
==== Front
Neurol Ther
Neurol Ther
Neurology and Therapy
2193-8253
2193-6536
Springer Healthcare Cheshire

39136813
652
10.1007/s40120-024-00652-3
Original Research
Quantitative Analysis of White Matter Hyperintensities as a Predictor of 1-Year Risk for Ischemic Stroke Recurrence
Sun Yi 1
Xia Wenping 2
Wei Ran 1
Dai Zedong 1
Sun Xilin 1
Zhu Jie 1
Song Bin songbin@fudan.edu.cn

1
http://orcid.org/0000-0002-5438-1211
Wang Hao wang_h@fudan.edu.cn

1
1 https://ror.org/013q1eq08 grid.8547.e 0000 0001 0125 2443 Department of Radiology, Minhang Hospital, Fudan University, 170 Xinsong Road, Shanghai, 201199 People’s Republic of China
2 Department of Radiology, Ningbo Yinzhou No. 2 Hospital, Ningbo, China
13 8 2024
13 8 2024
10 2024
13 5 14671482
3 6 2024
25 7 2024
© The Author(s) 2024
2024
https://creativecommons.org/licenses/by-nc/4.0/ Open Access This article is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, which permits any non-commercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc/4.0/.
Introduction

This study evaluates the role of quantitative characteristics of white matter hyperintensities (WMHs) in predicting the 1-year recurrence risk of ischemic stroke.

Methods

We conducted a retrospective analysis of 1061 patients with ischemic stroke from January 2018 to April 2021. WMHs were automatically segmented using a cluster-based method to quantify their volume and number of clusters (NoC). Additionally, two radiologists independently rated periventricular and deep WMHs using the Fazekas scale. The cohort was divided into a training set (70%) and a testing set (30%). We employed Cox proportional hazards models to develop predictors based on quantitative WMH characteristics, Fazekas scores, and clinical factors, and compared their performance using the concordance index (C-index).

Results

A total of 180 quantitative variables related to WMHs were extracted. A higher NoC in deep white matter and brainstem, advanced age (> 90 years old), specific stroke subtypes, and absence of discharge antiplatelets showed stronger associations with the risk of ischemic stroke recurrence within 1 year. The nomogram incorporating quantitative WMHs data showed superior discrimination compared to those based on the Fazekas scale or clinical factors alone, with C-index values of 0.709 versus 0.647 and 0.648, respectively, in the testing set. Notably, a combined model including both WMHs and clinical factors achieved the highest predictive accuracy, with a C-index of 0.735 in the testing set.

Conclusion

Quantitative assessment of WMHs provides a valuable neuro-imaging tool for enhancing the prediction of ischemic stroke recurrence risk.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40120-024-00652-3.

Keywords

White matter hyperintensities
Ischemic stroke
Recurrence
Magnetic resonance imaging
The Natural Science Foundation of Minhang Hospital, Fudan University2022MHPY04 2022MHBJ04 Sun Yi Wang Hao Science and Health Commission of Minhang Districtmwyjyx17 Sun Yi issue-copyright-statement© Springer Healthcare Ltd., part of Springer Nature 2024
==== Body
pmcKey Summary Points

Why carry out this study?	
White matter hyperintensity is associated with poor outcomes after acute ischemic stroke.	
We hypothesize that quantitative analysis of WMHs could be used for the estimation of ischemic stroke recurrence.	
What was learned from this study?	
The nomogram based on quantitative WMHs demonstrated superior discriminatory ability in predicting stroke recurrence compared to the Fazekas scale-based nomogram.	
The total number of clusters observed in deep white matter and brainstem exhibited a stronger association with the occurrence of ischemic stroke.	

Introduction

White matter hyperintensities (WMHs), commonly observed on fluid-attenuated inversion recovery (FLAIR) images from brain magnetic resonance imaging (MRI), increase in prevalence with age. For instance, among individuals over 60 years, WMHs appear in about 30% of healthy subjects [1]. Extensive research has linked WMHs with various neurological conditions: they are indicators of cerebral small vessel disease (CSVD) stages [2–8], associated with cognitive impairment [5, 9, 10], influence the disease severity in frontotemporal dementia [2], and heighten the risk of stroke [3, 4, 7, 8, 11, 12]. Neuroimaging studies have identified several key features indicative of CSVD, including recent subcortical infarcts, lacunar infarctions, white matter hyperintensities, perivascular spaces, microbleeds, and brain atrophy [13]. Although the etiology underlying WMHs remains unclear, presumed pathological changes in structures such as gliosis, axonal degeneration, myelin loss, and vacuolation may reflect “macrostructural” damage [8, 14, 15]. Furthermore, the morphological heterogeneity of WMHs—varying shapes, distributions, and sizes—suggests different underlying parenchymal changes [7, 14, 16]. Histopathological studies show that periventricular WMHs (PV-WMHs) correlate with mild non-ischemic changes, implying additional gliosis [4, 15], while the confluence in deep WMHs (D-WMHs) indicates ischemic damage, often accompanied by significant fiber loss and arteriolosclerosis [15].

Stroke remains a leading global cause of death and disability [17, 18], with ischemic stroke accounting for approximately 80% of all stroke cases [18]. Despite this, stroke recurrence receives less attention than initial stroke events [19]. Recurrent strokes carry a higher risk of disability, dementia, and mortality [20]. A meta-analysis reported a 1-year recurrence risk of 11% [21], contrasted by a recent Danish national registry study, which calculated a 1-year recurrence risk of 4% when adjusting for competing risks [19].

Increasing evidence has consistently demonstrated that the volume of WMHs serves as a robust neuroimaging marker for predicting functional outcomes following ischemic stroke [5, 6, 22–26]. A multicenter quantitative brain MRI study to demonstrate that advanced WMH affects post-stroke outcomes, both in the general stroke population and with possibly differential impact in different stroke subtypes [27]. Another study suggests that quantitative and qualitative assessments are highly correlated and comparable in patients with TIA/minor stroke. White matter lesion burden is associated with short-term outcomes of patients with good pre-stroke function in the presence of intracranial stenosis/occlusion [28]. A comprehensive study involving a large-scale sample of 7101 patients reported an independent association between WMH volume load and increased risk of stroke recurrence and mortality [29]. Given these associations, this study hypothesizes that the spatial distribution and quantification of WMHs, in terms of both volume and number of clusters (NoC), could significantly influence the probability of recurrent ischemic stroke. Consequently, this investigation aims to examine the prognostic significance of these quantitative WMH characteristics in predicting ischemic stroke recurrence.

Methods

Study Population

This current retrospective study was approved by the Institutional Ethics Committee of Minhang Hospital, Fudan University, with a waiver for informed consent granted (approval number: 2023-022-01 K). The study was performed in accordance with the 1964 Declaration of Helsinki and its later amendments.

This retrospective study enrolled 1580 patients (≥ 18 years old) with first-ever ischemic stroke confirmed by our stroke center between January 2018 and April 2021. Exclusion criteria included cerebral hemorrhage, traumatic brain injury, previous neurological or psychiatric disorder, cerebral tumor, history of substance abuse, severe MRI artifacts, contradiction to MR examination, death during follow-up, or loss to follow-up.

