Identification of Predictors for Hemorrhagic Transformation in Patients with Acute Ischemic Stroke After Endovascular Therapy Using the Decision Tree Model

Prediction of Symptomatic Intracranial Hemorrhage Before Mechanical Thrombectomy Using Machine Learning in Patients with Anterior Circulation Large Vessel Occlusion

BACKGROUND

Symptomatic intracranial hemorrhage (sICH) after mechanical thrombectomy (MT) is associated with worse outcomes. We sought to develop and internally validate a machine learning (ML) model to predict sICH prior to MT in patients with anterior circulation large vessel occlusion.

METHODS

Consecutive adults who underwent MT for internal carotid artery/M1/M2 occlusions at a single institution were reviewed. The data was split into 80% training and 20% hold-out test sets. 9 ML models were screened. The top performing ML model was compared to logistic regression and previously described clinical prediction models. SHapley Additive exPlanations were used to identify the most predictive features in the ML model.

RESULTS

A total of 497 patients met inclusion criteria. The top performing ML model was extreme gradient boosting. The area under the receiver operating characteristics curve for the ML model on the test set was 0.79 (95% confidence interval [CI] 0.67-0.89), which was significantly higher (P < 0.001) than the logistic regression model (0.54 [95% CI 0.33-0.76]). The ML model also performed significantly better than the TAG = TICI-ASPECTS-glucose score (0.69 [95% CI 0.55-0.85], P < 0.001), Systolic blood pressure-Time-Blood glucose-ASPECTS score (0.45 [95% CI 0.30-0.60], P < 0.001), and ChatGPT 4.0 (0.60 [95% CI 0.48-0.68], P < 0.001). Based on SHapley Additive exPlanations values the most predictive features of sICH in the ML model were lower Alberta Stroke Program Early CT score, lower collateral score, and higher presenting National Institutes of Health Stroke Scale.

CONCLUSIONS

An ML model accurately predicted sICH prior to MT. It performed better than a standard statistical model and previously described clinical prediction models.

Integrating Neutrophil-To-Albumin Ratio and Triglycerides: A Novel Indicator for Predicting Spontaneous Hemorrhagic Transformation in Acute Ischemic Stroke Patients

BACKGROUND

Hemorrhagic transformation (HT) is a tragic complication of acute ischemic stroke (AIS), with spontaneous HT (sHT) occurring even without reperfusion therapies. Despite evidence suggesting that several inflammation biomarkers are closely related to HT, its utility in sHT risk stratification remains unclear. This study aimed to identify and integrate effective inflammatory biomarkers associated with sHT and to develop a novel nomogram model for the early detection of sHT.

METHODS

We conducted a retrospective observational cohort study of AIS patients receiving conventional medical treatment solely from March 2022 to March 2023, using a prospectively maintained database. All patients underwent CT follow-up within 7 days after admission, with sHT occurrence within this period as the outcome. Data on demographics, clinical information, laboratory results, and imaging were collected. The cohort was divided into training and validation sets (7:3). Least absolute shrinkage and selection operator (LASSO) regression selected inflammatory biomarkers for a novel index. Univariable and multivariable logistic regressions were conducted to identify independent sHT risk factors. Receiver operating characteristic (ROC) analysis determined optimal cut-off values for continuous factors. A nomogram was developed and validated internally and externally. Predictive accuracy was assessed using the area under the ROC curve (AUC) and calibration plots. Decision curve analysis (DCA) evaluated clinical usefulness.

RESULTS

Of 803 AIS patients, 325 were included in the final analysis. sHT was found in 9.5% (31 patients). Training (n = 228) and validation (n = 97) cohorts showed no significant demographic or clinical differences. LASSO regression integrated neutrophil-to-albumin ratio (NAR) and triglycerides (TGs) into a novel index-NATG. Independent sHT risk factors included baseline National Institute of Health Stroke Scale (NIHSS) (OR = 1.09, 95% CI (1.02, 1.16), p = 0.0095), NATG (OR = 1534.87, 95% CI (5.02, 469638.44), p = 0.0120), D-dimer (DD) (OR = 1.12, 95% CI (1.01, 1.25), p = 0.0249), and total cholesterol (TC) (OR = 1.01, 95% CI (1.00, 1.01), p = 0.0280), with their respective optimal cut-off values being 13, 0.059, 0.86, and 3.6. These factors were used to develop the nomogram in the training cohort, which achieved an AUC of 0.804 (95% CI, 0.643-0.918) in the training cohort and 0.713 (95% CI, 0.499-0.868) in the validation cohort, demonstrating consistent calibration. DCA confirmed the nomogram's clinical applicability in both cohorts.

CONCLUSIONS

A novel indicator combining NAR and TG is positively associated with sHT in AIS patients. The constructed nomogram, integrating this novel indicator with other risk factors, provides a valuable tool for identifying sHT risk, aiding in clinical decision-making.

