Vascular cognitive impairment (VCI) is the second most prevalent form of dementia, presenting significant challenges for diagnosis and management. VCI encompasses a continuum from subjective cognitive decline to full-blown dementia, primarily caused by cerebrovascular damage. With projections estimating 1.5 billion individuals over 65 by 2050, the burden of VCI is expected to rise dramatically, particularly in low- and middle-income countries. Traditional diagnostic methods, including magnetic resonance imaging (MRI), are limited in predicting disease progression. Radiomics, a novel approach that quantitatively analyses imaging data, offers new opportunities for early detection, diagnosis and differentiation of VCI. By extracting and modelling imaging features, radiomics may provide critical insights into brain changes linked with cognitive decline.
Understanding VCI and Radiomics
VCI arises from risk factors such as hypertension, diabetes and hyperlipidaemia, which lead to cerebrovascular injury and subsequent cognitive impairment. It includes three main subtypes: vascular cognitive impairment no dementia (VCIND), vascular dementia (VD) and mixed dementia. VCIND represents a milder form that does not interfere with daily activities but poses a high risk for progression. VD, on the other hand, results in significant cognitive dysfunction and is the second leading cause of dementia globally. Mixed dementia often combines vascular pathology with Alzheimer's disease (AD).
Radiomics involves the high-throughput extraction of quantitative features from medical images and applying these to clinical decision-making. In MRI radiomics, key steps include image acquisition, segmentation of biologically significant regions, feature extraction, selection and model construction. Tools such as ITK-SNAP and 3D Slicer facilitate segmentation, while MATLAB and PyRadiomics support feature extraction. Feature selection is crucial to avoid overfitting, and modelling often employs support vector machines (SVM), logistic regression (LR) or random forests (RF) to analyse disease-related patterns.
Applications in Prediction, Diagnosis and Differentiation
MRI radiomics is increasingly applied to cognitive disorders, including VCI. Research shows that radiomics can predict the development of VCI by analysing pathology types such as infarcts and myelin loss, particularly those linked to small vessel disease. One study used deep learning to assess white matter hyperintensities, building an LR model that demonstrated high accuracy in detecting cognitive impairment. Other studies have focused on radiomics analysis of lenticulostriate arteries (LSA) and amyloid plaques, further reinforcing the value of this technique. For instance, radiomics features from T1 and T2 FLAIR sequences, combined with clinical data, improved predictions of amyloid positivity, a marker of cerebral amyloid angiopathy and VCI risk.
Must Read: Advancing Cerebral Artery Imaging with 4D CTA
In diagnosis, radiomics has been used alongside deep learning to identify VCI more accurately. Algorithms such as a multi-modal convolutional neural network (CNN) significantly enhanced image segmentation precision. These improvements translated into more accurate clinical assessments, with clear distinctions between VCI and non-VCI groups. Furthermore, studies have employed radiomics to examine microstructural brain changes, such as in the amygdala and thalamus, identifying features that serve as reliable biomarkers. For subcortical ischemic vascular cognitive impairment without dementia (SIVCIND), high-resolution MRI combined with machine learning achieved high diagnostic accuracy.
Radiomics also plays a role in differentiating VD from AD, which often present with overlapping symptoms. Structural MRI features analysed through radiomics and SVM models enabled efficient classification between these dementia subtypes. This capability is particularly valuable in clinical settings, where determining the underlying cause of dementia is essential for guiding treatment.
Challenges and Future Directions
Despite its promise, MRI radiomics faces several limitations. Many studies are retrospective with small sample sizes, leading to potential bias and false positives. There is also a tendency to rely on single-sequence imaging, which may limit the depth of analysis. Clinical adoption remains limited due to these methodological constraints and the need for further validation. In addition, the shortage of interdisciplinary professionals skilled in both medical and computational sciences poses a barrier to progress.
However, the future of radiomics in VCI research is promising. Advances in image segmentation, the use of multi-sequence data, longer follow-up periods and growing computational resources are expected to improve its clinical utility. As radiomics becomes more standardised and integrated with clinical factors, it may emerge as a key tool for early screening, accurate diagnosis and tailored intervention for VCI.
MRI radiomics offers a transformative approach to the early detection, diagnosis and differentiation of vascular cognitive impairment. By leveraging quantitative imaging features and machine learning models, radiomics enhances the precision of clinical assessments and supports proactive care. Although challenges remain, ongoing advancements in technology, methodology and interdisciplinary collaboration are likely to make radiomics an integral part of future diagnostic pathways for VCI.
Source: Insights into Imaging
Image Credit: iStock