Clinical Data

Demographic and clinical data were collected, including age, sex, alcohol and tobacco use history, hypertension, hyperlipidemia, diabetes, atrial fibrillation, the initial National Institutes of Health Stroke Scale (NIHSS) score [30], Trial of Org 10,172 in Acute Stroke Treatment (TOAST) classification of stroke subtype [22], post-discharge medications (antiplatelets, anticoagulants, statins), circulation territory involved (anterior, posterior, or both), and the modified Rankin Scale (mRS) score [31] at 90 days. The workflow of the study is illustrated in Fig. 1.Fig. 1 Study population flowchart

MRI Acquisition

Brain MRI was performed on scanner 1 (EXCITE HD 1.5 T MRI; GE Healthcare, Milwaukee, WI, USA) and scanner 2 (uMR780 3.0 T MRI; United Imaging Healthcare, Shanghai, China). Scanning protocol included axial T2-weighted imaging (T2-WI), T1-weighted imaging (T1-WI), FLAIR, and diffusion-weighted imaging (DWI). The scan parameters are shown in the supplementary material (e-1).

Outcome Assessment

Within the initial 21-day period following the first stroke, recurrent ischemic stroke was defined as the emergence of a novel neurological deficit or deterioration of an existing neurological deficit accompanied by newly identified discrete lesions on DWI. Beyond this period, recurrent ischemic stroke was characterized by the sudden onset of focal neurologic deficits corresponding to confirmed infarct lesions on DWI [29].

We captured mRS at 90 days and recorded stroke recurrence 1 year after the qualifying event by conducting in-person or telephone interviews with patients or their caregivers. Verification of all recurrent ischemic stroke cases was performed using DWI at our center or through reports from other institutions.

Fazekas Scale Assessment

Two radiologists independently evaluated the severity of D-WMHs and PV-WMHs, assigning scores from 0 to 3 [8, 32]. To ensure the reliability of these assessments, interobserver agreement for 100 randomly selected patients was evaluated using intraclass correlation coefficients.

WMHs Segmentation and Extraction

WMHs segmentation was performed using the UBO detector toolbox (https://cheba.unsw.edu.au/group/neuroimaging-pipeline) [33]. The process involved multiple steps:

WMHs Segmentation

The segmentation pipeline for WMHs is illustrated in Fig. 2. The comprehensive details of the segmentation process are provided below:Coregistration: A linear registration was performed with T1-WI as the reference image and FLAIR as the source image. Subsequently, the re-sliced FLAIR image was further resampled.

Normalization: Gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) segmentation on individual T1-WI sequences were initially performed using Statistical Parametric Mapping (SPM12; University College London). Subsequent steps included employing Diffeomorphic Anatomical Registration Through Exponentiated Lie algebra (DARTEL) to align individual T1-WI images to a common DARTEL template and generate flow fields.

Registration to DARTEL Space: Using the generated flow fields, co-registered FLAIR images were mapped to DARTEL space.

Segmentation: WMHs were differentiated from white and gray matter plus cerebrospinal fluid using FMRIB's Automated Segmentation Tool (FAST) on FLAIR images.

WMHs identification: The K-nearest neighbors (K-NN) algorithm, with a threshold of 0.8, identified WMHs.

Inverse deformation: This process mapped WMHs from DARTEL back to FLAIR space.

Manual correction: Manual correction was performed to exclude acute infarct lesions on FLAIR, referencing corresponding DWI images.

Re-registration: Finally, WMHs in FLAIR space were registered into the DARTEL standard space.

Fig. 2 WHMs segmentation pipeline. Orig original, Flair fluid-attenuated inversion recovery, MNI Montreal Neurological Institute, FAST FMRIB’s Automated Segmentation Tool, KNN K-nearest neighbors

WMHs Features Extraction

Anatomical locations and cluster sizes of WMHs were systematically quantified. The anatomical location included the PV-WMHs, D-WMHs, and lobes (periventricular, and deep white matter as well as different lobes, PDL strategy), as well as arterial methods. In the PDL strategy, a distance threshold of 10 mm from the ventricular border was utilized to identify PV-WMHs and D-WMHs [33]. In the arterial strategy [34], the WMHs maps were segmented into eight arterial territories on the left or right hemisphere [anterior artery hemisphere (AAH), middle artery hemisphere (MAH); posterior artery hemisphere (PAH), anterior artery callosal (AAC), middle artery lateral lenticulostriate (MALL), posterior artery thalamic and midbrain perforators (PATMP), anterior artery medial lenticulostriate (AAML), posterior artery callosal (PAC)].

WMHs clusters were defined as 26-connected neighboring voxels (six faces, twelve edges, and eight vertices) [33]. The NoC is a measure of interest in different anatomical locations. WMHs clusters size was classified as punctate (< 3 voxels on DARTEL space, i.e., < 10.125 mm3), focal (< 9 voxels, i.e., < 30.375 mm3), medium (< 15 voxels, i.e., < 50.625 mm3), and confluent (≥ 15 voxels, i.e., ≥ 50.625 mm3) WMHs [33].

To ensure clarity in hemispheric analysis, particularly in stroke-affected cases, references were made to the contralateral hemisphere. We calculated the volumes of all WMHs in various anatomical locations. Additionally, we computed the NoC with different sizes and the total NoC using both PDL and arterial strategies.

Statistical Analysis

The statistical analysis was performed with R software (v.3.6.2, http://www.R-project.org). All tests of statistical significance were two-sided, with p values less than 0.05 considered statistically significant.

Data Distribution

The study participants were randomly divided into two sets, the training set (70% of cases) and the testing set (30% of cases). The differences in age, gender, stage, circulation subtypes, stroke subtypes, previous alcohol intake, previous smoking, hypertension, diabetes mellitus, hyperlipidemia, atrial fibrillation, discharge medications, mRS at 90 days, and Fazekas scale between the training and testing set were evaluated by using an independent samples t test, χ2 test, or Mann–Whitney U test, where appropriate.

Construction of Cox Proportional Hazards Model

All continuous variables, including WMHs volume, NoC, and age were classified into high-risk or low-risk groups according to stroke recurrence event by using Kaplan–Meier survival analysis, whose optimal threshold was identified corresponding to the minimum p value. The least absolute shrinkage and selection operator (LASSO) Cox regression model, which is suitable for the regression of high-dimensional data (a total of 180 WMHs variables in this study, supplementary material e-2), was used to select the most useful prognostic features in the training dataset. Features between the two groups were selected by univariate Cox regression analysis (p < 0.05). Spearman correlation analysis was used to further eliminate the features with correlation > 0.8. Multivariate Cox hazards regression (backward step-down selection) was used to select the features and construct the prediction model. The quantitative WMHs model, clinical characteristics-based models, and a Fazekas scale-based model were finally constructed.