A weakly supervised deep learning model integrating noncontrasted computed tomography images and clinical factors facilitates haemorrhagic transformation prediction after intravenous thrombolysis in acute ischaemic stroke patients

BACKGROUND

Haemorrhage transformation (HT) is a serious complication of intravenous thrombolysis (IVT) in acute ischaemic stroke (AIS). Accurate and timely prediction of the risk of HT before IVT may change the treatment decision and improve clinical prognosis. We aimed to develop a deep learning method for predicting HT after IVT for AIS using noncontrast computed tomography (NCCT) images.

METHODS

We retrospectively collected data from 828 AIS patients undergoing recombinant tissue plasminogen activator (rt-PA) treatment within a 4.5-h time window (n = 665) or of undergoing urokinase treatment within a 6-h time window (n = 163) and divided them into the HT group (n = 69) and non-HT group (n = 759). HT was defined based on the criteria of the European Cooperative Acute Stroke Study-II trial. To address the problems of indiscernible features and imbalanced data, a weakly supervised deep learning (WSDL) model for HT prediction was constructed based on multiple instance learning and active learning using admission NCCT images and clinical information in addition to conventional deep learning models. Threefold cross-validation and transfer learning were performed to confirm the robustness of the network. Of note, the predictive value of the commonly used scales in clinics associated with NCCT images (i.e., the HAT and SEDAN score) was also analysed and compared to measure the feasibility of our proposed DL algorithms.

RESULTS

Compared to the conventional DL and ML models, the WSDL model had the highest AUC of 0.799 (95% CI 0.712-0.883). Significant differences were observed between the WSDL model and five ML models (P < 0.05). The prediction performance of the WSDL model outperforms the HAT and SEDAN scores at the optimal operating point (threshold = 1.5). Further subgroup analysis showed that the WSDL model performed better for symptomatic intracranial haemorrhage (AUC = 0.833, F1 score = 0.909).

CONCLUSIONS

Our WSDL model based on NCCT images had relatively good performance for predicting HT in AIS and may be suitable for assisting in clinical treatment decision-making.

Assessing the Relationship between LAMS and CT Perfusion Parameters in Acute Ischemic Stroke Secondary to Large Vessel Occlusion

BACKGROUND

The Los Angeles Motor Scale (LAMS) is a rapid pre-hospital scale used to predict stroke severity which has also been shown to accurately predict large vessel occlusions (LVOs). However, to date there is no study exploring whether LAMS correlates with the computed tomography perfusion (CTP) parameters in LVOs.

METHODS

Patients with LVO between September 2019 and October 2021 were retrospectively reviewed and included if the CTP data and admission neurologic exams were available. The LAMS was documented based on emergency personnel exams or scored retrospectively using an admission neurologic exam. The CTP data was processed by RAPID (IschemaView, Menlo Park, CA, USA) with an ischemic core volume (relative cerebral blood flow [rCBF] < 30%), time-to-maximum (Tmax) volume (Tmax > 6 s delay), hypoperfusion index (HI), and cerebral blood volume (CBV) index. Spearman's correlations were performed between the LAMS and CTP parameters.

RESULTS

A total of 85 patients were included, of which there were 9 intracranial internal carotid artery (ICA), 53 proximal M1 branch middle cerebral artery M1, and 23 proximal M2 branch occlusions. Overall, 26 patients had LAMS 0-3, and 59 had LAMS 4-5. In total, LAMS positively correlated with CBF < 30% (Correlation Coefficient (CC): 0.32, p < 0.01), Tmax > 6 s (CC:0.23, p < 0.04), HI (CC:0.27, p < 0.01), and negatively correlated with the CBV index (CC:-0.24, p < 0.05). The relationships between LAMS and CBF were < 30% and the HI was more pronounced in M1 occlusions (CC:0.42, p < 0.01; 0.34, p < 0.01 respectively) and proximal M2 occlusions (CC:0.53, p < 0.01; 0.48, p < 0.03 respectively). The LAMS also correlated with a Tmax > 6 s in M1 occlusions (CC:0.42, p < 0.01), and negatively correlated with the CBV index in M2 occlusions (CC:-0.69, p < 0.01). There were no significant correlations between the LAMS and intracranial ICA occlusions.

CONCLUSIONS

The results of our preliminary study indicate that the LAMS is positively correlated with the estimated ischemic core, perfusion deficit, and HI, and negatively correlated with the CBV index in patients with anterior circulation LVO, with stronger relationships in the M1 and M2 occlusions. This is the first study showing that the LAMS may be correlated with the collateral status and estimated ischemic core in patients with LVO.

