Unsupervised machine learning model to predict cognitive impairment in subcortical ischemic vascular disease.

Objectives: Subcortical ischemic vascular disease (SIVD) is frequently associated with cognitive impairment. Rating scales to grade cerebrovascular disease are available, but their sensitivity to the clinical features of mild SIVD is unclear. The aim of the study is to devise and validate a computed tomography (CT)-based visual rating scale sensitive to SIVD in patients with mild cognitive deterioration. Methods: Subjects were 122 consecutive outpatients of a memory clinic (mean age ± SD 77 ± 8 years, 71% females, mean ± SD score of mental state exam 22 ± 3, 61% with clinical dementia rating of 0.5 and 39% of 1). Diffuse leukoaraiosis, fuzzy and patchy lesions, and lacunes were assessed on ordinary CT films and weights were computed based on clinical indicators of SIVD. A continuous score and an ordinal class (0–3) with higher values indicating higher vascular damage were derived. Results: Intraclass correlation coefficients for intra- and interrater reliability of the subcortical vascular score and class ranged from 0.84 to 0.88. Convergent validity versus the scale of Wahlund et al. was good (Spearman’s correlation coefficient between 0.53 and 0.73, p < 0.0005). Known-group validity was assessed versus clinical diagnoses of degenerative (mild cognitive impairment determined by Petersen et al. and NINCDS-ADRDA probable Alzheimer’s disease; n = 58), mixed (possible Alzheimer’s disease with cerebrovascular disease; n = 21), and vascular cognitive impairment (criteria of Erkinjuntti et al. for subcortical vascular dementia and cognitive impairment; n = 43). Patients with degenerative cognitive impairment were more often in the lower subcortical vascular classes (43% in class 0, 43% in class 1, 14% in class 2, and 0% in class 3), while patients with mixed (0, 24, 24, and 52%) and vascular cognitive impairment were in increasingly higher subcortical vascular classes (0, 12, 35, 53%; p < 0.00005). Criterion-related validity was assessed versus clinical indicators of cerebrovascular disease. Hypertension, platelet aggregation inhibitor use, balance, gait, and bradykinesia increased linearly with increasing subcortical vascular class (p for trend <0.02). Conclusions: This rating scale is valid and sensitive to capture different degrees of SIVD associated with mild cognitive deterioration.

Vascular cognitive impairment (VCI) describes all forms of cognitive impairment caused by any type of cerebrovascular disease. Early identification of VCI is quite difficult due to the lack of both...

ObjectiveEmerging evidence suggests that white matter (WM) disruption is associated with the incidence of subcortical vascular cognitive impairment (SVCI). However, our knowledge regarding this relationship in the early stage of SVCI is limited. We aimed to investigate the associations between WM disruptions and cognitive declines at the early stage of SVCI.MethodWe performed a case–control study, involving 22 cases and 19 controls. The cases were patients at the early stage of SVCI, which was defined as subcortical ischemic vascular disease with normal global cognitive measures (pre-SVCI). The controls were healthy people matched by age, sex, and education years. We assessed the differences in a battery of neuropsychological tests between the two groups, investigated the diffusion changes in 40 WM tracts among the participants via an atlas-based segmentation strategy, and compared the differences between the cases and controls by multiple linear regression analysis. We then evaluated the relationships between diffusion indices and cognitive assessment scores by Pearson’s correlation.ResultsThe pre-SVCI group exhibited significant differences in the Montreal cognitive assessment (MoCA), Rey–Osterrieth Complex Figure (R-O)-copy, and Trail Making Test (TMT)-B test compared with the controls. Compared with the controls, some long associative and projective bundles, such as the right anterior corona radiata (ACR), the right inferior fronto-occipital fasciculus (IFOF), and the left external capsule (EC), were extensively damaged in cases after Bonferroni correction (p < 0.05/40). Damages to specific fibers, such as the right ACR, IFOF, and posterior thalamic radiation (PTR), exhibited significant correlations with declines in MoCA, R-O delay, and the Mini-Mental State Examination (MMSE), respectively, after Bonferroni correction (p < 0.05/14).ConclusionLong WM tracts, especially those in the right hemisphere, were extensively damaged in the pre-SVCI patients and correlated with declines in executive functions and spatial processing. Patients of pre-SVCI are likely at an ultra-early stage of SVCI, and there is a very high risk of this condition becoming SVCI.