Performance and Comparison in Different Prediction Models for Ischemic Stroke Recurrence Estimation

Based on the multivariate Cox analysis, a total of five different prediction models including a clinical characteristics-based model, a subjective Fazekas scale-based model, a quantitative WMHs-based model, a Fazekas scale combined clinical characteristics, and WMHs combined with clinical characteristics were constructed and compared. Moreover, nomograms were presented in the testing dataset. The Harrell concordance index (C-index) was calculated to demonstrate the discrimination performance, along with the concordance probability estimate considering the high degree of censoring in our data (1 indicates perfect concordance; 0.5 indicates no better concordance than chance) [35]. A decision curve analysis (DCA) was utilized to assess the clinical usefulness of the five nomograms by quantifying the net benefits for multiple threshold probabilities obtained by the DCA.

Results

Data Distribution Results

No difference was found between the training dataset and the testing dataset in demographic and clinical characteristics (all p > 0.05; supplementary material e-3).

Demographic and Clinical Characteristics

A total of 519 patients were excluded from the study because of the following reasons: (1) cerebral hemorrhage (n = 21); (2) traumatic brain injury (n = 7); (3) previous neurological or psychiatric disorders (n = 163); (4) severe MRI artifacts (n = 17); (5) contradiction to MR examination (n = 9); (6) lost to follow-up (n = 280); and (7) death (n = 22).

Finally, a total of 1061 patients (mean age 65.84 ± 12.14) were included in the current study, 88 (8.3%) of whom were with ischemic stroke recurrence within 1 year after the qualifying event. Patients with recurrent stroke (69.97 ± 12.49) were older than non-recurrence patients (65.46 ± 12.04). Univariate Cox proportional hazard regression analysis showed that stroke subtypes such as large artery atherosclerosis (LAA) and cardioembolism (CE) had a higher risk of recurrent stroke compared to other stroke subtypes (p < 0.05). We also found that previous alcohol intake and lack of antiplatelets after discharge showed a higher risk of recurrence (p < 0.05) (Table 1).Table 1 Univariate Cox regression analysis between variables and stroke recurrence within 1 year after ischemic stroke

Variable	Stroke recurrence	Hazard ratio	95%CI	p value	
Lower	Upper	
No (n = 973)	Yes (n = 88)	
Age, years	65.46 ± 12.04	69.97 ± 12.49	6.51	2.04	20.79	0.002	
Men, n (%)	671 (69.0)	59 (67.0)	1.15	0.67	1.98	0.607	
Circulation territory	0.96	0.60	1.53	0.856	
Anterior, n (%)	641 (65.9)	61 (69.3)					
Posterior, n (%)	306 (31.4)	23 (26.1)					
 Both, n (%)	26 (2.7)	4 (4.5)					
Stroke subtypes (TOAST)	0.66	0.51	0.85	0.002	
LAA, n (%)	481 (49.4)	56 (63.6)					
CE, n (%)	75 (7.7)	8 (9.1)					
SVO, n (%)	335 (34.4)	22 (25.0)					
OD, n (%)	10 (1.0)	0 (0.0)					
Undetermined, n (%)	72 (7.4)	2 (2.3)					
Previous alcohol intake, n (%)	137 (14.1)	8 (9.1)	0.20	0.05	0.81	0.024	
Previous smoking, n (%)	358 (36.8)	37 (42.0)	1.05	0.63	1.76	0.849	
Hypertension, n (%)	659 (67.7)	59 (67.0)	0.95	0.56	1.61	0.847	
Diabetes mellitus, n (%)	333 (34.2)	31 (35.2)	0.81	0.47	1.40	0.450	
Hyperlipidemia, n (%)	268 (27.5)	22 (25.0)	0.83	0.46	1.49	0.532	
Atrial fibrillation, n (%)	98 (10.1)	12 (13.6)	1.27	0.60	2.67	0.529	
Discharge statins	611 (62.8)	66 (75.0)	1.72	0.97	3.04	0.063	
Discharge anticoagulants	40 (4.1)	8 (9.1)	1.77	0.71	4.42	0.222	
Discharge antiplatelets	889 (91.4)	73 (83.0)	0.42	0.22	0.79	0.007	
Admissio n NIHSS scorea	3 (1, 4)	3 (1, 4)	0.97	0.89	1.05	0.461	
mRS at 90 daysa	0 (0, 1)	0 (0, 1)	1.03	0.85	1.25	0.769	
Fazekas_pva	2 (1, 2)	2 (1, 3)	1.72	1.25	2.36	< 0.001	
Fazekas_deepa	1 (1, 2)	2 (1, 2)	1.32	1.12	1.54	< 0.001	
TOAST trial of org 10,172 in acute stroke treatment stroke, LAA large artery atherosclerosis, CE cardioembolism, SVO small vessel occlusion, OD other determined, NIHSS the National Institutes of Health Stroke Scale, mRS modified Rankin Scale

aWe employ median (lower quartile, upper quartile) as a statistical measure to depict ranked data such as ‘Admission NIHSS score’, ‘mRS at 90 days’, ‘Fazekas_pv’, and ‘Fazekas_deep’

Fazekas Scale Assessment and Univariate Cox Proportional Hazard Regression

The interobserver reproducibility for the assessment of the Fazekas scale was 0.89. Therefore, the Fazekas scale was recorded based on the assessments of the first radiologist (Y.S.). Univariate Cox proportional hazard regression showed that, compared to a lower Fazekas scale, both PV-WMHs (HR = 1.72, 95% CI 1.25–2.36) and D-WMHs (HR = 1.32, 95% CI 1.12–1.54) with a higher Fazekas scale had a higher risk of ischemic stroke recurrence (p < 0.05) (Table 1).

WMHs Extraction and Univariate Cox Proportional Hazard Regression

A total of 30 related to WMHs volume variables and 30 related to WMHs NoC variables (PDL strategy, n = 14; arterials strategy, n = 16) were extracted. In addition, 120 WMHs NoC with different cluster sizes [punctate, focal, medium, and confluent; n = (14 + 16) × 4 = 120] in different lobe and arterial territories were further extracted. Finally, a total of 180 WMHs quantitative variables were extracted (supplementary material e-2).

Univariate Cox proportional hazard regression analysis showed that 59 WMHs variables were associated with ischemic stroke recurrence (supplementary material e-2).

