perfusion imaging can add a physiological dimension by estimating how much tissue is receiving delayed or insufficient blood flow. A patient with mild symptoms but a large hypoperfusion territory, poor collateral support or exhausted cerebrovascular reactivity may deserve especially close monitoring and rapid discussion with an endovascular team. Yet imaging must be interpreted alongside symptom severity, disability in daily life, clot location, time since symptom onset, intravenous thrombolysis eligibility and the possibility of neurological recovery without intervention. Perfusion maps are not direct photographs of dead and living neurons. They are model-based estimates influenced by contrast timing, motion, cardiac output, arterial input selection and software algorithms. A threshold that labels tissue as endangered in one system may not be interchangeable with a threshold from another.

The decisive evidence will have to come from randomized trials that compare medical management with immediate or imaging-selected thrombectomy in patients with low NIHSS scores and LVO. The review identifies ongoing efforts, including the MILD-MT and MOSTE trial programs, as important attempts to resolve the uncertainty. Such studies can determine whether a high-risk perfusion signature is a treatment effect modifier—an imaging feature that identifies people who benefit especially strongly from thrombectomy—or simply a marker of poor prognosis. They must also measure outcomes beyond survival and a conventional modified Rankin Scale score, because small neurological changes can have major consequences for communication, dexterity, work and independence. Until those trials report, the message is both urgent and restrained: a mild examination should not obscure a dangerous blockage, and advanced imaging may reveal that hidden risk, but no scan alone currently justifies a universal policy of intervention. The review, published in the Journal of Neurology, turns a clinical blind spot into a testable question—whether restoring blood flow before subtle deficits become catastrophic can change the fate of patients who initially look almost well.

Subject of Research: Perfusion imaging-guided risk stratification and endovascular decision-making in adults with low-NIHSS large-vessel occlusion stroke

Article Title: Perfusion imaging-guided risk stratification and endovascular decision-making in low-NIHSS large-vessel occlusion stroke: a systematic review and exploratory meta-analysis

Article References: Liu, C., Hu, T., Wu, X. et al. Journal of Neurology 273, 551 (2026). Original research article

Image Credits: AI Generated

DOI: 10.1007/s00415-026-14075-7

Keywords: low-NIHSS stroke, large-vessel occlusion, perfusion imaging, early neurological deterioration, hypoperfusion, hemodynamic reserve, mechanical thrombectomy, endovascular therapy

Tags: collateral blood vessels in strokeearly neurological deterioration in strokeischemic stroke blood flow patternslarge-vessel occlusion strokelow NIHSS stroke patients with large-vessel occlusionmechanical thrombectomy decision-makingperfusion imaging in stroke managementpredicting stroke outcomes with perfusion imagingrisks of invasive procedures in mild strokerole of hemodynamic imaging in stroke treatmentstroke imaging and treatment selectiontissue-at-risk identification in stroke