BackgroundTo date, the relationship between thyroid dysfunction and subcortical ischemic vascular disease (SIVD)‐induced cognitive impairments still remains elusive.MethodsCognitive performances were examined in 215 participants, including 54 healthy participants, 52 SIVD patients with no dementia (SIVDND), 55 patients with mild cognitive impairment (SVMCI), and 54 patients with vascular dementia (VD). Serum thyroid‐stimulating hormone (TSH), total triiodothyronine (TT3), free triiodothyronine (FT3), total thyroxine (TT4) and free thyroxine (FT4), thyroglobulin antibody (TGA), and antithyroid peroxidase antibody (TPO‐Abs) were quantified by radioimmunoassay or ELISA.ResultsA close correlation between thyroid status and cognitive dysfunction in SIVD was observed. Serum TT3 and FT3 levels decreased, whereas serum TSH level increased, with the decline in cognitive functions. Furthermore, TT3 levels showed a positive correlation, whereas TSH level showed a negative correlation, with the Mini‐Mental State Examination (MMSE) scores. Our results suggested that thyroid function was associated with cognitive impairments induced by SIVD. Also, thyroid function and thyroid hormone level could be a risk factor in the development of SIVD. Serum TT3 and TSH levels might also be used as biomarkers for cognitive dysfunction.ConclusionsThese findings might contribute to a more accurate clinical diagnosis and differentiation among normal controls, SIVDND, SVMCI, and VD patients, in order to develop appropriate intervention approaches for SIVD therapeutic treatment.

IntroductionIn swine disease surveillance, obtaining labeled data for supervised learning models can be challenging because many farms lack standardized diagnostic routines and consistent health monitoring systems. Unsupervised learning is particularly suitable in such scenarios because it does not require labeled data, allowing for detecting anomalies without predefined labels. This study evaluates the effectiveness of unsupervised machine learning models in detecting anomalies in productivity indicators in swine breeding herds.MethodsAnomalies, defined as deviations from expected patterns, were identified in indicators such as abortions per 1000 sows, prenatal losses, preweaning mortality, total born, liveborn, culled sows per 1000 sows, and dead sows per 1000 sows. Three unsupervised models - Isolation Forest, Autoencoder, and K-Nearest Neighbors (KNN) - were applied to data from two swine production systems. The herd-week was used as the unit of analysis, and anomaly scores above the 75th percentile were used to flag anomalous weeks. A permutation test assessed differences between anomalous and non-anomalous weeks. Performance was evaluated using F1-score, precision, and recall, with true anomalous weeks defined as those coinciding with reported health challenges, including porcine reproductive and respiratory syndrome (PRRS) and Seneca Valley virus outbreaks. A total of 8,044 weeks were analyzed.ResultsThe models identified 336 anomalous weeks and 1,008 non-anomalous weeks in Production System 1, and 1,675 anomalous weeks and 5,025 non-anomalous weeks in Production System 2. The results from the permutation test revealed significant differences in productivity indicators between anomalous and non-anomalous weeks, especially during PRRS outbreaks, with more subtle changes observed during Seneca Valley virus outbreaks. The models performed well in detecting the PRRSV anomaly, achieving perfect precision (100%) across all models for both production systems. For anomalies like SVV the models showed lower performance compared to PRRSV.DiscussionThese findings suggest that unsupervised machine learning models are promising tools for early disease detection in swine herds, as they can identify anomalies in productivity data that may signal health challenges.

