CT-perfusion parameters and reperfusion therapy in acute ischemic stroke: 24-hour NIHSS dynamics in a retrospective cohort

Full Text

Neurological improvement in patients with ischemic stroke depends on multiple clinical, laboratory, and neuroimaging factors, including baseline neurological deficit, infarct volume, success of reperfusion therapy, and the presence of collateral circulation [1]. In the era of an extended time window for reperfusion therapy, understanding the factors influencing short- and long-term neurological recovery becomes critically important. Ischemic stroke due to large vessel occlusion results in a mosaic of irreversibly damaged ischemic core and potentially salvageable brain tissue, the penumbra, and the ratio between these components determines the effectiveness of reperfusion therapy regardless of treatment timing [2]. The state of this dynamic balance between the zone of irreversible damage and the threatened brain tissue provides the biological rationale for patient selection for reperfusion therapy using perfusion neuroimaging.

Computed tomography perfusion (CTP) is the tool for in vivo assessment of this ratio, allowing quantitative measurement of ischemic core volume, critical hypoperfusion volume, and penumbra volume. CT perfusion parameters have an established prognostic value for selecting patients for reperfusion therapy in the extended time window in accordance with the DEFUSE 3, DAWN, and EXTEND trials, in which inclusion was based on the presence of a mismatch between penumbra or clinical deficit and infarct volume according to perfusion imaging. These trials demonstrated that in patients with preserved penumbra and relatively small infarct core, endovascular treatment and/or intravenous thrombolysis remain effective even when initiated 6–24 hours after symptom onset. The use of standardized thresholds (reduction of regional cerebral blood flow to less than 30% of the contralateral value for the infarct core and Tmax delay greater than 6 seconds for the critically hypoperfused area), validated in clinical trials using automated post-processing software, ensures measurement reproducibility. At the same time, the prognostic value of different perfusion parameters for short-term clinical improvement as measured by the National Institutes of Health Stroke Scale (NIHSS) during the first 24 hours remains insufficiently investigated and debated. [3–7].

The effectiveness of thrombolytic therapy and mechanical thrombectomy in preventing infarct growth and restoring neurological function is now well established [8]. Successful recanalization according to the modified Treatment in Cerebral Infarction (mTICI) scale, which reflects the degree of reperfusion, is considered one of the key determinants of clinical outcome. However, despite technically successful recanalization, 30–52% of patients exhibit the phenomenon of “futile recanalization,” in which restoration of blood flow does not translate into clinical improvement or is associated with an unfavorable 90-day functional outcome [9]. According to a 2023 meta-analysis including data from more than 8000 patients, the leading predictors of futile recanalization were severe baseline neurological deficit, large infarct core volume, high admission blood glucose, and prolonged time from symptom onset to recanalization [10]. This underscores the need to identify additional predictors of clinical response to reperfusion therapy beyond the technical success of recanalization alone.

One of the mechanisms explaining the discrepancy between successful recanalization and clinical effect is the no-reflow phenomenon (persistent tissue-level hypoperfusion despite restoration of arterial patency), which is detected on CT perfusion in up to 80% of patients with successful recanalization after mechanical thrombectomy [11]. In addition, reperfusion injury may occur after blood flow restoration and is accompanied by increased brain edema and the release of reactive oxygen species [12]. Taken together, these phenomena indicate that neither baseline neuroimaging parameters nor the degree of recanalization alone can fully predict short-term NIHSS dynamics, and that their integrative assessment within regression models is of substantial scientific interest.

It should be noted that the impact of CT perfusion parameters on NIHSS dynamics during the first 24 hours, as opposed to the conventional 90-day functional outcomes, has been addressed in relatively few studies [13]. At the same time, 24-hour NIHSS is recognized as a clinically meaningful surrogate marker associated with long-term functional outcome and can be used as an early indicator of treatment effectiveness in real-world clinical practice [6, 14, 15].

