Predicting Functional Outcome Using 24-Hour Post-Treatment Characteristics: Application of Machine Learning Algorithms in the STRATIS Registry.

FOR SOCIAL MEDIA: @AliciaCastongu2, @FazalZaidi9, @oozaidat, @Mouhammad_Jumaa OBJECTIVE: Machine learning (ML) algorithms have emerged as powerful predictive tools in the field stroke. Here, we examine the predictive accuracy of ML models for predicting functional outcomes using 24-hour post-treatment characteristics in the Systematic Evaluation of Patients Treated With Neurothrombectomy Devices for Acute Ischemic Stroke (STRATIS) Registry. METHODS: ML models, adaptive boost, random forest (RF), classification and regression trees (CART), C5.0 decision tree (C5.0), support vector machine (SVM), least absolute shrinkage and selection operator (LASSO), and logistic regression (LR), and traditional LR models were used to predict 90-day functional outcome (modified Rankin Scale score 0-2). Twenty-four-hour National Institutes of Health Stroke Scale (NIHSS) was examined as a continuous or dichotomous variable in all models. Model accuracy was assessed using the area under characteristic curve (AUC). RESULTS: The 24-hour NIHSS score was a top-predictor of functional outcome in all models. ML models using the continuous 24-hour NIHSS scored showed moderate-to-good predictive performance (range mean AUC: 0.76-0.92); however, RF (AUC: 0.92 ± 0.028) outperformed all ML models, except LASSO (AUC: 0.89 ± 0.023, p = 0.0958). Importantly, RF demonstrated a significantly higher predictive value than LR (AUC: 0.87 ± 0.031, p = 0.048) and traditional LR (AUC: 85 ± 0.06, p = 0.035) when using the 24-hour continuous NIHSS score. Predictive accuracy was similar between the 24-hour NIHSS score dichotomous and continuous ML models. INTERPRETATION: In this substudy, we found similar predictive accuracy for functional outcome when using the 24-hour NIHSS score as a continuous or dichotomous variable in ML models. ML models had moderate-to-good predictive accuracy, with RF outperforming LR models. External validation of these ML models is warranted. ANN NEUROL 2023;93:40-49.

Association of chronic kidney disease with cerebral microbleeds in patients with primary intracerebral hemorrhage.

BACKGROUND AND PURPOSE: To investigate the relationship between chronic kidney disease (CKD) and MRI-defined cerebral microbleeds (CMB), a harbinger of future intracerebral hemorrhage (ICH), among patients with a recent history of primary ICH. METHODS: Using data from a predominantly black cohort of patients with a recent ICH-enrolled in an observational study between September 2007 and June 2011, we evaluated the association between CKD (defined as estimated low glomerular filtration rate<60 mL/min per 1.73 m(2)) and CMB on gradient-echo MRI. Multivariable models were generated to determine the contribution of CKD to the presence, number, and location of CMB. RESULTS: Of 197 subjects with imaging data, mean age was 59 years, 48% were women, 73% were black, 114 (58%) had ≥1 CMBs, and 52 (26%) had CKD. Overall, CKD was associated with presence of CMB (adjusted odds ratio, 2.70; 95% confidence interval [CI], 1.10-6.59) and number of CMB (adjusted relative risk, 2.04; 95% CI, 1.27-3.27). CKD was associated with CMB presence (adjusted odds ratio, 3.44; 95% CI, 1.64-7.24) and number (adjusted relative risk, 2.46; 95% CI, 1.11-5.42) in black patients, but not CMB presence (adjusted odds ratio, 3.00; 95% CI, 0.61-14.86) or number (adjusted relative risk, 1.03; 95% CI: 0.22-4.89) in non-Hispanic white patients (interactions by race were statistically not significant). CONCLUSIONS: CKD is associated with a greater presence and number of CMB in ICH patients, particularly in patients of black race. Future studies should assess whether low estimated glomerular filtration rate may be a CMB risk marker or potential therapeutic target for mitigating the development of CMB.

Predictors of highly prevalent brain ischemia in intracerebral hemorrhage.