The genetic basis of amyloid β (Aβ) deposition in subcortical vascular cognitive impairment (SVCI) is still unknown. Here, we investigated genetic variants involved in Aβ deposition in patients with SVCI. We recruited a total of 110 patients with SVCI and 424 patients with Alzheimer's disease-related cognitive impairment (ADCI), who underwent Aβ positron emission tomography and genetic testing. Using candidate AD-associated single nucleotide polymorphisms (SNPs) that were previously identified, we investigated Aβ-associated SNPs that were shared or distinct between patients with SVCI and those with ADCI. Replication analyses were performed using the Alzheimer's Disease Neuroimaging Initiative (ADNI) and Religious Orders Study and Rush Memory and Aging Project cohorts (ROS/MAP). We identified a novel SNP, rs4732728, which showed distinct associations with Aβ positivity in patients with SVCI (P interaction = 1.49 × 10-5); rs4732728 was associated with increased Aβ positivity in SVCI but decreased Aβ positivity in ADCI. This pattern was also observed in ADNI and ROS/MAP cohorts. Prediction performance for Aβ positivity in patients with SVCI increased (area under the receiver operating characteristic curve = 0.780; 95% confidence interval = 0.757-0.803) when rs4732728 was included. Cis-expression quantitative trait loci analysis demonstrated that rs4732728 was associated with EPHX2 expression in the brain (normalized effect size = -0.182, P = 0.005). The novel genetic variants associated with EPHX2 showed a distinct effect on Aβ deposition between SVCI and ADCI. This finding may provide a potential pre-screening marker for Aβ positivity and a candidate therapeutic target for SVCI.

Previous studies on Alzheimer’s disease-type cognitive impairment (ADCI) and subcortical vascular cognitive impairment (SVCI) has rarely explored spatiotemporal heterogeneity. This study aims to identify distinct spatiotemporal cortical atrophy patterns in ADCI and SVCI. 1,338 participants (713 ADCI, 208 SVCI, and 417 cognitively unimpaired elders) underwent brain magnetic resonance imaging (MRI), amyloid positron emission tomography, and neuropsychological tests. Using MRI, this study measures cortical thickness in five brain regions (medial temporal, inferior temporal, posterior medial parietal, lateral parietal, and frontal areas) and utilizes the Subtype and Stage Inference (SuStaIn) model to predict the most probable subtype and stage for each participant. SuStaIn identifies two distinct cortical thinning patterns in ADCI (medial temporal: 65.8%, diffuse: 34.2%) and SVCI (frontotemporal: 47.1%, parietal: 52.9%) patients. The medial temporal subtype of ADCI shows a faster decline in attention, visuospatial, visual memory, and frontal/executive domains than the diffuse subtype (p-value < 0.01). However, there are no significant differences in longitudinal cognitive outcomes between the two subtypes of SVCI. Our study provides valuable insights into the distinct spatiotemporal patterns of cortical thinning in patients with ADCI and SVCI, suggesting the potential for individualized therapeutic and preventive strategies to improve clinical outcomes.

Purpose: The aim of this study was to evaluate microstructural changes of major white matter (WM) tracts in patients with vascular cognitive impairment (VCI).Method: Diffusion tensor imaging (DTI) data were obtained from 24 subjects with subcortical ischemic vascular disease (SIVD), including 13 subjects with VCI-no dementia (VCIND) and 11 subjects with normal cognition (as a control group). A tract-based spatial statistics approach was performed to investigate WM microstructure in VCIND by integrating multiple indices including fractional anisotropy (FA) and mean diffusivity (MD), which are intra-voxel metrics, and local diffusion homogeneity (LDH), which is an inter-voxel metric.Results: The VCIND group had decreased FA and increased MD values throughout widespread WM areas predominately in the corpus callosum, bilateral internal capsule/corona radiata/posterior thalamic radiation/inferior fronto-occipital fasciculus and right inferior/superior longitudinal fasciculus. There was a slight discrepancy between the distribution of areas with decreased FA and LDH. The FA, MD and LDH values were significantly correlated with cognitive test results. According to a WM tract atlas, 10 major tracts were identified as tracts of interest in which three diffusion metrics simultaneously differed between groups, including bilateral anterior thalamic radiation, forceps minor, right corticospinal tract, bilateral inferior fronto-occipital fasciculus, left inferior and superior longitudinal fasciculus, and bilateral uncinate fasciculus. Receiver operating characteristic (ROC) analysis demonstrated the feasibility of using diffusion metrics along the forceps minor and left anterior thalamic radiation for separating two groups.Conclusion: The results suggest WM microstructural abnormalities contribute to cognitive impairments in SIVD patients. DTI parameters may be potential biomarkers for detecting VCIND from SIVD.

