Network impact score is an independent predictor of post-stroke cognitive impairment: A multicenter cohort study in 2341 patients with acute ischemic stroke

BACKGROUND

Post-stroke cognitive impairment (PSCI) is a common consequence of stroke. Accurate prediction of PSCI risk is challenging. The recently developed network impact score, which integrates information on infarct location and size with brain network topology, may improve PSCI risk prediction.

AIMS

To determine if the network impact score is an independent predictor of PSCI, and of cognitive recovery or decline.

METHODS

We pooled data from patients with acute ischemic stroke from 12 cohorts through the Meta VCI Map consortium. PSCI was defined as impairment in ≥ 1 cognitive domain on neuropsychological examination, or abnormal Montreal Cognitive Assessment. Cognitive recovery was defined as conversion from PSCI < 3 months post-stroke to no PSCI at follow-up, and cognitive decline as conversion from no PSCI to PSCI. The network impact score was related to serial measures of PSCI using Generalized Estimating Equations (GEE) models, and to PSCI stratified according to post-stroke interval (<3, 3-12, 12-24, >24 months) and cognitive recovery or decline using logistic regression. Models were adjusted for age, sex, education, prior stroke, infarct volume, and study site.

RESULTS

We included 2341 patients with 4657 cognitive assessments. PSCI was present in 398/844 patients (47%) <3 months, 709/1640 (43%) at 3-12 months, 243/853 (28%) at 12-24 months, and 208/522 (40%) >24 months. Cognitive recovery occurred in 64/181 (35%) patients and cognitive decline in 26/287 (9%). The network impact score predicted PSCI in the univariable (OR 1.50, 95%CI 1.34-1.68) and multivariable (OR 1.27, 95%CI 1.10-1.46) GEE model, with similar ORs in the logistic regression models for specified post-stroke intervals. The network impact score was not associated with cognitive recovery or decline.

CONCLUSIONS

The network impact score is an independent predictor of PSCI. As such, the network impact score may contribute to a more precise and individualized cognitive prognostication in patients with ischemic stroke. Future studies should address if multimodal prediction models, combining the network impact score with demographics, clinical characteristics and other advanced brain imaging biomarkers, will provide accurate individualized prediction of PSCI. A tool for calculating the network impact score is freely available at https://metavcimap.org/features/software-tools/lsm-viewer/.

Strategic infarct locations for post-stroke cognitive impairment: a pooled analysis of individual patient data from 12 acute ischaemic stroke cohorts

BACKGROUND

Post-stroke cognitive impairment (PSCI) occurs in approximately half of people in the first year after stroke. Infarct location is a potential determinant of PSCI, but a comprehensive map of strategic infarct locations predictive of PSCI is unavailable. We aimed to identify infarct locations most strongly predictive of PSCI after acute ischaemic stroke and use this information to develop a prediction model.

METHODS

In this large-scale multicohort lesion-symptom mapping study, we pooled and harmonised individual patient data from 12 cohorts through the Meta-analyses on Strategic Lesion Locations for Vascular Cognitive Impairment using Lesion-Symptom Mapping (Meta VCI Map) consortium. The identified cohorts (as of Jan 1, 2019) comprised patients with acute symptomatic infarcts on CT or MRI (with available infarct segmentations) and a cognitive assessment up to 15 months after acute ischaemic stroke onset. PSCI was defined as performance lower than the fifth percentile of local normative data, on at least one cognitive domain on a multidomain neuropsychological assessment or on the Montreal Cognitive Assessment. Voxel-based lesion-symptom mapping (VLSM) was used to calculate voxel-wise odds ratios (ORs) for PSCI that were mapped onto a three-dimensional brain template to visualise PSCI risk per location. For the prediction model of PSCI risk, a location impact score on a 5-point scale was derived from the VLSM results on the basis of the mean voxel-wise coefficient (ln[OR]) within each patient's infarct. We did combined internal-external validation by leave-one-cohort-out cross-validation for all 12 cohorts using logistic regression. Predictive performance of a univariable model with only the location impact score was compared with a multivariable model with addition of other clinical PSCI predictors (age, sex, education, time interval between stroke onset and cognitive assessment, history of stroke, and total infarct volume). Testing of visual ratings was done by three clinicians, and accuracy, inter-rater reliability, and intra-rater reliability were assessed with Cohen's weighted kappa.