OBJECTIVE: This study was undertaken to determine the prevalence, characteristics, risk factors, and temporal profile of concurrent ischemic lesions in patients with acute primary intracerebral hemorrhage (ICH). METHODS: Patients were recruited within a prospective, longitudinal, magnetic resonance imaging (MRI)-based study of primary ICH. Clinical, demographic, and MRI data were collected on all subjects at baseline and 1 month. RESULTS: Of the 138 patients enrolled, mean age was 59 years, 54% were male, 73% were black, and 84% had a history of hypertension. At baseline, ischemic lesions on diffusion-weighted imaging (DWI) were found in 35% of patients. At 1 month, lesions were present in 27%, and of these lesions, 83% were new and not present at baseline. ICH volume (p = 0.025), intraventricular hemorrhage (p = 0.019), presence of microbleeds (p = 0.024), and large, early reductions in mean arterial pressure (p = 0.003) were independent predictors of baseline DWI lesions. A multivariate logistical model predicting the presence of 1-month DWI lesions included history of any prior stroke (p = 0.012), presence of 1 or more microbleeds (p = 0.04), black race (p = 0.641), and presence of a DWI lesion at baseline (p = 0.007). INTERPRETATION: This study demonstrates that >⅓ of patients with primary ICH have active cerebral ischemia at baseline remote from the index hematoma, and » of patients experience ongoing, acute ischemic events at 1 month. Multivariate analyses implicate blood pressure reductions in the setting of an active vasculopathy as a potential underlying mechanism. Further studies are needed to determine the impact of these lesions on outcome and optimal management strategies to arrest vascular damage.

Poor long-term blood pressure control after intracerebral hemorrhage.

BACKGROUND AND PURPOSE: Hypertension is the most important risk factor associated with intracerebral hemorrhage. We explored racial differences in blood pressure (BP) control after intracerebral hemorrhage and assessed predictors of BP control at presentation, 30 days, and 1 year in a prospective cohort study. METHODS: Subjects with spontaneous intracerebral hemorrhage were identified from the DiffErenCes in the Imaging of Primary Hemorrhage based on Ethnicity or Race (DECIPHER) Project. BP was compared by race at each time point. Multivariable linear regression was used to determine predictors of presenting mean arterial pressure, and longitudinal linear regression was used to assess predictors of mean arterial pressure at follow-up. RESULTS: A total of 162 patients were included (mean age, 59 years; 53% male; 77% black). Mean arterial pressure at presentation was 9.6 mm Hg higher in blacks than whites despite adjustment for confounders (P=0.065). Fewer than 20% of patients had normal BP (<120/80 mm Hg) at 30 days or 1 year. Although there was no difference at 30 days (P=0.331), blacks were more likely than whites to have Stage I/II hypertension at 1 year (P=0.036). Factors associated with lower mean arterial pressure at follow-up in multivariable analysis were being married at baseline (P=0.032) and living in a facility (versus personal residence) at the time of BP measurement (P=0.023). CONCLUSIONS: Long-term BP control is inadequate in patients after intracerebral hemorrhage, particularly in blacks. Further studies are needed to understand the role of social support and barriers to control to identify optimal approaches to improve BP in this high-risk population.

Use of MRI in the assessment of patients with stroke.

Multimodal neuroimaging is assuming an increasingly important role in the initial evaluation and management of acute stroke patients in parallel with the expansion of therapeutic options. Multimodal MRI can identify the type of stroke (ischemia or hemorrhage), severity and location of the lesion, the patency of the intracranial vessels, the degree of cerebral perfusion, and the presence and size of the ischemic penumbra. This information can be used to guide both acute and long-term treatment decisions for stroke patients.

Development and validation of a simple conversion model for comparison of intracerebral hemorrhage volumes measured on CT and gradient recalled echo MRI.