Alzheimer’s disease (AD) is characterized by the presence of β-amyloid (Aβ) and tau, and subcortical vascular cognitive impairment (SVCI) is characterized by cerebral small vessel disease (CSVD). They are the most common causes of cognitive impairment in the elderly population. Concurrent CSVD burden is more commonly observed in AD-type dementia than in other neurodegenerative diseases. Recent developments in Aβ and tau positron emission tomography (PET) have enabled the investigation of the relationship between AD biomarkers and CSVD in vivo. In this review, we focus on the interaction between AD and CSVD markers and the clinical effects of these two markers based on molecular imaging studies. First, we cover the frequency of AD imaging markers, including Aβ and tau, in patients with SVCI. Second, we discuss the relationship between AD and CSVD markers and the potential distinct pathobiology of AD markers in SVCI compared to AD-type dementia. Next, we discuss the clinical effects of AD and CSVD markers in SVCI, and hemorrhagic markers in cerebral amyloid angiopathy. Finally, this review provides both the current challenges and future perspectives for SVCI.

Abnormal structures in the cortical and subcortical regions have been identified in subcortical vascular cognition impairment (SVCI). However, little is known about the functional alterations in SVCI, and no study refers to the functional connectivity in the prefrontal and subcortical regions in this context. The medial prefrontal cortex (MPFC) is an important region of the executive network and default mode network, and the subcortical thalamus plays vital roles in mediating or modulating these two networks. To investigate both thalamus- and MPFC-related functional connectivity as well as its relationship with cognition in SVCI, 32 SVCI patients and 23 control individuals were administered neuropsychological assessments. They also underwent structural and functional magnetic resonance imaging scans. Voxel-based morphometry and functional connectivity analysis were performed to detect gray matter (GM) atrophy and to characterize the functional alterations in the thalamus and the MPFC. For structural data, we observed that GM atrophy was distributed in both cortical regions and subcortical areas. For functional data, we observed that the thalamus functional connectivity in SVCI was significantly decreased in several cortical regions [i.e., the orbitofrontal lobe (OFL)], which are mainly involved in executive function and memory function. However, connectivity was increased in several frontal regions (i.e., the inferior frontal gyrus), which may be induced by the compensatory recruitment of the decreased functional connectivity. The MPFC functional connectivity was also decreased in executive- and memory-related regions (i.e., the anterior cingulate cortex) along with a motor region (i.e., the supplementary motor area). In addition, the cognitive performance was closely correlated with functional connectivity between the left thalamus and the left OFL in SVCI. The present study, thus, provides evidence for an association between structural and functional alterations, and sheds light on the underlying neural mechanisms of executive dysfunction in SVCI.

The clinical effects of β-amyloid positivity (Aβ+) on copathologies in various dementias remain relatively underexamined. Thus, the present study was conducted to investigate the prevalence and clinical effects of Aβ+ in subcortical vascular cognitive impairment (SVCI) and frontotemporal dementia (FTD). We enrolled SVCI (n = 583), FTD (n = 152), and cognitively unimpaired (CU) participants (n = 1,249) who underwent Aβ PET scans. The odds of having Aβ+ were subsequently compared among the diagnostic groups (CU, SVCI, and FTD) according to age and apolipoprotein E genotype. Additionally, a linear mixed-effects model was used to investigate the effects of Aβ+ on cognitive trajectories in SVCI and FTD. Compared with CU, the SVCI group had a higher prevalence of Aβ+ in the 75 to 90 years age group (adjusted odds ratio, 1.97; 95% confidence interval, 1.36-2.85; P < 0.001), as well as within the apolipoprotein E ε3/ε3 group (adjusted odds ratio, 1.78; 95% confidence interval, 1.20-2.63; P = 0.001), whereas the FTD group showed no difference in Aβ+ prevalence. Aβ+ was associated with a worse cognitive trajectory in SVCI (adjusted β-coefficient = -0.6424; P < 0.001), but not in FTD. These findings contribute to our understanding of Aβ biomarker traits in various dementias in Korea.