Construction and Validation of Different Prediction Models

Quantitative WMHs-based model (Model 1): LASSO multiple Cox proportional hazard regression analysis demonstrated that three WMHs volume variables (Ltemporal_WMHvol_mm3, Lparietal_WMHvol_mm3, rMALL_WMHvol_mm3) and six WMHs NoC variables (lAAH_WMHnoc_total, Rfrontal_WMHnoc_punctuate, rPAH_WMHnoc_focal, wholeBrain_WMHnoc_medium, lAAH_WMHnoc_medium, lMALL_WMHnoc_confluent) were associated with risk of ischemic stroke recurrence (all p < 0.05; Table 2).Table 2 Bivariate Cox regression analysis between variables and stroke recurrence at 1 year after ischemic stroke

Model	Variable	Hazard ratio	95% CI	P value	
Model 1	Ltemporal_WMHvol_mm3	0.18	[0.05, 0.644]	0.008	
Lparietal_WMHvol_mm3	2.00	[1.13, 3.52]	0.017	
rAAH_WMHvol_mm3	307,322.11	[0.00, 7.39E186]	0.953	
rMALL_WMHvol_mm3	2.47	[1.42, 4.30]	0.001	
lAAH_WMHnoc_total	0.20	[0.10, 0.43]	 < 0.001	
Rfrontal_WMHnoc_punctuate	2.90	[1.45, 5.79]	0.003	
rPAH_WMHnoc_focal	4.54	[1.87, 11.04]	0.001	
wholeBrain_WMHnoc_medium	0.41	[0.18, 0.93]	0.032	
lAAH_WMHnoc_medium	4.15	[1.80, 9.58]	0.001	
lMALL_WMHnoc_confluent	2.77	[1.22, 6.29]	0.015	
Model 2	Age	1.04	[1.01, 1.06]	0.002	
Stroke subtype (TOAST)	0.74	[0.58, 0.95]	0.019	
Smoke status	1.74	[1.03, 2.93]	0.039	
Model 3	Fazekas_total	1.33	[1.13, 1.57]	0.001	
Model 4	Ltemporal_WMHvol_mm3	0.12	[0.03, 0.42]	0.001	
Lparietal_WMHvol_mm3	2.06	[1.18, 3.60]	0.011	
rAAH_WMHvol_mm3	364,791.96	[0.00, 5.68E187]	0.952	
rMALL_WMHvol_mm3	2.35	[1.35, 4.09]	0.003	
lAAH_WMHnoc_total	0.19	[0.09, 0.40]	 < 0.001	
Rfrontal_WMHnoc_punctuate	2.56	[1.33, 4.92]	0.005	
rPAH_WMHnoc_focal	3.85	[1.57, 9.45]	0.003	
lAAH_WMHnoc_medium	4.26	[1.86, 9.80]	0.001	
lMALL_WMHnoc_confluent	3.12	[1.35, 7.22]	0.008	
Stroke subtype (TOAST)	0.70	[0.55, 0.90]	0.005	
Model 5	Stroke subtype (TOAST)	0.72	[0.57, 0.93]	0.01	
Discharge_Anticoagulants	2.42	[1.04, 5.64]	0.041	
Fazekas_total	1.34	[1.13, 1.58]	0.001	
Model 1 quantitative WMHs-based model. Model 2 clinical characteristics-based model, Model 3 subjective Fazekas scale-based model, Model 4 quantitative WMHs combined with clinical characteristics model, Model 5 Fazekas scale combined with clinical characteristics-based model, WMHs white matter hyperintensities, Ltemporal_WMHvol_mm3 WMHs volume in left temporal lobe, Lparietal_WMHvol_mm3 WMHs volume in left parietal lobe, rAAH_WMHvol_mm3 WMHs volume in right anterior artery hemisphere, rMALL_WMHvol_mm3 WMHs volume in right middle artery lateral lenticulostriate territory, lAAH_WMHnoc_total the total number of WMHs clusters in left anterior artery hemisphere territory, Rfrontal_WMHnoc_punctuate the number of WMHs clusters with punctuate size in right frontal lobe, rPAH_WMHnoc_focal the number of WMHs clusters with punctuate size in right posterior artery hemisphere territory, wholeBrain_WMHnoc_medium the number of WMHs clusters with medium size in whole brain, lAAH_WMHnoc_medium the number of WMHs clusters with medium size in left anterior artery hemisphere territory, lMALL_WMHnoc_confluent the number of WMHs clusters with confluent size in left middle artery lateral lenticulostriate territory, TOAST trial of org 10,172 in acute stroke treatment stroke, Discharge_Anticoagulants implementation of post-discharge anticoagulant therapy for secondary prevention

Clinical characteristics-based model (Model 2): this model indicated that older age (HR = 1.04, 95% CI 1.01–1.06) and smoking (HR = 1.74, 95% CI 1.03–2.93) were associated with higher recurrence risk. Stroke subtypes other than large artery atherosclerosis (LAA) and cardioembolism (CE) showed a decreased risk (HR = 0.74, 95% CI 0.58–0.95) (all p < 0.05; Table 2).

Subjective Fazekas scale-based model (Model 3): higher total Fazekas scale scores were linked to an increased risk of recurrence (HR = 1.33, 95% CI 1.13–1.57) (all p < 0.05; Table 2).

Quantitative WMHs combined with clinical characteristics model (Model 4): LASSO multivariate Cox regression analysis showed significant associations with three WMHs volume variables, five NoC variables, and stroke subtypes in predicting recurrence (all p < 0.05; Table 2).

Fazekas scale combined with clinical characteristics-based model (Model 5): higher Fazekas scale scores (HR = 1.34, 95% CI 1.13–1.58) and post-discharge use of anticoagulants (HR = 2.42, 95% CI 1.03–5.64) were associated with a higher risk of recurrence. In contrast, other and underdetermined stroke subtypes showed a lower risk compared to LAA and CE (HR = 0.72, 95% CI 0.57–0.93) (all p < 0.05; Table 2).

Comparison of Different Prediction Models

Survival curves for the various prediction models are depicted in Fig. 3, demonstrating significant p values for both the training and testing sets. The C-index for the five prediction models is detailed in Table 3. The models based on quantitative WMHs and those combining quantitative WMHs with clinical characteristics exhibited superior discrimination capability compared to those based solely on the Fazekas scale or clinical characteristics. The nomogram integrating quantitative WMHs with clinical characteristics is shown in Fig. 4. The DCA indicated that this combined nomogram provided a higher overall net benefit compared to the other models, across a broad range of reasonable threshold probabilities (Fig. 5).Fig. 3 Kaplan–Meier survival curves of patients with ischemic stroke. Survival curves for A quantitative WMHs-based model (Model 1), B clinical characteristics-based model (Model 2); C subjective Fazekas scale-based model (Model 3): D quantitative WMHs combined with clinical characteristics model (Model 4); E Fazekas scale combined with clinical characteristics-based model (Model 5)

Table 3 Performance of prediction model

Model	Training cohort	Validation cohort	
C-index	95%CI	P value	C-index	95%CI	P value	
Model 1	0.81	0.76, 0.86	< 0.001	0.71	0.61, 0.81	0.006	
Model 2	0.64	0.57, 0.71	0.001	0.68	0.54, 0.74	0.010	
Model 3	0.62	0.55, 0.70	0.001	0.65	0.55, 0.75	0.007	
Model 4	0.83	0.78, 0.88	< 0.001	0.74	0.64, 0.83	0.002	
Model 5	0.67	0.59, 0.73	< 0.001	0.70	0.60, 0.80	0.006	
Model 1 quantitative WMHs-based model, Model 2 clinical characteristics-based model, Model 3 subjective Fazekas scale-based model, Model 4 quantitative WMHs combined with clinical characteristics model, Model 5 Fazekas scale combined with clinical characteristics-based model