Study objective—to assess the association and prognostic significance of CT perfusion parameters, as well as the use of intravenous thrombolysis and mechanical thrombectomy (considering the degree of recanalization on the TICI scale), in relation to NIHSS dynamics during the first 24 hours after admission in patients with acute ischemic stroke due to large vessel occlusion.

METHODS

Study design

This study of patients with ischemic stroke admitted to a regional stroke center is a retrospective, single-center cohort study without matched control groups.

Study settings

The study was conducted at the City Clinical Hospital named after S.S. Yudin, Moscow Healthcare Department (Moscow, Russia). Patient inclusion was carried out from April 30 to December 30, 2024, while data processing, analysis, and manuscript preparation were performed from January to March 2026.

Inclusion and exclusion criteria

Inclusion criteria. A clinical diagnosis of ischemic stroke with symptom onset no more than 24 hours before admission to the emergency department; age ≥18 years; completion of the required brain imaging protocol, including non-contrast CT, CT angiography of the brachiocephalic arteries, and CT perfusion; presence of an occlusion in one of the following arteries: internal carotid artery, middle cerebral artery (segments M1–M3), anterior cerebral artery (segments A1–A2), or posterior cerebral artery (segment P1).

Exclusion criteria. Presence of intracranial hemorrhage or any other major brain pathology unrelated to ischemic stroke; inadequate image quality of at least one required study that prevents reliable analysis; lack of coverage of the vascular territory of the occluded artery within the CT perfusion scan range; absence of occlusion in one of the following arteries: internal carotid, middle cerebral (segments M1–M3), anterior cerebral (segments A1–A2), or posterior cerebral (segment P1); missing data on the dynamics of neurological status (NIHSS) during the first 24 hours after admission.

Description of eligibility criteria. The inclusion criteria are consistent with standard requirements for studies of acute ischemic stroke due to large vessel occlusion. The requirement that symptom onset be no more than 24 hours before admission is in line with current clinical guidelines and the criteria used in the DAWN and DEFUSE 3 trials [16–18].

The age criterion (≥18 years) is conventional for studies in adult populations and reflects the profile of a medical facility providing emergency care to adult patients. The mandatory neuroimaging protocol is determined by clinical necessity: non-contrast CT is used to exclude hemorrhagic stroke; CT angiography is used to verify large vessel occlusion; CT perfusion provides information on infarct core and penumbra. This three-step protocol is consistent with the approach used in similar retrospective cohort studies [19].

The list of target arteries (internal carotid; middle cerebral, segments M1–M3; anterior cerebral, segments A1–A2; posterior cerebral, segment P1) encompasses proximal occlusions in the anterior and posterior circulation that are amenable to endovascular treatment and allow adequate calculation of core and penumbra parameters (CT perfusion has limited applicability in the posterior fossa). This list matches that used in similar retrospective studies of predictors of NIHSS improvement [20].

The exclusion criteria were formulated to ensure homogeneity of the sample with respect to pathology type and data quality. Excluding patients with intracranial hemorrhage or other brain disorders that contraindicate thrombolytic therapy and/or mechanical thrombectomy was intended to eliminate the influence of competing causes of neurological deficit. Exclusion of studies with inadequate image quality and insufficient CT perfusion coverage was driven by the technical requirements for perfusion map calculations. Excluding patients without confirmed large vessel occlusion ensured clinical homogeneity of the sample. Exclusion of patients without 24-hour NIHSS data was dictated by the definition of the dependent variable in this study.

Non-inclusion criteria were not explicitly specified, as they were effectively predefined before data collection and applied during the initial screening stage: some of the above exclusion criteria (presence of hemorrhage, absence of occlusion, inadequate image quality) could by their nature be considered criteria for non-inclusion.

A formal, separate list of non-inclusion criteria, distinct from the exclusion criteria, was not planned. The exclusion criteria were partially defined before data collection (as requirements for the imaging protocol and presence of occlusion) and partially refined during the retrospective analysis (inadequate CT perfusion quality, missing 24-hour NIHSS data).