FINDINGS

In our sample of 2950 patients (mean age 66·8 years [SD 11·6]; 1157 [39·2%] women), 1286 (43·6%) had PSCI. We achieved high lesion coverage of the brain in our analyses (86·9%). Infarcts in the left frontotemporal lobes, left thalamus, and right parietal lobe were strongly associated with PSCI (after false discovery rate correction, q<0·01; voxel-wise ORs >20). On cross-validation, the location impact score showed good correspondence, based on visual assessment of goodness of fit, between predicted and observed risk of PSCI across cohorts after adjusting for cohort-specific PSCI occurrence. Cross-validations showed that the location impact score by itself had similar performance to the combined model with other PSCI predictors, while allowing for easy visual assessment. Therefore the univariable model with only the location impact score was selected as the final model. Correspondence between visual ratings and actual location impact score (Cohen's weighted kappa: range 0·88-0·92), inter-rater agreement (0·85-0·87), and intra-rater agreement (for a single rater, 0·95) were all high.

INTERPRETATION

To the best of our knowledge, this study provides the first comprehensive map of strategic infarct locations associated with risk of PSCI. A location impact score was derived from this map that robustly predicted PSCI across cohorts. Furthermore, we developed a quick and reliable visual rating scale that might in the future be applied by clinicians to identify individual patients at risk of PSCI.

FUNDING

The Netherlands Organisation for Health Research and Development.

Automatic spatial estimation of white matter hyperintensities evolution in brain MRI using disease evolution predictor deep neural networks

Previous studies have indicated that white matter hyperintensities (WMH), the main radiological feature of small vessel disease, may evolve (i.e., shrink, grow) or stay stable over a period of time. Predicting these changes are challenging because it involves some unknown clinical risk factors that leads to a non-deterministic prediction task. In this study, we propose a deep learning model to predict the evolution of WMH from baseline to follow-up (i.e., 1-year later), namely "Disease Evolution Predictor" (DEP) model, which can be adjusted to become a non-deterministic model. The DEP model receives a baseline image as input and produces a map called "Disease Evolution Map" (DEM), which represents the evolution of WMH from baseline to follow-up. Two DEP models are proposed, namely DEP-UResNet and DEP-GAN, which are representatives of the supervised (i.e., need expert-generated manual labels to generate the output) and unsupervised (i.e., do not require manual labels produced by experts) deep learning algorithms respectively. To simulate the non-deterministic and unknown parameters involved in WMH evolution, we modulate a Gaussian noise array to the DEP model as auxiliary input. This forces the DEP model to imitate a wider spectrum of alternatives in the prediction results. The alternatives of using other types of auxiliary input instead, such as baseline WMH and stroke lesion loads are also proposed and tested. Based on our experiments, the fully supervised machine learning scheme DEP-UResNet regularly performed better than the DEP-GAN which works in principle without using any expert-generated label (i.e., unsupervised). However, a semi-supervised DEP-GAN model, which uses probability maps produced by a supervised segmentation method in the learning process, yielded similar performances to the DEP-UResNet and performed best in the clinical evaluation. Furthermore, an ablation study showed that an auxiliary input, especially the Gaussian noise, improved the performance of DEP models compared to DEP models that lacked the auxiliary input regardless of the model's architecture. To the best of our knowledge, this is the first extensive study on modelling WMH evolution using deep learning algorithms, which deals with the non-deterministic nature of WMH evolution.