Fig. 4 The nomogram based on white matter hyperintensities and clinical characteristics. A Fazekas scale-based nomogram, B quantitative WMHs-based nomogram. The two nomograms combining Fazekas, quantitative white matter hyperintensities and clinical characteristics were developed in the training cohort. The points for each were determined by drawing a vertical line to the points’ axis. The final total was then located on the total point axis, which indicates the probability of ischemic stroke recurrence

Fig. 5 The decision curve of the five models. The net benefit was calculated by summing the benefits (true-positive results) and subtracting the harms (false-positive results). The quantitative WMHs combined with clinical characteristics model (Model 4) showed the highest net benefit compared with other models

Discussion

The current study has explored different spatial distributions of the volume and NoC of WMHs to estimate ischemic stroke recurrence. Our findings highlight that the quantitative measurements of WMHs, encompassing both volumes and NoC, serve as predictive markers for ischemic stroke recurrence within 1 year post-event. Notably, a nomogram based on quantitative WMHs assessments demonstrated superior predictive performance compared to the traditional subjective Fazekas scale.

Demographic and clinical variables, such as age, stroke subtype, and post-discharge antiplatelet therapy, were also significantly linked to stroke recurrence. Advanced age and the absence of post-discharge antiplatelet therapy were hypothesized to correlate positively with increased recurrence risk, a finding supported by existing literature that suggests a relationship between reduced antiplatelet use at discharge and higher recurrence rates [36]. Our results also confirm that LAA and cardioembolic strokes exhibit a higher recurrence rate compared to other subtypes [29].

Numerous studies have consistently demonstrated a significant association between WMHs and cognitive impairment, dementia, stroke occurrence, and post-stroke functional outcome, as well as motor symptoms in Parkinson's disease [6, 9, 10, 26, 37–42]. However, few studies have specifically explored the relationship between WMHs and recurrent stroke risk. A substantial recent study involving over 7,000 patients established WMHs as an independent risk factor for stroke recurrence [29], aligning with our findings. A higher volume burden could represent more severe brain tissue damage. In our analysis, not only was WMH volume considered, but the spatial distribution and NoC were also evaluated using an automated extraction method (UBO detector) [33]. The presumed vascular origin and heterogeneity of WMHs corresponded to different underlying brain parenchymal changes, likely indicative of varying numbers and sizes of WMH clusters [5, 33]. Our analysis showed that a greater NoC in the deep white matter and brainstem was strongly associated with recurrent stroke risk. Prior research has demonstrated that a more intricate shape of WMHs may indicate more profound underlying cerebral alterations [7]. Histopathological investigations could offer evidence supporting ischemic damage, myelin loss, and incomplete parenchymal destruction in WMHs with more irregular shapes [5]. Accordingly, a high total NoC might indicate severe underlying vascular impairments, increasing the risk of stroke recurrence. The amount of D-WMHs, typically linked to ischemic origins [15], reflects the degree of ischemic damage, with extensive fiber loss and arteriolosclerosis marking the most severe cases. Moreover, recent evidence underscores a significant correlation between WMH presence in the brainstem and adverse functional outcomes, including a heightened stroke risk [12]. The etiology of WMHs remains elusive, but it is hypothesized that chronic white matter damage impairs brain plasticity and its capacity to compensate for ischemic insults, partly due to disrupted organization of white matter fibers and neuronal network integrity [12]. Conversely, an increased total NoC in specific regions like the right anterior artery callosal, focal clusters in the left parietal lobe, and left middle artery hemisphere showed an inverse association with recurrent stroke risk, a relationship that remains not well understood. Despite the higher NoC in these smaller regions, the overall WMHs burden was presumed low, suggesting a limited impact on stroke outcomes.

This study developed two nomograms: one based on quantitative WMHs and the other on the widely used Fazekas scale, a validated visual rating system compatible with volumetric measurements and valued for its cost-effectiveness and reliability [8, 32]. Comparative analysis demonstrated that the quantitative WMHs nomogram outperformed the Fazekas scale nomogram in predicting ischemic stroke recurrence, evidenced by a higher C-index of 0.732 versus 0.634 in the testing set. While the Fazekas scale provides valuable semi-quantitative information about periventricular and deep white matter WMHs, the integration of cluster-based WMHs extraction, which includes topographic and vascular territory data, offers significant insights into advanced neuroimaging techniques, enhancing our understanding of the pathology.

Our study possesses several notable strengths. Primarily, it leverages a large sample size of 1,061 patients with ischemic stroke, facilitating a robust analysis of WMHs' impact on stroke recurrence. Additionally, the use of an automated UBO detector ensures precise and efficient segmentation of WMHs, boosting the analytical accuracy. By incorporating multiple quantitative variables of WMHs, such as volume and NoC with diverse sizes and distributions, we extend clinical and research applications.

However, the study also faces limitations. It is retrospective and single-centered; although we employed training and testing sets for internal validation, external validation with data from various stroke centers and MRI scanners remains essential. Moreover, the exclusion of patients with hemorrhagic stroke from follow-up confines our findings to recurrent ischemic stroke risks. Future longitudinal studies are planned to overcome this limitation. Additionally, data variability due to the use of two different MRI scanners and varying scan parameters, along with a standard slice thickness of 5 mm on FLAIR and T1-WI, could impact segmentation precision. Efforts to normalize T1-WI and co-register FLAIR sequences were made, but further verification using 3D T1-WI and thinner FLAIR sequences is required.

Conclusion

Our findings confirm that a higher WMH burden significantly correlates with increased ischemic stroke recurrence risk. The presence of WMHs, particularly in deep white matter or the brainstem, is closely linked to the heightened risk of recurrence. The quantitative analysis of WMHs introduces a promising neuro-imaging tool for predicting stroke recurrence. The observed spatial heterogeneity of WMHs suggests distinct underlying pathologies, warranting further prospective investigation to clarify these mechanisms.

Supplementary Information

Below is the link to the electronic supplementary material.Supplementary file1 (PDF 308 KB)

Acknowledgements

The authors express their gratitude to the doctors Yuzeng Liu and Yi Jin (Department of Radiology, Minhang Hospital, Fudan University) for searching data. We also thank Guoqing Wu (Fudan University) for invaluable help in manuscript preparation.

Medical Writing

The authors didn’t receive any medical writing or editorial assistance for this article.

Author Contribution

Conceptualization: Hao Wang; Methodology: Yi Sun; Formal analysis: Bin Song, Wenping Xia; Investigation: Yi Sun, Ran Wei, Zedong Dai, Xilin Sun, Jie Zhu; Project administration: Bin Song; Writing—original draft preparation: Yi Sun, Wenping Xia, Hao Wang; Writing—review and editing: Bin Song, Hao Wang; Funding acquisition: Yi Sun, Hao Wang.