Allocation of participants to groups. Given the study design, all patients who met the eligibility criteria formed a single cohort for regression analysis. Assignment to a particular therapeutic subgroup (intravenous thrombolysis, mechanical thrombectomy, combined therapy, or conservative treatment) was treated as an independent predictor variable in the regression model rather than as a criterion for forming separate comparison groups.

Study intervention

A total of 52 patients were included in the study.

Neurological status was assessed using the NIHSS at admission and again at 24 hours. The NIHSS assessment was performed by the treating neurologists as part of the standard clinical protocol of the institution. Data on NIHSS dynamics were collected retrospectively from medical records. Thus, the primary outcome was measured using secondary data sources (medical documentation).

Information on the fact and characteristics of reperfusion therapy (administration of intravenous thrombolysis, attempt of mechanical thrombectomy, degree of recanalization on the TICI scale) was also extracted from medical records. The TICI (Thrombolysis in Cerebral Infarction) scale is a standardized angiographic grading system for the degree of reperfusion, where 0 indicates no perfusion; 1, minimal perfusion; 2a, partial reperfusion of less than 50% of the affected territory; 2b, partial reperfusion of 50% or more; and 3, complete reperfusion. The TICI grade was assigned by interventional neuroradiologists during digital subtraction angiography and documented in the medical records.

No specific measures to minimize measurement bias (such as assessor blinding or randomization of the timing/order of assessments) were implemented due to the retrospective design of the study.

Study outcomes

The primary outcome of the study was the 24-hour change in NIHSS score (ΔNIHSS) from admission, defined as the difference between NIHSS at 24 hours and baseline NIHSS (ΔNIHSS=NIHSS₂₄−NIHSS₀). A negative ΔNIHSS value indicates an improvement in neurological status. This quantitative measure is a standardized surrogate marker of short-term effectiveness of reperfusion therapy and is widely used in the literature as a dependent variable when evaluating predictors of early neurological improvement. According to published data, short-term NIHSS improvement, including definitions such as a decrease of ≥4 points or achieving NIHSS ≤1 at 24 hours, is a significant predictor of long-term functional outcomes on the modified Rankin Scale (mRS) at 90 days [15, 20–22].

The NIHSS is a validated tool for assessing the severity of acute ischemic stroke with good reproducibility (inter-rater reliability κ=0.6–0.8) when performed by certified raters. In the present study, ΔNIHSS was treated as a continuous variable, which allows full use of the range of changes and increases statistical power compared with dichotomization [15, 23, 24].

As secondary outcomes, we evaluated: (1) standardized (normalized) ridge regression coefficients for each predictor to quantify the relative contribution of each variable (perfusion parameters, characteristics of reperfusion therapy, baseline NIHSS) to the explained variance of the dependent variable; (2) the coefficient of determination (R²) of the regression model to assess how much of the variability in ΔNIHSS was explained by the model; and (3) the Pearson correlation matrix between predictors to assess the degree of multicollinearity, which influenced the choice of regression method.

Additional outcomes, such as 90-day mRS or the rate of symptomatic intracranial hemorrhage, were not analyzed because of the retrospective design and limited availability of long-term follow-up data.

Methods for measuring outcomes. CT perfusion data were processed using the ArchiMed PRO Chronos software complex (Russia), which has demonstrated good agreement with one of the most widely used CT perfusion post-processing platforms, Vitrea (Canon Medical Systems, USA, Japan) [25], and standard threshold values were applied (rCBF <30% and Tmax >6 s for infarct core and critically hypoperfused tissue, respectively; penumbra volume was calculated as the difference between these volumes).

All patients underwent non-contrast CT, CT angiography of extra- and intracranial arteries, and CT perfusion of the brain on a multislice CT scanner Aquilion Prime SP (Canon Medical Systems, Japan) according to the institution’s standard protocol. Raw dynamic CT perfusion datasets were exported in DICOM format and processed in the ArchiMed PRO Chronos software (Russia) using built-in algorithms for automatic calculation of perfusion parameters and lesion volumes.