BACKGROUND AND PURPOSE: Gradient recalled echo MRI (GRE) has been shown to be as accurate as CT for the detection of acute intracerebral hemorrhage (ICH). However, because of the differences in the signal parameter being detected, apparent hemorrhage size is expected to vary by imaging modality, with GRE providing larger volumes attributable to susceptibility effects. METHODS: Image data from patients participating in 3 ICH studies were retrospectively reviewed. Patients with acute ICH were included if (1) concurrent MRI and CT were performed within 72 hours of symptom onset, and (2) each modality was performed within 240 minutes of each other. ICH volumes were calculated using a semiautomated image analysis program. The least squares method was used to develop a conversion equation based on a linear regression of GRE volume on CT volume. RESULTS: Thirty-six patients met inclusion criteria. MRI was performed first in 18, CT first in 18. Mean hemorrhage volume was 25.2 cc (range 0.1 to 83.9 cc) on CT and 32.7 cc (range 0.1 to 98.7 cc) measured on GRE. A linear relationship defined by CT Volume=GRE Volume*0.8 (Spearman's correlation coefficient=0.992, P<0.001) was derived. CONCLUSIONS: Acute ICH volumes as measured on GRE pulse sequences are consistently larger than CT volumes. A simple mathematical conversion model has been developed: CT volume=0.8*GRE volume. This formula can be used in studies using both imaging modalities, across different studies, or to track ICH growth over time independent of imaging modality in an individual patient.

High rate of microbleed formation following primary intracerebral hemorrhage.

BACKGROUND: We sought to investigate the frequency of microbleed development following intracerebral hemorrhage in a predominantly African-American population and to identify predictors of new microbleed formation. AIMS AND/OR HYPOTHESIS: To investigate the frequency and predictors of new microbleeds following intracerebral hemorrhage. METHODS: The DECIPHER study was a prospective, longitudinal, magnetic resonance-based cohort study designed to evaluate racial/ethnic differences in risk factors for microbleeds and to evaluate the prognostic impact of microbleeds in this intracerebral hemorrhage population. We evaluated new microbleed formation in two time periods: from baseline to 30 days and from 30 days to year 1. RESULTS: Of 200 subjects enrolled in DECIPHER, 84 had magnetic resonance imaging at all required time points to meet criteria for this analysis. In the baseline to day 30 analysis, 11 (13·1%) had new microbleeds, compared with 25 (29·8%) in the day 30 to year 1 analysis. Logistic regression analysis demonstrated that baseline number of microbleeds [odds ratio 1·05 (95% confidence interval 1·01, 1·08), P = 0·01] was associated with new microbleed formation at 30 days. A logistic regression model predicting new microbleed at one-year included baseline number of microbleeds [odds ratio 1·05 (1·00, 1·11), P = 0·046], baseline age [odds ratio 1·05 (1·00, 1·10), P = 0·04], and white matter disease score [odds ratio 1·18 (0·96, 1·45). P = 0·115]. Overall, 28 of 84 (33·3%) intracerebral hemorrhage subjects formed new microbleeds at some point in the first year post-intracerebral hemorrhage. CONCLUSIONS: We found that one-third of intracerebral hemorrhage subjects in this cohort surviving one-year developed new microbleeds, which suggests a dynamic and rapidly progressive vasculopathy. Future studies are needed to examine the impact of new microbleed formation on patient outcomes.

Racial differences by ischemic stroke subtype: a comprehensive diagnostic approach.

Background. Previous studies have suggested that black populations have more small-vessel and fewer cardioembolic strokes. We sought to analyze racial differences in ischemic stroke subtype employing a comprehensive diagnostic workup with magnetic resonance-imaging-(MRI-) based evaluation including diffusion-weighted imaging (DWI). Methods. 350 acute ischemic stroke patients admitted to an urban hospital with standardized comprehensive diagnostic evaluations were retrospectively analyzed. Ischemic stroke subtype was determined by three Trial of Org 10172 in Acute Stroke Treatment (TOAST) classification systems. Results. We found similar proportions of cardioembolic and lacunar strokes in the black and white cohort. The only subtype category with a significant difference by race was "stroke of other etiology," more common in whites. Black stroke patients were more likely to have an incomplete evaluation, but this did not reach significance. Conclusions. We found similar proportions by race of cardioembolic and lacunar strokes when employing a full diagnostic evaluation including DWI MRI. The relatively high rate of cardioembolism may have been underappreciated in black stroke patients when employing a CT approach to stroke subtype diagnosis. Further research is required to better understand the racial differences in frequency of "stroke of other etiology" and explore disparities in the extent of diagnostic evaluations.