Funding

This study and the journal’s Rapid Service Fee was funded by the Natural Science Foundation of Minhang Hospital, Fudan University (2022MHBJ04 and 2022MHPY04), Science and Health Commission of Minhang District, Shanghai (mwyjyx17).

Data Availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Conflict of Interest

All authors, including Yi Sun, Wenping Xia, Ran Wei, Jie Zhu, Bin Song, and Hao Wang, Zedong Dai and Xilin Sun have declared no potential competing interests. The manuscript has been approved for submission by all authors. No commercial or financial conflicts of interest were identified in relation to this research.

Ethical Approval

This current retrospective study was approved by the Institutional Ethics Committee of Minhang Hospital, Fudan University, with a waiver for informed consent granted (approval number: 2023-022-01 K). The study was performed in accordance with the 1964 Declaration of Helsinki and its later amendments.

Thanking Patients

We thank the participants of the study.

Yi Sun and Wenping Xia contributed equally to this work and share the first authorship.
==== Refs
References

1. Lee Y Ko J Choi YE Areas of white matter hyperintensities and motor symptoms of Parkinson disease Neurology 2020 95 e291 e298 10.1212/WNL.0000000000009890 32576636
Lee Y, Ko J, Choi YE, et al. Areas of white matter hyperintensities and motor symptoms of Parkinson disease. Neurology. 2020;95:e291–8.32576636 10.1212/WNL.0000000000009890
2. Uretsky M Bouix S Killiany RJ Association between antemortem flair white matter hyperintensities and neuropathology in brain donors exposed to repetitive head impacts Neurology 2022 98 e27 e39 10.1212/WNL.0000000000013012 34819338
Uretsky M, Bouix S, Killiany RJ, et al. Association between antemortem flair white matter hyperintensities and neuropathology in brain donors exposed to repetitive head impacts. Neurology. 2022;98:e27–39.34819338 10.1212/WNL.0000000000013012
3. Zhao L Wong A Luo Y The additional contribution of white matter hyperintensity location to post-stroke cognitive impairment: Insights from a multiple-lesion symptom mapping study Front Neurosci 2018 12 290 10.3389/fnins.2018.00290 29765301
Zhao L, Wong A, Luo Y, et al. The additional contribution of white matter hyperintensity location to post-stroke cognitive impairment: Insights from a multiple-lesion symptom mapping study. Front Neurosci. 2018;12:290.29765301 10.3389/fnins.2018.00290
4. Ghaznawi R Geerlings MI Jaarsma-Coes MG The association between lacunes and white matter hyperintensity features on MRI: the SMART-MR study J Cereb Blood Flow Metab 2019 39 2486 2496 10.1177/0271678X18800463 30204039
Ghaznawi R, Geerlings MI, Jaarsma-Coes MG, et al. The association between lacunes and white matter hyperintensity features on MRI: the SMART-MR study. J Cereb Blood Flow Metab. 2019;39:2486–96.30204039 10.1177/0271678X18800463
5. Ghaznawi R Geerlings MI Jaarsma-Coes M Hendrikse J de Bresser J Group UC-SS Association of white matter hyperintensity markers on mri and long-term risk of mortality and ischemic stroke: the SMART-MR Study Neurology 2021 96 e2172 e2183 10.1212/WNL.0000000000011827 33727406
Ghaznawi R, Geerlings MI, Jaarsma-Coes M, Hendrikse J, de Bresser J, Group UC-SS. Association of white matter hyperintensity markers on mri and long-term risk of mortality and ischemic stroke: the SMART-MR Study. Neurology. 2021;96:e2172–83.33727406 10.1212/WNL.0000000000011827
6. Derraz I Abdelrady M Ahmed R Impact of white matter hyperintensity burden on outcome in large-vessel occlusion stroke Radiology 2022 304 210419 10.1148/radiol.210419
Derraz I, Abdelrady M, Ahmed R, et al. Impact of white matter hyperintensity burden on outcome in large-vessel occlusion stroke. Radiology. 2022;304:210419.10.1148/radiol.210419
7. Inglese F Jaarsma-Coes MG Steup-Beekman GM Neuropsychiatric systemic lupus erythematosus is associated with a distinct type and shape of cerebral white matter hyperintensities Rheumatology (Oxford) 2022 61 2663 2671 10.1093/rheumatology/keab823 34730801
Inglese F, Jaarsma-Coes MG, Steup-Beekman GM, et al. Neuropsychiatric systemic lupus erythematosus is associated with a distinct type and shape of cerebral white matter hyperintensities. Rheumatology (Oxford). 2022;61:2663–71.34730801 10.1093/rheumatology/keab823
8. Hasan TF Barrett KM Brott TG Severity of white matter hyperintensities and effects on all-cause mortality in the mayo clinic florida familial cerebrovascular diseases registry Mayo Clin Proc 2019 94 408 416 10.1016/j.mayocp.2018.10.024 30832790
Hasan TF, Barrett KM, Brott TG, et al. Severity of white matter hyperintensities and effects on all-cause mortality in the mayo clinic florida familial cerebrovascular diseases registry. Mayo Clin Proc. 2019;94:408–16.30832790 10.1016/j.mayocp.2018.10.024
9. Lampe L Kharabian-Masouleh S Kynast J Lesion location matters: the relationships between white matter hyperintensities on cognition in the healthy elderly J Cereb Blood Flow Metab 2019 39 36 43 10.1177/0271678X17740501 29106319
Lampe L, Kharabian-Masouleh S, Kynast J, et al. Lesion location matters: the relationships between white matter hyperintensities on cognition in the healthy elderly. J Cereb Blood Flow Metab. 2019;39:36–43.29106319 10.1177/0271678X17740501
10. Melazzini L Mackay CE Bordin V White matter hyperintensities classified according to intensity and spatial location reveal specific associations with cognitive performance NeuroImage Clinical 2021 30 102616 10.1016/j.nicl.2021.102616 33743476
Melazzini L, Mackay CE, Bordin V, et al. White matter hyperintensities classified according to intensity and spatial location reveal specific associations with cognitive performance. NeuroImage Clinical. 2021;30: 102616.33743476 10.1016/j.nicl.2021.102616
11. Wiegertjes K Dinsmore L Drever J Diffusion-weighted imaging lesions and risk of recurrent stroke after intracerebral haemorrhage J Neurol Neurosurg Psychiatry 2021 92 950 955 10.1136/jnnp-2021-326116 34103345
Wiegertjes K, Dinsmore L, Drever J, et al. Diffusion-weighted imaging lesions and risk of recurrent stroke after intracerebral haemorrhage. J Neurol Neurosurg Psychiatry. 2021;92:950–5.34103345 10.1136/jnnp-2021-326116
