Background and Purpose—: The aim of the present study was to investigate atherosclerotic plaque characteristics in patients with artery-to-artery (A-to-A) embolic infarction by whole-brain high-resolution magnetic resonance imaging. Methods—: Seventy-four patients (mean age, 54.7±12.1 years; 59 men) with recent stroke in the territory of middle cerebral artery because of intracranial atherosclerotic disease were prospectively enrolled. Whole-brain high-resolution magnetic resonance imaging was performed in all the patients both precontrast and postcontrast administration by using a 3-dimensional T1-weighted vessel wall magnetic resonance imaging technique known as inversion-recovery prepared sampling perfection with application-optimized contrast using different flip angle evolutions. Patients were divided into A-to-A embolic infarction and non–A-to-A embolic infarction groups based on diffusion-weighted imaging findings. The characteristics of the intracranial atherosclerotic plaques were compared between groups. Results—: A total of 74 intracranial atherosclerotic plaques were analyzed (36 in A-to-A embolism group and 38 in non–A-to-A embolism group). Hyperintense plaques (HIPs) were more frequently observed in A-to-A embolism group (75.0% versus 21.1%; P <0.001). Eighteen of the 27 HIPs (66.7%) demonstrated hyperintense spots or areas located adjacent to the lumen versus 9 HIPs (33.3%) located within the plaque in A-to-A embolism group. Furthermore, a higher prevalence of plaque surface irregularity was also observed in A-to-A embolism group (41.7% versus 18.4%; P =0.029). Logistic regression analysis showed that HIP was the most powerful independent predictor of A-to-A embolic infarction ( P <0.001), with the odds ratio of 11.2 (95% confidence interval, 3.5–36.2). Conclusions—: A-to-A embolic infarction has distinct vulnerable plaque characteristics compared with non–A-to-A embolic infarction. HIP and plaque surface irregularity may predict A-to-A embolic infarction.

Background: Intracranial atherosclerotic disease tends to affect multiple arterial segments. Using whole‐brain vessel wall imaging, we sought to study the differences in plaque features among various types of plaques in patients with a recent unilateral anterior circulation ischemic stroke. Methods and Results: Sixty‐one patients with unilateral anterior circulation ischemic stroke were referred to undergo whole‐brain vessel wall imaging (before and after contrast) within 1 month of symptom onset for intracranial atherosclerotic disease evaluations. Each plaque was classified as a culprit, probably culprit, or nonculprit lesion, according to its likelihood of causing the stroke. The associations between plaque features (thickening pattern, plaque‐wall contrast ratio, high signal on T1‐weighted images, plaque contrast enhancement ratio, enhancement grade, and enhancement pattern) and culprit lesions were estimated using mixed multivariable logistic regression after adjustment for maximum wall thickness. In 52 patients without motion corruption in whole‐brain vessel wall imaging, a total of 178 intracranial plaques in the anterior circulation were identified, including 52 culprit lesions (29.2%), 51 probably culprit lesions (28.7%), and 75 nonculprit lesions (42.1%). High signal on T1‐weighted images (adjusted odds ratio, 9.1; 95% confidence interval, 1.9–44.1; P =0.006), grade 2 (enhancement ratio of plaque ≥ enhancement ratio of pituitary) contrast enhancement (adjusted odds ratio, 17.4; 95% confidence interval, 1.8–164.9; P =0.013), and type 2 (≥50% cross‐sectional wall involvement) enhancement pattern (adjusted odds ratio, 10.1; 95% confidence interval, 1.3–82.2; P =0.030) were independently associated with culprit lesions. Conclusions: High signal on T1‐weighted images, grade 2 contrast enhancement, and type 2 enhancement pattern are associated with cerebrovascular ischemic events, which may provide valuable insights into risk stratification.

Background and Purpose—: We aimed to systematically investigate the characteristics of cervicocranial artery dissection (CCAD) on high-resolution magnetic resonance imaging that are associated with acute ischemic stroke. Methods—: Patients with CCAD were recruited and divided into stroke and nonstroke groups. The lesion location, the presence of a double lumen, intimal flap, intramural hematoma, pseudoaneurysm, irregular surface, intraluminal thrombus, and other quantitative parameters of each dissected segment were reviewed. Multiple logistic regression was used to examine the association between imaging features of CCAD and ischemic stroke. Results—: A total of 145 affected vessels from 118 patients with CCAD were analyzed. Anterior circulation, intramural hematoma, irregular surface, intraluminal thrombus, and severe stenosis (>70%) on high-resolution magnetic resonance imaging were more prevalent in CCAD patient with stroke (54.4% versus 36.4%; P =0.030, 96.2% versus 84.8%; P =0.017, 74.7% versus 37.9%; P <0.001, 44.3% versus 4.5%; P <0.001, and 54.4% versus 31.8%; P =0.008, respectively). In multivariable logistic regression analysis, the presence of irregular surface and intraluminal thrombus on imaging were independently associated with acute ischemic stroke in CCAD with odds ratios of 4.29 (95% CI, 1.61–11.46, P =0.004) and 7.48 (95% CI, 1.64–34.07, P =0.009). Conclusions—: The current findings supported that the presence of irregular surface and intraluminal thrombus were related to stroke occurrence in patients with CCAD. High-resolution magnetic resonance imaging might give insights into pathogenesis of ischemic stroke in CCAD. It may be useful for individual prediction of ischemic stroke early in CCAD.