Symptomatic Intracranial Hemorrhage After Endovascular Stroke Treatment: External Validation of Prediction Models

BACKGROUND

Symptomatic intracranial hemorrhage (sICH) is a severe complication of reperfusion therapy for ischemic stroke. Multiple models have been developed to predict sICH or intracranial hemorrhage (ICH) after reperfusion therapy. We provide an overview of published models and validate their ability to predict sICH in patients treated with endovascular treatment in daily clinical practice.

METHODS

We conducted a systematic search to identify models either developed or validated to predict sICH or ICH after reperfusion therapy (intravenous thrombolysis and/or endovascular treatment) for ischemic stroke. Models were externally validated in the MR CLEAN Registry (n=3180; Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands). The primary outcome was sICH according to the Heidelberg Bleeding Classification. Model performance was evaluated with discrimination (c-statistic, ideally 1; a c-statistic below 0.7 is considered poor in discrimination) and calibration (slope, ideally 1, and intercept, ideally 0).

RESULTS

We included 39 studies describing 40 models. The most frequently used predictors were baseline National Institutes of Health Stroke Scale (NIHSS; n=35), age (n=22), and glucose level (n=22). In the MR CLEAN Registry, sICH occurred in 188/3180 (5.9%) patients. Discrimination ranged from 0.51 (SPAN-100 [Stroke Prognostication Using Age and National Institutes of Health Stroke Scale]) to 0.61 (SITS-SICH [Safe Implementation of Treatments in Stroke Symptomatic Intracerebral Hemorrhage] and STARTING-SICH [STARTING Symptomatic Intracerebral Hemorrhage]). Best calibrated models were IST-3 (intercept, -0.15 [95% CI, -0.01 to -0.31]; slope, 0.80 [95% CI, 0.50-1.09]), SITS-SICH (intercept, 0.15 [95% CI, -0.01 to 0.30]; slope, 0.62 [95% CI, 0.38-0.87]), and STARTING-SICH (intercept, -0.03 [95% CI, -0.19 to 0.12]; slope, 0.56 [95% CI, 0.35-0.76]).

CONCLUSIONS

The investigated models to predict sICH or ICH discriminate poorly between patients with a low and high risk of sICH after endovascular treatment in daily clinical practice and are, therefore, not clinically useful for this patient population.

Risk factors of hemorrhagic transformation in acute ischaemic stroke: A systematic review and meta-analysis

Background

Hemorrhagic transformation (HT) following reperfusion therapies for acute ischaemic stroke often predicts a poor prognosis. This systematic review and meta-analysis aims to identify risk factors for HT, and how these vary with hyperacute treatment [intravenous thrombolysis (IVT) and endovascular thrombectomy (EVT)].

Methods

Electronic databases PubMed and EMBASE were used to search relevant studies. Pooled odds ratio (OR) with 95% confidence interval (CI) were estimated.

Results

A total of 120 studies were included. Atrial fibrillation and NIHSS score were common predictors for any intracerebral hemorrhage (ICH) after reperfusion therapies (both IVT and EVT), while a hyperdense artery sign (OR = 2.605, 95% CI 1.212-5.599, I ² = 0.0%) and number of thrombectomy passes (OR = 1.151, 95% CI 1.041-1.272, I ² = 54.3%) were predictors of any ICH after IVT and EVT, respectively. Common predictors for symptomatic ICH (sICH) after reperfusion therapies were age and serum glucose level. Atrial fibrillation (OR = 3.867, 95% CI 1.970-7.591, I ² = 29.1%), NIHSS score (OR = 1.082, 95% CI 1.060-1.105, I ² = 54.5%) and onset-to-treatment time (OR = 1.003, 95% CI 1.001-1.005, I ² = 0.0%) were predictors of sICH after IVT. Alberta Stroke Program Early CT score (ASPECTS) (OR = 0.686, 95% CI 0.565-0.833, I ² =77.6%) and number of thrombectomy passes (OR = 1.374, 95% CI 1.012-1.866, I ² = 86.4%) were predictors of sICH after EVT.

Conclusion

Several predictors of ICH were identified, which varied by treatment type. Studies based on larger and multi-center data sets should be prioritized to confirm the results.

Systematic review registration

https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=268927, identifier: CRD42021268927.

Baseline Characteristics Associated with Good Collateral Status Using Hypoperfusion Index as an Outcome

Up to 30% of ischemic stroke cases are due to large vessel occlusion (LVO), causing significant morbidity. Studies have shown that the collateral circulation of patients with acute ischemic stroke (AIS) secondary to LVO can predict their clinical and radiological outcomes. The aim of this study is to identify baseline patient characteristics that can help predict the collateral status of these patients for improved triage. In this IRB approved retrospective study, consecutive patients presenting with AIS secondary to anterior circulation LVO were identified between September 2019 and August 2021. The baseline patient characteristics, laboratory values, imaging features and outcomes were collected using a manual chart review. From the 181 consecutive patients initially reviewed, 54 were confirmed with a clinical diagnosis of AIS and anterior circulation LVO. In patients with poor collateral status, the body mass index (BMI) was found to be significantly lower compared to those with good collateral status (26.4 ± 5.6 vs. 31.7 ± 12.3; p = 0.045). BMI of >35 kg/m2 was found to predict the presence of good collateral status. Age was found to be significantly higher (70.5 ± 9.6 vs. 58.9 ± 15.6; p = 0.034) in patients with poor collateral status and M1 strokes associated with older age and BMI.