Limited One-time Sampling Irregularity Map (LOTS-IM) for Automatic Unsupervised Assessment of White Matter Hyperintensities and Multiple Sclerosis Lesions in Structural Brain Magnetic Resonance Images

We present the application of limited one-time sampling irregularity map (LOTS-IM): a fully automatic unsupervised approach to extract brain tissue irregularities in magnetic resonance images (MRI), for quantitatively assessing white matter hyperintensities (WMH) of presumed vascular origin, and multiple sclerosis (MS) lesions and their progression. LOTS-IM generates an irregularity map (IM) that represents all voxels as irregularity values with respect to the ones considered "normal". Unlike probability values, IM represents both regular and irregular regions in the brain based on the original MRI's texture information. We evaluated and compared the use of IM for WMH and MS lesions segmentation on T2-FLAIR MRI with the state-of-the-art unsupervised lesions' segmentation method, Lesion Growth Algorithm from the public toolbox Lesion Segmentation Toolbox (LST-LGA), with several well established conventional supervised machine learning schemes and with state-of-the-art supervised deep learning methods for WMH segmentation. In our experiments, LOTS-IM outperformed unsupervised method LST-LGA on WMH segmentation, both in performance and processing speed, thanks to the limited one-time sampling scheme and its implementation on GPU. Our method also outperformed supervised conventional machine learning algorithms (i.e., support vector machine (SVM) and random forest (RF)) and deep learning algorithms (i.e., deep Boltzmann machine (DBM) and convolutional encoder network (CEN)), while yielding comparable results to the convolutional neural network schemes that rank top of the algorithms developed up to date for this purpose (i.e., UResNet and UNet). LOTS-IM also performed well on MS lesions segmentation, performing similar to LST-LGA. On the other hand, the high sensitivity of IM on depicting signal change deems suitable for assessing MS progression, although care must be taken with signal changes not reflective of a true pathology.

Exome Chip Analysis Identifies Low-Frequency and Rare Variants in MRPL38 for White Matter Hyperintensities on Brain Magnetic Resonance Imaging

Background and Purpose- White matter hyperintensities (WMH) on brain magnetic resonance imaging are typical signs of cerebral small vessel disease and may indicate various preclinical, age-related neurological disorders, such as stroke. Though WMH are highly heritable, known common variants explain a small proportion of the WMH variance. The contribution of low-frequency/rare coding variants to WMH burden has not been explored. Methods- In the discovery sample we recruited 20 719 stroke/dementia-free adults from 13 population-based cohort studies within the Cohorts for Heart and Aging Research in Genomic Epidemiology consortium, among which 17 790 were of European ancestry and 2929 of African ancestry. We genotyped these participants at ≈250 000 mostly exonic variants with Illumina HumanExome BeadChip arrays. We performed ethnicity-specific linear regression on rank-normalized WMH in each study separately, which were then combined in meta-analyses to test for association with single variants and genes aggregating the effects of putatively functional low-frequency/rare variants. We then sought replication of the top findings in 1192 adults (European ancestry) with whole exome/genome sequencing data from 2 independent studies. Results- At 17q25, we confirmed the association of multiple common variants in TRIM65, FBF1, and ACOX1 ( P<6×10^-7). We also identified a novel association with 2 low-frequency nonsynonymous variants in MRPL38 (lead, rs34136221; P_EA=4.5×10^-8) partially independent of known common signal ( P_{EA(conditional)}=1.4×10^-3). We further identified a locus at 2q33 containing common variants in NBEAL1, CARF, and WDR12 (lead, rs2351524; P_all=1.9×10^-10). Although our novel findings were not replicated because of limited power and possible differences in study design, meta-analysis of the discovery and replication samples yielded stronger association for the 2 low-frequency MRPL38 variants ( P_rs34136221=2.8×10^-8). Conclusions- Both common and low-frequency/rare functional variants influence WMH. Larger replication and experimental follow-up are essential to confirm our findings and uncover the biological causal mechanisms of age-related WMH.