12. Giralt-Steinhauer E Medrano S Soriano-Tarraga C Brainstem leukoaraiosis independently predicts poor outcome after ischemic stroke Eur J Neurol 2018 25 1086 1092 10.1111/ene.13659 29660221
Giralt-Steinhauer E, Medrano S, Soriano-Tarraga C, et al. Brainstem leukoaraiosis independently predicts poor outcome after ischemic stroke. Eur J Neurol. 2018;25:1086–92.29660221 10.1111/ene.13659
13. Wardlaw JM Smith EE Biessels GJ Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration Lancet Neurol 2013 12 822 838 10.1016/S1474-4422(13)70124-8 23867200
Wardlaw JM, Smith EE, Biessels GJ, et al. Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration. Lancet Neurol. 2013;12:822–38.23867200 10.1016/S1474-4422(13)70124-8
14. Bauer CE Zachariou V Seago E Gold BT White matter hyperintensity volume and location: associations with WM microstructure, brain iron, and cerebral perfusion Front Aging Neurosci 2021 13 617947 10.3389/fnagi.2021.617947 34290597
Bauer CE, Zachariou V, Seago E, Gold BT. White matter hyperintensity volume and location: associations with WM microstructure, brain iron, and cerebral perfusion. Front Aging Neurosci. 2021;13: 617947.34290597 10.3389/fnagi.2021.617947
15. Veldsman M Kindalova P Husain M Kosmidis I Nichols TE Spatial distribution and cognitive impact of cerebrovascular risk-related white matter hyperintensities NeuroImage Clin 2020 28 102405 10.1016/j.nicl.2020.102405 32971464
Veldsman M, Kindalova P, Husain M, Kosmidis I, Nichols TE. Spatial distribution and cognitive impact of cerebrovascular risk-related white matter hyperintensities. NeuroImage Clin. 2020;28: 102405.32971464 10.1016/j.nicl.2020.102405
16. de Bresser J Kuijf HJ Zaanen K White matter hyperintensity shape and location feature analysis on brain MRI; proof of principle study in patients with diabetes Sci Rep 2018 8 1893 10.1038/s41598-018-20084-y 29382936
de Bresser J, Kuijf HJ, Zaanen K, et al. White matter hyperintensity shape and location feature analysis on brain MRI; proof of principle study in patients with diabetes. Sci Rep. 2018;8:1893.29382936 10.1038/s41598-018-20084-y
17. Collaborators GBDS Global, regional, and national burden of stroke and its risk factors, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019 Lancet Neurol 2021 20 795 820 10.1016/S1474-4422(21)00252-0 34487721
Collaborators GBDS. Global, regional, and national burden of stroke and its risk factors, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Neurol. 2021;20:795–820.34487721 10.1016/S1474-4422(21)00252-0
18. Saini V Guada L Yavagal DR Global epidemiology of stroke and access to acute ischemic stroke interventions Neurology 2021 97 S6 S16 10.1212/WNL.0000000000012781 34785599
Saini V, Guada L, Yavagal DR. Global epidemiology of stroke and access to acute ischemic stroke interventions. Neurology. 2021;97:S6–16.34785599 10.1212/WNL.0000000000012781
19. Skajaa N Adelborg K Horvath-Puho E Risks of stroke recurrence and mortality after first and recurrent strokes in Denmark: a nationwide registry study Neurology 2022 98 e329 e342 10.1212/WNL.0000000000013118 34845054
Skajaa N, Adelborg K, Horvath-Puho E, et al. Risks of stroke recurrence and mortality after first and recurrent strokes in Denmark: a nationwide registry study. Neurology. 2022;98:e329–42.34845054 10.1212/WNL.0000000000013118
20. Skoog I Madsen TE Risk of recurrent stroke: the critical need for continued efforts in secondary prevention Neurology 2022 98 133 134 10.1212/WNL.0000000000013116 34845053
Skoog I, Madsen TE. Risk of recurrent stroke: the critical need for continued efforts in secondary prevention. Neurology. 2022;98:133–4.34845053 10.1212/WNL.0000000000013116
21. Mohan KM Wolfe CD Rudd AG Heuschmann PU Kolominsky-Rabas PL Grieve AP Risk and cumulative risk of stroke recurrence: a systematic review and meta-analysis Stroke 2011 42 1489 1494 10.1161/STROKEAHA.110.602615 21454819
Mohan KM, Wolfe CD, Rudd AG, Heuschmann PU, Kolominsky-Rabas PL, Grieve AP. Risk and cumulative risk of stroke recurrence: a systematic review and meta-analysis. Stroke. 2011;42:1489–94.21454819 10.1161/STROKEAHA.110.602615
22. Boulouis G Bricout N Benhassen W White matter hyperintensity burden in patients with ischemic stroke treated with thrombectomy Neurology 2019 93 e1498 e1506 10.1212/WNL.0000000000008317 31519778
Boulouis G, Bricout N, Benhassen W, et al. White matter hyperintensity burden in patients with ischemic stroke treated with thrombectomy. Neurology. 2019;93:e1498–506.31519778 10.1212/WNL.0000000000008317
23. Etherton MR Wu O Giese AK Rost NS Normal-appearing white matter microstructural injury is associated with white matter hyperintensity burden in acute ischemic stroke Int J Stroke 2021 16 184 191 10.1177/1747493019895707 31847795
Etherton MR, Wu O, Giese AK, Rost NS. Normal-appearing white matter microstructural injury is associated with white matter hyperintensity burden in acute ischemic stroke. Int J Stroke. 2021;16:184–91.31847795 10.1177/1747493019895707
24. Zerna C Yu AYX Hong ZM White matter hyperintensity volume influences symptoms in patients presenting with minor neurological deficits Stroke 2020 51 409 415 10.1161/STROKEAHA.119.027213 31795896
Zerna C, Yu AYX, Hong ZM, et al. White matter hyperintensity volume influences symptoms in patients presenting with minor neurological deficits. Stroke. 2020;51:409–15.31795896 10.1161/STROKEAHA.119.027213
25. Griessenauer CJ McPherson D Berger A Effects of white matter hyperintensities on 90-day functional outcome after large vessel and non-large vessel stroke Cerebrovasc Dis 2020 49 419 426 10.1159/000509071 32694259
Griessenauer CJ, McPherson D, Berger A, et al. Effects of white matter hyperintensities on 90-day functional outcome after large vessel and non-large vessel stroke. Cerebrovasc Dis. 2020;49:419–26.32694259 10.1159/000509071
26. Giese AK Schirmer MD Dalca AV White matter hyperintensity burden in acute stroke patients differs by ischemic stroke subtype Neurology 2020 95 e79 e88 10.1212/WNL.0000000000009728 32493718
Giese AK, Schirmer MD, Dalca AV, et al. White matter hyperintensity burden in acute stroke patients differs by ischemic stroke subtype. Neurology. 2020;95:e79–88.32493718 10.1212/WNL.0000000000009728
27. Ryu WS Woo SH Schellingerhout D Stroke outcomes are worse with larger leukoaraiosis volumes Brain 2017 140 158 170 10.1093/brain/aww259 28008000