Prognostic Role of the Platelet-Lymphocyte Ratio in Acute Ischemic Stroke Patients Undergoing Reperfusion Therapy: A Meta-Analysis

Background

Both inflammation and thrombotic/hemostatic mechanisms may play a role in acute ischemic stroke (AIS) pathogenesis, and a biomarker, such as the platelet-to-lymphocyte ratio (PLR), considering both mechanisms may be of clinical utility.

Objectives

This meta-analysis sought to examine the effect of PLR on functional outcomes, early neurological changes, bleeding complications, mortality, and adverse outcomes in AIS patients treated with reperfusion therapy (RT).

Design

Systematic Review and Meta-Analysis.

Data Sources and Methods

Individual studies were retrieved from the PubMed/Medline, EMBASE and Cochrane databases. References thereof were also consulted. Data were extracted using a standardised data sheet, and systematic reviews and meta-analyses on the association of admission (pre-RT) or delayed (post-RT) PLR with defined clinical and safety outcomes were conducted. In the case of multiple delayed PLR timepoints, the timepoint closest to 24 hours was selected.

Results

Eighteen studies (n=4878) were identified for the systematic review, of which 14 (n=4413) were included in the meta-analyses. PLR collected at admission was significantly negatively associated with 90-day good functional outcomes (SMD=-.32; 95% CI = -.58 to -.05; P=.020; z=-2.328), as was PLR collected at delayed timepoints (SMD=-.43; 95% CI = -.54 to -.32; P<.0001; z=-7.454). PLR at delayed timepoints was also significantly negatively associated with ENI (SMD=-.18; 95% CI = -.29 to -.08; P=.001. Conversely, the study suggested that a higher PLR at delayed timepoints may be associated with radiological bleeding and mortality. The results varied based on the type of RT administered.

Conclusions

A higher PLR is associated with worse outcomes after stroke in terms of morbidity, mortality, and safety outcomes after stroke.

Role of Neutrophil-Lymphocyte Ratio in the Prognosis of Acute Ischaemic Stroke After Reperfusion Therapy: A Systematic Review and Meta-analysis

Background

Inflammation may mediate response to acute reperfusion therapy (RT) in acute cerebral ischaemia. Neutrophil-lymphocyte ratio (NLR), an inflammatory biomarker, may play an important role in acute ischaemic stroke (AIS) prognostication.

Objective

This meta-analysis sought to examine the effect of NLR on functional outcomes, mortality and adverse outcomes in AIS patients receiving RT.

Methods

Individual studies were retrieved from PubMed/Medline, EMBASE and Cochrane databases. Data were extracted using a standardised data sheet and meta-analysis on association of admission (pre-RT) or delayed (post-RT) NLR with clinical/safety outcomes after RT was conducted.

Results

Thirty-five studies (n = 10 308) were identified for the systematic review with 27 (n = 8537) included in the meta-analyses. Lower admission NLR was associated with good functional outcomes (GFOs), defined as 3-month modified Rankin scale (mRS) 0–2 (SMD = −.46; 95% CI = −.62 to −.29; P < .0001), mRS 0–1 (SMD = −.44; 95% CI = −.66 to −.22; P < .0001) and early neurological improvement (ENI) (SMD = −.55; 95 %CI = −.84 to −.25; P < .0001). Lower delayed admission NLR was also associated with GFOs (SMD = −.80; 95%CI = −.91 to −.68; P < .0001). Higher admission NLR was significantly associated with mortality (SMD = .49; 95%CI = .12 to .85; P = .009), intracerebral haemorrhage (ICH) (SMD = .34; 95% CI = .09 to .59; P = .007), symptomatic ICH (sICH) (SMD = .48; 95% CI = .07 to .90; P = .022) and stroke-associated infection or pneumonia (SMD = .85; 95% CI = .50, 1.19; P < .0001). Higher delayed NLR was significantly associated with sICH (SMD = 1.40; 95% CI = .60 to 2.19; P = .001), ICH (SMD = .94; 95% CI = .41 to 1.46; P < .0001) and mortality (SMD = 1.12; 95% CI = .57 to 1.67; P < .0001). There were variations in outcomes across RT groups.

Conclusion

Higher admission or delayed NLR is significantly associated with worse morbidity, mortality and safety outcomes in AIS patients receiving RT.