In ArchiMed PRO Chronos, the ischemic core was defined as tissue with a relative cerebral blood flow (rCBF) less than 30% of the value in the contralateral hemisphere. This threshold (rCBF <30%) is standard and is implemented in the RAPID software (iSchemaView, USA), which has been validated in randomized trials (EXTEND-IA, SWIFT-PRIME, DEFUSE 3). Critically hypoperfused tissue (tissue at risk) was defined as an increase in Tmax to >6 seconds (time to the maximum of the residue function), which is also a standard threshold that has been validated in the DAWN and DEFUSE 3 trials. Penumbra volume was calculated as the difference between the volume of critical hypoperfusion (Tmax >6 s) and infarct core volume (rCBF <30%), corresponding to the “mismatch volume” metric in the respective clinical trials [4].

Sensitivity analysis

A formal sensitivity analysis was not performed in the present study. This decision was driven by the retrospective design and the relatively small sample size (52 patients, of whom 34 were included in the primary regression analysis). Conducting additional multivariable subgroup analyses with this sample size was considered inappropriate because of the high risk of model overfitting.

Statistical analysis

Planned sample size. A formal a priori sample size calculation was not carried out at the planning stage. The retrospective cohort included all consecutively hospitalized patients during the study period who met the inclusion criteria and did not meet the exclusion criteria, resulting in a final sample of 52 observations. This approach was chosen to maximize the use of available clinical data while maintaining sample homogeneity with respect to key characteristics (stroke type, presence of large vessel occlusion, standardized imaging protocol). No stopping rules were defined, since the analysis was performed after inclusion of all eligible cases.

Statistical methods. Statistical analysis was performed using Python with the libraries numpy, pandas, statsmodels, scipy, seaborn, and matplotlib. The significance level for all tests was set at p <0.05, and all p-values were reported as two-sided. Given the identified multicollinearity between predictors (r >0.8 for several pairs of variables, particularly between critical hypoperfusion volume and penumbra volume, as well as between attempted mechanical thrombectomy and final TICI grade), ridge regression (Tikhonov regularization) was used to assess the effects of predictors on ΔNIHSS, as this method stabilizes coefficient estimates in the presence of multicollinearity. The regularization parameter (λ) was selected using cross-validation. Model performance was evaluated using the coefficient of determination R². For graphical comparison of predictor contributions, regression coefficients were normalized (standardized). Paired quantitative variables and the strength of linear association between them were analyzed using parametric or non-parametric tests for paired samples, depending on data distribution, and the degree of linear association was quantified with Pearson’s correlation coefficient (r). Missing data were not imputed: examinations with inadequate image quality or incomplete perfusion coverage were excluded at the sample-formation stage.

RESULTS

Study sample selection

During the study period, patients with a clinical diagnosis of acute ischemic stroke who were admitted within 24 hours of symptom onset and underwent a comprehensive CT protocol (non-contrast CT, CT angiography, and CT perfusion) were consecutively identified from the medical records of the City Clinical Hospital named after S.S. Yudin. In total, 52 patients met the eligibility criteria and were included in the study, of whom 34 (65.4%) had complete data required for the primary regression analysis: high-quality segmentation of perfusion data and NIHSS scores both at admission and at 24 hours. The remaining 18 patients (34.6%) were excluded from the regression analysis due to missing at least one of these components.

Characteristics of the study sample

Because this was a retrospective study and did not involve prospective definition of a target or source population for comparison with non-included patients, formal comparison between included and non-included patients was not feasible.

Main results

In all patients, in addition to perfusion parameters, neurological status was assessed using NIHSS at admission and at 24 hours, and the presence or absence of reperfusion therapy (intravenous thrombolysis and/or mechanical thrombectomy with assessment of recanalization degree on the TICI scale) was recorded. Of the 52 patients, only 34 had complete clinical and imaging data (baseline and 24-hour NIHSS, and adequate perfusion segmentation) and were therefore included in the regression analysis. When assessing correlations between the selected parameters, some variables showed a strong association (r >0.8; Fig. 1), specifically: critical hypoperfusion volume and penumbra volume (the latter is derived using the former), and performance of mechanical thrombectomy and final TICI reperfusion grade.