Ryu WS, Woo SH, Schellingerhout D, et al. Stroke outcomes are worse with larger leukoaraiosis volumes. Brain. 2017;140:158–70.28008000 10.1093/brain/aww259
28. Zerna C Yu AYX Modi J Association of white matter hyperintensities with short-term outcomes in patients with minor cerebrovascular events Stroke 2018 49 919 923 10.1161/STROKEAHA.117.017429 29540612
Zerna C, Yu AYX, Modi J, et al. Association of white matter hyperintensities with short-term outcomes in patients with minor cerebrovascular events. Stroke. 2018;49:919–23.29540612 10.1161/STROKEAHA.117.017429
29. Ryu WS Schellingerhout D Hong KS White matter hyperintensity load on stroke recurrence and mortality at 1 year after ischemic stroke Neurology 2019 93 e578 e589 10.1212/WNL.0000000000007896 31308151
Ryu WS, Schellingerhout D, Hong KS, et al. White matter hyperintensity load on stroke recurrence and mortality at 1 year after ischemic stroke. Neurology. 2019;93:e578–89.31308151 10.1212/WNL.0000000000007896
30. Brott T Adams HJ Olinger C Measurements of acute cerebral infarction: a clinical examination scale Stroke 1989 20 864 870 10.1161/01.STR.20.7.864 2749846
Brott T, Adams HJ, Olinger C, et al. Measurements of acute cerebral infarction: a clinical examination scale. Stroke. 1989;20:864–70.2749846 10.1161/01.STR.20.7.864
31. Farrell B Godwin J Richards S Warlow C The United Kingdom transient ischaemic attack (UK-TIA) aspirin trial: final results J Neurol Neurosurg Psychiatry 1991 54 1044 1054 10.1136/jnnp.54.12.1044 1783914
Farrell B, Godwin J, Richards S, Warlow C. The United Kingdom transient ischaemic attack (UK-TIA) aspirin trial: final results. J Neurol Neurosurg Psychiatry. 1991;54:1044–54.1783914 10.1136/jnnp.54.12.1044
32. Fazekas F Chawluk JB Alavi A Hurtig HI Zimmerman RA MR signal abnormalities at 1.5 T in Alzheimer's dementia and normal aging AJR Am J Roentgenol 1987 149 351 356 10.2214/ajr.149.2.351 3496763
Fazekas F, Chawluk JB, Alavi A, Hurtig HI, Zimmerman RA. MR signal abnormalities at 1.5 T in Alzheimer’s dementia and normal aging. AJR Am J Roentgenol. 1987;149:351–6.3496763 10.2214/ajr.149.2.351
33. Jiang J Liu T Zhu W UBO detector—a cluster-based, fully automated pipeline for extracting white matter hyperintensities Neuroimage 2018 174 539 549 10.1016/j.neuroimage.2018.03.050 29578029
Jiang J, Liu T, Zhu W, et al. UBO detector—a cluster-based, fully automated pipeline for extracting white matter hyperintensities. Neuroimage. 2018;174:539–49.29578029 10.1016/j.neuroimage.2018.03.050
34. Wen W Sachdev P The topography of white matter hyperintensities on brain MRI in healthy 60- to 64-year-old individuals Neuroimage 2004 22 144 154 10.1016/j.neuroimage.2003.12.027 15110004
Wen W, Sachdev P. The topography of white matter hyperintensities on brain MRI in healthy 60- to 64-year-old individuals. Neuroimage. 2004;22:144–54.15110004 10.1016/j.neuroimage.2003.12.027
35. Huang Y Liu Z He L Radiomics signature: a potential biomarker for the prediction of disease-free survival in early-stage (I or II) non-small cell lung cancer Radiology 2016 281 947 957 10.1148/radiol.2016152234 27347764
Huang Y, Liu Z, He L, et al. Radiomics signature: a potential biomarker for the prediction of disease-free survival in early-stage (I or II) non-small cell lung cancer. Radiology. 2016;281:947–57.27347764 10.1148/radiol.2016152234
36. Jing J Suo Y Wang A Imaging parameters predict recurrence after transient ischemic attack or minor stroke stratified by ABCD(2) score Stroke 2021 52 2007 2015 10.1161/STROKEAHA.120.032424 33947206
Jing J, Suo Y, Wang A, et al. Imaging parameters predict recurrence after transient ischemic attack or minor stroke stratified by ABCD(2) score. Stroke. 2021;52:2007–15.33947206 10.1161/STROKEAHA.120.032424
37. Habes M Sotiras A Erus G White matter lesions: Spatial heterogeneity, links to risk factors, cognition, genetics, and atrophy Neurology 2018 91 e964 e975 10.1212/WNL.0000000000006116 30076276
Habes M, Sotiras A, Erus G, et al. White matter lesions: Spatial heterogeneity, links to risk factors, cognition, genetics, and atrophy. Neurology. 2018;91:e964–75.30076276 10.1212/WNL.0000000000006116
38. Etherton MR Wu O Giese AK White matter integrity and early outcomes after acute ischemic stroke Transl Stroke Res 2019 10 630 638 10.1007/s12975-019-0689-4 30693424
Etherton MR, Wu O, Giese AK, et al. White matter integrity and early outcomes after acute ischemic stroke. Transl Stroke Res. 2019;10:630–8.30693424 10.1007/s12975-019-0689-4
39. Jeong SH Lee HS Jung JH White matter hyperintensities, dopamine loss, and motor deficits in de novo Parkinson’s disease Mov Disord 2021 36 1411 1419 10.1002/mds.28510 33513293
Jeong SH, Lee HS, Jung JH, et al. White matter hyperintensities, dopamine loss, and motor deficits in de novo Parkinson’s disease. Mov Disord. 2021;36:1411–9.33513293 10.1002/mds.28510
40. Weaver NA Doeven T Barkhof F Cerebral amyloid burden is associated with white matter hyperintensity location in specific posterior white matter regions Neurobiol Aging 2019 84 225 234 10.1016/j.neurobiolaging.2019.08.001 31500909
Weaver NA, Doeven T, Barkhof F, et al. Cerebral amyloid burden is associated with white matter hyperintensity location in specific posterior white matter regions. Neurobiol Aging. 2019;84:225–34.31500909 10.1016/j.neurobiolaging.2019.08.001
41. Gwo CY Zhu DC Zhang R Brain white matter hyperintensity lesion characterization in t2 fluid-attenuated inversion recovery magnetic resonance images: Shape, texture, and potential growth Front Neurosci 2019 13 353 10.3389/fnins.2019.00353 31057353
Gwo CY, Zhu DC, Zhang R. Brain white matter hyperintensity lesion characterization in t2 fluid-attenuated inversion recovery magnetic resonance images: Shape, texture, and potential growth. Front Neurosci. 2019;13:353.31057353 10.3389/fnins.2019.00353
42. Graff-Radford J Arenaza-Urquijo EM Knopman DS White matter hyperintensities: relationship to amyloid and tau burden Brain 2019 142 2483 2491 10.1093/brain/awz162 31199475
Graff-Radford J, Arenaza-Urquijo EM, Knopman DS, et al. White matter hyperintensities: relationship to amyloid and tau burden. Brain. 2019;142:2483–91.31199475 10.1093/brain/awz162