Few studies have investigated differences in functional connectivity (FC) between patients with subcortical ischemic vascular disease (SIVD) and Alzheimer’s disease (AD), especially in relation to apathy. Therefore, the aim of this study was to compare apathy-related FC changes among patients with SIVD, AD, and cognitively normal subjects. The SIVD group had the highest level of apathy as measured using the Apathy Evaluation Scale-clinician version (AES). Dementia staging, volume of white matter hyperintensities (WMH), and the Beck Depression Inventory were the most significant clinical predictors for apathy. Group-wise comparisons revealed that the SIVD patients had the worst level of “Initiation” by factor analysis of the AES. FCs from four resting state networks (RSNs) were compared, and the connectograms at the level of intra- and inter-RSNs revealed dissociable FC changes, shared FC in the dorsal attention network, and distinct FC in the salient network across SIVD and AD. Neuronal correlates for “Initiation” deficits that underlie apathy were explored through a regional-specific approach, which showed that the right inferior frontal gyrus, left middle frontal gyrus, and left anterior insula were the critical hubs. These findings broaden the disconnection theory by considering the effect of FC interactions across multiple RSNs on apathy formation.

Subcortical vascular cognitive impairment (SVCI) is characterized by the presence of cerebral small vessel disease (CSVD) markers. Some SVCI patients also show Alzheimer's disease and cerebral amyloid angiopathy markers. However, the effects of these imaging markers on long-term clinical outcomes have not yet been established. The present study, therefore, aimed to determine how these imaging markers influence functional disability and/or mortality. We recruited 194 participants with SVCI from the memory clinic and followed them up. All participants underwent brain magnetic resonance imaging at baseline, and 177 (91.2%) participants underwent beta-amyloid (Aβ) positron emission tomography. We examined the occurrence of ischemic or hemorrhagic strokes. We also evaluated functional disability and mortality using the modified Rankin scale. To determine the effects of imaging markers on functional disability or mortality, we used Fine and Gray competing regression or Cox regression analysis. During a 8.6-year follow-up period, 46 of 194 patients (23.7%) experienced a stroke, 110 patients (56.7%) developed functional disabilities and 75 (38.6%) died. Aβ positivity (subdistribution hazard ratio [SHR]=2.73), greater white matter hyperintensity (WMH) volume (SHR=3.11) and ≥3microbleeds (SHR=2.29) at baseline were independent predictors of functional disability regardless of the occurrence of stroke. Greater WMH volume (hazard ratio =2.07) was an independent predictor of mortality. Our findings suggest that diverse imaging markers may predict long-term functional disability and mortality in patients with SVCI, which in turn may provide clinicians with a more insightful understanding of the long-term outcomes of SVCI.

PurposeThis study aimed to investigate the dynamic functional connectivity (DFC) pattern in cerebral small vessel disease (CSVD) and explore the relationships between DFC temporal properties and cognitive impairment in CSVD.MethodsFunctional data were collected from 67 CSVD patients, including 35 patients with subcortical vascular cognitive impairment (SVCI) and 32 cognitively unimpaired (CU) patients, as well as 35 healthy controls (HCs). The DFC properties were estimated by k-means clustering analysis. DFC strength analysis was used to explore the regional functional alterations between CSVD patients and HCs. Correlation analysis was used for DFC properties with cognition and SVD scores, respectively.ResultsThe DFC analysis showed three distinct connectivity states (state I: sparsely connected, state II: strongly connected, state III: intermediate pattern). Compared to HCs, CSVD patients exhibited an increased proportion in state I and decreased proportion in state II. Besides, CSVD patients dwelled longer in state I while dwelled shorter in state II. CSVD subgroup analyses showed that state I frequently occurred and dwelled longer in SVCI compared with CSVD-CU. Also, the internetwork (frontal-parietal lobe, frontal-occipital lobe) and intranetwork (frontal lobe, occipital lobe) functional activities were obviously decreased in CSVD. Furthermore, the fractional windows and mean dwell time (MDT) in state I were negatively correlated with cognition in CSVD but opposite to cognition in state II.ConclusionPatients with CSVD accounted for a higher proportion and dwelled longer mean time in the sparsely connected state, while presented lower proportion and shorter mean dwell time in the strongly connected state, which was more prominent in SVCI. The changes in the DFC are associated with altered cognition in CSVD. This study provides a better explanation of the potential mechanism of CSVD patients with cognitive impairment from the perspective of DFC.

Donepezil in patients with subcortical vascular cognitive impairment: a randomised double-blind trial in CADASIL