 

Fig. 1. Pearson correlation between the studied parameters. Thrombolysis, intravenous thrombolytic therapy; Thrombectomy—mechanical thrombectomy; Thrombectomy TICI, TICI scale for mechanical thrombectomy; NIHSS, National Institutes of Health Stroke Scale.

 

Accordingly, ridge regression (Tikhonov regularization) was chosen to evaluate the effect of all selected predictors on change in NIHSS (calculated as the difference between the 24-hour and baseline scores), in order to mitigate the impact of multicollinearity. The resulting model had an R² of 0.44, indicating that 44% of the variance in the dependent variable (24-hour ΔNIHSS) could be explained by the predictors included in the model. Detailed assessment of individual predictors showed the following:

1. the strongest contribution was from the TICI grade reflecting the degree of arterial recanalization after mechanical thrombectomy: higher TICI grades were associated with a greater decrease in NIHSS at 24 hours (statistically significant predictor);
2. an attempt at mechanical thrombectomy, regardless of its technical success, was associated with a smaller improvement (i.e., less reduction) in NIHSS at 24 hours (statistically significant predictor);
3. higher baseline NIHSS scores were associated with a more pronounced reduction in NIHSS at 24 hours (statistically significant predictor);
4. infarct core volume on baseline CT perfusion was associated with less NIHSS improvement at 24 hours (negative effect), but this predictor did not reach statistical significance;
5. baseline penumbra volume on CT perfusion was associated with greater NIHSS improvement at 24 hours (positive effect), but did not reach statistical significance;
6. an attempt at intravenous thrombolytic therapy, irrespective of its success, was associated with greater NIHSS improvement at 24 hours (positive effect), but this predictor was not statistically significant;
7. baseline critical hypoperfusion volume on CT perfusion was likewise associated with greater NIHSS improvement at 24 hours (positive effect), but did not reach statistical significance.

For visual comparison of the influence of different predictors on NIHSS change, regression coefficients were additionally normalized and presented on a separate plot (Fig. 2). Taken together, these findings highlight the predominant role of recanalization degree on the TICI scale and baseline NIHSS severity, as well as the presence of directional but statistically non-significant associations between perfusion parameters, intravenous thrombolysis, and short-term neurological improvement.

 

Fig. 2. Contribution of different parameters in the ridge regression model for predicting 24-hour change in NIHSS score. Thrombolysis, intravenous thrombolytic therapy; Thrombectomy, mechanical thrombectomy; Thrombectomy TICI, TICI scale for mechanical thrombectomy; NIHSS, National Institutes of Health Stroke Scale.

 

DISCUSSION

Summary of the main study results

This retrospective study demonstrated that among the clinical and neuroimaging predictors evaluated, three parameters had a significant and independent impact on 24-hour NIHSS change: baseline neurological deficit at admission, degree of arterial recanalization on the TICI scale after mechanical thrombectomy, and the mere fact of attempting mechanical thrombectomy. Successful recanalization was associated with a significantly greater reduction in NIHSS, whereas an attempt at mechanical thrombectomy, irrespective of its angiographic outcome, had a negative effect on short-term neurological status, which may reflect both procedure-related risks and a more severe baseline profile of patients selected for the intervention. Perfusion parameters (core, penumbra, and critical hypoperfusion volumes) and the use of intravenous thrombolysis did not reach statistical significance, although the direction of the observed trends was consistent with previously published data. The ridge regression model explained 44% of the variance in the primary outcome [3, 20], indicating that nearly half of the variability in 24-hour NIHSS dynamics in our cohort could be accounted for by the included clinical, perfusion, and procedural factors.

