Danny R. Rose, Jr., MD

Tozer DJ, Zeestraten E, Lawrence AJ, Barrick TR, Markus HS. Texture Analysis of T1-Weighted and Fluid-Attenuated Inversion Recovery Images Detects Abnormalities That Correlate With Cognitive Decline in Small Vessel Disease. Stroke. 2018

Cerebral small vessel disease (SVD) is a well-established but relatively poorly understood cause of lacunar stroke and vascular dementia. While the complex structural changes underpinning this disease process have yet to be fully elucidated, several MRI imaging markers have been identified in association with SVD, including white matter hyperintensities (WMH), cerebral microbleeds, and evidence of lacunar infarction. While these imaging markers have been useful in assessing severity of disease, they are less strongly associated with cognitive impairment. One particular area of interest as further imaging research into SVD has progressed is the so-called “normal appearing white matter” (NAWM), specifically the areas of white matter on traditional MRI that do not show the typical T2/FLAIR hyperintense changes.

Given the understanding of SVD as a gradient of changes to white matter, researchers have investigated NAWM with more advanced imaging techniques. Diffusion tensor imaging (DTI) in particular has been found to detect changes in NAWM that correlate more strongly with cognitive impairment than T2 WMH load. While this new insight has advanced our understanding of the changes in this tissue before becomingly overtly evident on traditional MRI, DTI is not yet ubiquitous enough to be able to use for populations. Tozer et al. sought to address this by using texture analysis (TA), an imaging post-processing technique looking at the intensities of neighboring pixels. In the current paper, they propose TA as a tool to detect changes in NAWM and to assess if these changes were associated with cognitive changes.

The authors analyzed 99 symptomatic SVD subjects (defined as a history of lacunar syndrome or Fazekas Grade ≥2 for WMH) from the St George’s Cognition and Neuroimaging in Stroke study and 54 control subjects from the St. George’s Neuropsychology and Imaging in Elderly study. All subjects had cerebrovascular risk factors recorded and underwent annual neuropsychological assessment for 5 years to assess for cognitive decline and conversion to dementia. The texture parameters (TP) used in texture analysis were cross-sectionally compared between SVD subjects and controls. Additionally, in the SVD cohort, a multivariate regression analysis was performed to see which TP were independently predictive of cognitive measures, controlling for other imaging parameters, age, sex and premorbid IQ. Analysis comparing SVD subjects and controls revealed that several texture parameters derived from T1 and FLAIR sequences, with sum entropy in particular having a P<0.0001 for both image types. Specific TP including entropy, sum average and MCC were correlated with executive function and global cognition at baseline as well as predicting conversion to dementia. TP parameters did not correlate to processing speed, which has been shown in previous studies to be more strongly associated with lacunar infarcts than diffuse white matter changes.

This study represents a promising method of analysis using traditional 1.5T MRI sequences to assess the integrity of NAWM and suggests that certain parameters relate to cognitive decline. It is important to note that these texture parameters do not have a biological correlate with respect to structural changes in SVD, and the significance of each texture parameter in relationship to pathologic changes in the brain deserves further study. Multi-center studies are also needed to validate these results across different imaging facilities. Although DTI parameters (which were also included in the study) were found to be stronger predictors, TP could serve as an adequate replacement in clinical settings where DTI is not available.