Objective: To investigate postoperative functional connectivity (FC) alterations across impaired cognitive domains and their causal relationships with systemic inflammation. Background: Postoperative cognitive dysfunction commonly occurs after cardiac surgery, and both systemic and neuroinflammation may trigger its development. Whether FC alterations underlying deficits in specific cognitive domains after cardiac surgery are affected by inflammation remains unclear. Methods: Seventeen patients, who underwent cardiac valve replacement, completed a neuropsychological test battery and brain MRI scan before surgery and on days 7 and 30 after surgery compared to age-matched healthy controls. Blood samples were taken for tumor necrosis factor-a and interleukin-6 measurements. Seed-to-voxel FC of the left dorsolateral prefrontal cortex (DLPFC) was examined. Bivariate correlation and linear regression models were used to determine the relationships among cognitive function, FC alterations, and cytokines. Results: Executive function was significantly impaired after cardiac surgery. At day 7 follow-up, the surgical patients, compared to the controls, demonstrated significantly decreased DLPFC FC with the superior parietal lobe and attenuated negative connectivity in the default mode network, including the angular gyrus and posterior cingulate cortex. The left DLPFC enhanced the connectivity in the right DLPFC and posterior cingulate cortex, all of which were related to the increased tumor necrosis factor-a and decreased executive function up to day 7 after cardiac surgery. Conclusions: The decreased FC of executive control network and its anticorrelation with the default mode network may contribute to executive function deficits after cardiac surgery. Systemic inflammation may trigger these transient FC changes and executive function impairments.

Cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) and heterozygous HTRA1 mutation-related cerebral small vessel disease (CSVD) are the two types of dominant hereditary CSVD. Blood–brain barrier (BBB) failure has been hypothesized in the pathophysiology of CSVD. However, it is unclear whether there is BBB damage in the two types of hereditary CSVD, especially in heterozygous HTRA1 mutation-related CSVD. In this study, a case-control design was used with two disease groups including CADASIL (n = 24), heterozygous HTRA1 mutation-related CSVD (n = 9) and healthy controls (n = 24). All participants underwent clinical cognitive assessments and brain MRI. Diffusion-prepared pseudo-continuous arterial spin labelling was used to estimate the water exchange rate across the BBB (k_w). Correlation and multiple linear regression analyses were used to examine the association between k_w and disease burden and neuropsychological performance, respectively. Compared with the healthy controls, k_w in the whole brain and multiple brain regions was decreased in both CADASIL and heterozygous HTRA1 mutation-related CSVD patients (Bonferroni-corrected P < 0.007). In the CADASIL group, decreased k_w in the whole brain (β = −0.634, P = 0.001), normal-appearing white matter (β = −0.599, P = 0.002) and temporal lobe (β = −0.654, P = 0.001) was significantly associated with higher CSVD score after adjusting for age and sex. Reduced k_w in the whole brain was significantly associated with poorer neuropsychological performance after adjusting for age, sex and education in both CADASIL and heterozygous HTRA1 mutation-related CSVD groups (β = 0.458, P = 0.001; β = 0.884, P = 0.008). This study showed that there was decreased water exchange rate across the BBB in both CADASIL and heterozygous HTRA1 mutation-related CSVD patients, suggesting a common pathophysiological mechanism underlying the two types of hereditary CSVD. These results highlight the potential use of k_w for monitoring the course of CADASIL and heterozygous HTRA1 mutation-related CSVD, a possibility which should be tested in future research.

Artificial intelligence (AI), particularly deep learning, has demonstrated remarkable performance in medical imaging across a variety of modalities, including X-ray, computed tomography (CT), magnetic resonance imaging (MRI), ultrasound, positron emission tomography (PET), and pathological imaging. However, most existing state-of-the-art AI techniques are task-specific and focus on a limited range of imaging modalities. Compared to these task-specific models, emerging foundation models represent a significant milestone in AI development. These models can learn generalized representations of medical images and apply them to downstream tasks through zero-shot or few-shot fine-tuning. Foundation models have the potential to address the comprehensive and multifactorial challenges encountered in clinical practice. This article reviews the clinical applications of both task-specific and foundation models, highlighting their differences, complementarities, and clinical relevance. We also examine their future research directions and potential challenges. Unlike the replacement relationship seen between deep learning and traditional machine learning, task-specific and foundation models are complementary, despite inherent differences. While foundation models primarily focus on segmentation and classification, task-specific models are integrated into nearly all medical image analyses. However, with further advancements, foundation models could be applied to other clinical scenarios. In conclusion, all indications suggest that task-specific and foundation models, especially the latter, have the potential to drive breakthroughs in medical imaging, from image processing to clinical workflows.