Interpretation of the results

The results of this study showed that among all evaluated parameters, only three factors had a statistically significant impact on change in neurological status, measured as 24-hour NIHSS dynamics: baseline NIHSS score at admission, performance of mechanical thrombectomy, and the degree of arterial recanalization on the TICI scale after mechanical thrombectomy. Notably, the latter two variables had opposite effects in the regression model, indicating a complex interplay between the mere fact of attempting endovascular treatment and its technical success. Successful recanalization (as reflected by the TICI grade) was associated with a statistically significant reduction in NIHSS at 24 hours, whereas an attempt at mechanical thrombectomy, regardless of its outcome, negatively affected short-term prognosis, which may be related both to procedural risks and to the more severe baseline profile of patients selected for the intervention [26]. This paradoxical finding underscores the need for careful patient selection for mechanical thrombectomy and raises important questions about balancing procedural risks against potential benefits.

Importantly, none of the perfusion parameters (ischemic core volume, critical hypoperfusion volume, or penumbra volume) demonstrated an independently statistically significant effect on 24-hour NIHSS change, although trends were observed in the expected directions. Specifically, ischemic core volume showed a negative association with NIHSS improvement (negative regression coefficient, p >0.05), whereas penumbra volume showed a positive association (positive regression coefficient, p >0.05), which is consistent with literature on the prognostic value of these parameters [27]. Similarly, intravenous thrombolytic therapy demonstrated a positive trend (positive regression coefficient, p >0.05), although this effect did not reach statistical significance. The most likely explanation for the lack of statistical significance is the relatively small sample size, which limits the statistical power to detect moderate effects of individual predictors.

The trends observed in the regression model are in line with extensive literature supporting the effectiveness of both forms of reperfusion therapy. It is well established that thrombolytic therapy, particularly when administered within the therapeutic window (up to 4.5 hours from symptom onset), substantially improves neurological outcomes in patients with acute ischemic stroke [28]. Systematic reviews and meta-analyses have shown that intravenous thrombolysis with recombinant tissue plasminogen activator (rt-PA) is associated with a 2- to 3-fold increase in the likelihood of a favorable functional outcome at 90 days [29–31]. Ischemic core volume and, especially, penumbra volume are regarded as critical parameters for patient selection in the extended time window, as demonstrated by key randomized clinical trials [32]. For example, the DEFUSE 3 criteria required a ratio of critically hypoperfused tissue to core ≥1.8, a penumbra volume >15 mL, and a core volume <70 mL, thereby selecting patients with the highest potential benefit from mechanical thrombectomy in the extended window [1]. In DEFUSE 3, patients meeting these criteria exhibited significantly improved functional outcomes at 90 days [32]. Nevertheless, in our study, the effect of mechanical thrombectomy (particularly successful recanalization) outweighed the impact of individual perfusion parameters, which may be explained by several factors. First, restoration of blood flow in the occluded artery remains the primary mechanism for salvaging the penumbra and improving neurological function, whereas baseline perfusion measurements may be insufficiently sensitive to the dynamic processes occurring in the first hours after recanalization [27]. Second, the dynamic nature of ischemic stroke implies that core and penumbra volumes measured at baseline do not fully reflect the tissue state at the time of the intervention, especially in the setting of delayed hospital arrival or fluctuations in blood pressure [9].

In the context of the extended time window, particular attention has been paid to the phenomenon of “super-selection” of patients, whereby perfusion imaging not only identifies ischemic penumbra but also enables highly accurate prediction of clinical response to large vessel recanalization [2]. Patients selected according to DAWN and DEFUSE 3 criteria, despite late presentation (6–24 hours after symptom onset), achieved functional outcomes comparable to or even better than those of patients treated within the standard window (up to 6 hours), highlighting the critical importance of preserving a substantial penumbra volume [9]. In our study, the directions of regression coefficients were consistent with these data, showing a detrimental effect of larger core volumes and a beneficial effect of larger penumbra volumes; however, statistical power was insufficient to achieve formal significance.

The lack of statistical significance for the effect of thrombolytic therapy in our cohort also warrants interpretation in the context of existing evidence. It is well established that thrombolytic therapy, especially when combined with mechanical thrombectomy as part of a stepwise reperfusion strategy, leads to significant improvements in outcomes in both early and late therapeutic windows [33–37]. A meta-analysis of several randomized controlled trials demonstrated that thrombolytic therapy combined with mechanical thrombectomy is associated with higher odds (odds ratio 1.4–1.6) of a good functional outcome (mRS 0–2) at 90 days compared with mechanical thrombectomy alone [37–39]. The positive regression coefficient for thrombolytic therapy in our model is consistent with these findings and suggests that, with a larger sample size, this effect might have reached statistical significance.

An important clinical observation is that baseline NIHSS score at admission emerged as a strong independent predictor of 24-hour NIHSS improvement, with patients presenting with higher initial scores exhibiting more pronounced reductions. This is consistent with previous reports indicating that patients with more severe initial deficits, who typically have more extensive brain injury, also have greater potential for improvement after successful recanalization [15, 23, 40, 41]. In addition, the well-known ceiling effect in patients with initially low NIHSS scores limits the extent of possible improvement because the minimum attainable score is 0.

The paradoxical negative effect of simply attempting mechanical thrombectomy (regardless of its angiographic success) on short-term NIHSS improvement requires critical consideration. Although successful recanalization clearly improved outcomes, the procedure itself was associated with worse short-term prognosis, which may be due to several mechanisms. First, patients selected for mechanical thrombectomy likely had more severe disease (occlusions of larger arteries, more pronounced neurological deficits) than those treated with intravenous thrombolysis alone or managed conservatively, which could bias the model toward a negative contribution of the “mechanical thrombectomy attempt” variable [42–45]. Second, procedure-related complications such as arterial dissection, vessel perforation, distal embolization of thrombus fragments, or contrast-induced encephalopathy may worsen neurological status, although in some cases long-term outcomes may still improve due to successful reperfusion [46–50]. Third, even after technically successful recanalization, reperfusion injury may occur, characterized by the release of reactive oxygen species, activation of inflammatory cascades, and additional brain edema, which may adversely affect NIHSS scores during the first 24 hours, even though some tissue may subsequently recover [51–54].

Study limitations

The sample size was 52 patients, of whom 34 were included in the primary regression analysis. At this sample size, statistical power to detect moderate effects of individual predictors, particularly perfusion parameters and thrombolytic therapy, was insufficient. The single-center design and inclusion of consecutively hospitalized patients from one hospital limit the generalizability of the findings to the broader population of patients with acute ischemic stroke [55–57].

Allocation of patients to different reperfusion strategies (intravenous thrombolysis, mechanical thrombectomy, combined therapy, conservative treatment) was not randomized and followed routine clinical practice. Patients selected for mechanical thrombectomy were likely to have had more severe baseline neurological deficits and more proximal occlusions than those treated with thrombolysis alone or managed conservatively, introducing treatment-selection bias that cannot be fully eliminated by regression modeling at this sample size. Use of ridge regression allowed partial control of multicollinearity among predictors but could not remove confounding by unmeasured variables such as collateral circulation, stroke etiology according to the TOAST classification, or time from onset to recanalization [58, 59].

The primary endpoint (24-hour ΔNIHSS) is a surrogate marker that does not fully capture functional outcome; the correlation between short-term NIHSS improvement and long-term functional status (90-day mRS) is only moderate. NIHSS scores were abstracted from routine medical records, which introduces a risk of documentation bias, and inter-rater standardization across different clinicians was not controlled [22, 24, 60].

CONCLUSION

The findings of this study reinforce the central role of successful recanalization in determining short-term clinical prognosis in acute ischemic stroke. Although perfusion parameters and intravenous thrombolytic therapy did not show independent statistical significance in predicting early improvement on the National Institutes of Health Stroke Scale (NIHSS), the observed trends are consistent with extensive literature supporting the effectiveness of reperfusion therapy and the importance of quantifying penumbra volume at admission. These results highlight the need to integrate clinical data (NIHSS), neuroimaging parameters, and information on the technical success of large-vessel recanalization to optimize outcome prediction in patients with acute ischemic stroke.

ADDITIONAL INFORMATION

Author contributions: I.L. Gubskiy, conceptualization, data analysis, manuscript revision and editing, study supervision; M.M. Beregov, data analysis, data curation, manuscript revision and editing; N.E. Staroverov, I.A. Larionov, data analysis, data curation; K.Yu. Kazachkov, A.P. Stepanchenko, V.A. Nechaev, investigation, validation; N.A. Marskaya, data curation, validation; N.A. Shamalov, study supervision, validation. Thereby, all authors provided approval of the version to be published and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Ethics approval: The study was approved by the Local Ethics Committee of the Pirogov Russian National Research Medical University (Pirogov University), Ministry of Health of the Russian Federation (Protocol No. 256 dated December 15, 2025).

Funding source: The work was partially carried out within the framework of the research project “TREATMENT EFFECT 2025–2027,” registration No. 125052706451-2.

Disclosure of interests: The authors declare that they have no competing interests.

Statement of originality: The authors declare no conflicts of interest related to the publication of this article.

Data availability statement: The authors provide limited access to the data (on request). Information containing personal information of patients is not provided.

Generative AI: Generative AI tools were used in the preparation of the manuscript. For searching and the initial analysis of the scientific literature on automated CT perfusion analysis, the following tools were used: Perplexity AI (web service https://www.perplexity.ai, developed by Perplexity AI, Inc., San Francisco, USA), used in February–March 2026; Scopus AI, a generative AI assistant integrated into the Scopus database (developed by Elsevier, Netherlands; accessed via https://www.scopus.com), also used in February–March 2026. All texts, references, and interpretations suggested by the AI tools during the literature search were critically reviewed and edited by the authors, who bear full responsibility for the final content of the manuscript.

About the authors

Ilya L. Gubskiy

The Russian National Research Medical University named after N.I. Pirogov

Author for correspondence.
Email: gubskiy.ilya@gmail.com
ORCID iD: 0000-0003-1726-6801
SPIN-code: 9181-3091

MD, PhD

Russian Federation, Moscow

Mikhail M. Beregov

Federal Center of Brain Research and Neurotechnologies

Email: mik.beregov@gmail.com
ORCID iD: 0000-0003-1899-8131
SPIN-code: 2559-0307
Russian Federation, Moscow

Nikolai E. Staroverov

Saint Petersburg Electrotechnical University LETI

Email: nik0205st@mail.ru
ORCID iD: 0000-0002-4404-5222
SPIN-code: 3147-2108

PhD

Russian Federation, Saint Petersburg

Ivan A. Larionov

Saint Petersburg Electrotechnical University LETI

Email: ivan.al.larionov@gmail.com
ORCID iD: 0000-0001-9620-9471
SPIN-code: 8210-9840

PhD

Russian Federation, Saint Petersburg

Kirill Yu. Kazachkov

City Clinical Hospital named after S.S. Yudin

Email: Kir-82@mail.ru
ORCID iD: 0009-0009-0142-1177
SPIN-code: 1257-8900
Russian Federation, Moscow

Andrey P. Stepanchenko

City Clinical Hospital named after S.S. Yudin

Email: aps65@mail.ru
ORCID iD: 0000-0001-5655-2929
SPIN-code: 7369-4096

MD, PhD

Russian Federation, Moscow

Valentin A. Nechaev

City Clinical Hospital named after S.S. Yudin

Email: dfkz2005@gmail.com
ORCID iD: 0000-0002-6716-5593
SPIN-code: 2527-0130

MD, PhD

Russian Federation, Moscow

Nataliya A. Marskaya

Federal Center of Brain Research and Neurotechnologies

Email: marskayana@gmail.com
ORCID iD: 0000-0002-0789-4823
SPIN-code: 5578-2649
Russian Federation, Moscow

Nikolay A. Shamalov

Federal Center of Brain Research and Neurotechnologies

Email: shamalovn@gmail.com
ORCID iD: 0000-0001-6250-0762
SPIN-code: 2865-9817

MD, PhD

Russian Federation, Moscow

References

Supplementary files