Using Cluster Analysis To Identify Imaging Subgroups Of Age-related Cerebral Small Vessel Disease:A Pilot Study

Introduction:Cerebral small vessel disease(CSVD)is a group of pathological processes with various aetiologies that affect the small arteries,arterioles,venules,and capillaries of the brain.Typical symptoms include headache,dizziness,gait dysfunction,and declined executive function and information processing speed.The most common type of CSVD is known as age-related and vascular risk-factors-related CSVD.Early diagnosis and accurate clinical assessments of CSVD are crucial in the context of there is no reliable treatments of CSVD available at present.Typical imaging apparences of age-related CSVD include white matter hyperintensities(WMH)on the fluid attenuated inversion recovery imaging(FLAIR),dilated perivascular spaces(DPVS)on the T2 weighted imaging,and cerebral microbleeds(CMBs)on the susceptibility weighted imaging(SWI).Visual rating scales of these imaging markers are widely used in daily clinical assessments.However,these visual rating scales are time-consuming to apply,having a ceiling effect and using rather broad categories for severity.In order to obtain a better assessments of age-related CSVD patients,here we performed a Two Step Cluster Analysis to cluster the results of visual rating scales for identifying clinical subgroups of age-related CSVDObjectives:To identify the subgroups of age-related CSVD based on daily imaging visual rating scales,in order to obtain a better clinical assessments of age-related CSVDMethods:From Feb 2017 to Dec 2018,264 subjects with headache,dizziness,gait dysfunction,and declined executive function and information processing speed were enrolled in this study.Typical imaging apparences like WMH,DPVS and CMBs were obtained by MRI exams.A battery of neuropsychological assessments scales were performed including the Trail Making test(TMT)and Stroop color-ward test(SCWT)for assessments of executive function and processing speed.Imaging visual rating scales including the Fazekas scale for WMH,DPVS scale,and Microbleed Anatomical Rating Scale(MARS)were also applied for images assessments.Two step cluster analysis was performed by the SPSS to indentify any possible subgroups of age-related CSVD under 0%,5%,10%and 15%noise threshold.Tract-Based Spatial Statistics(TBSS)and diffusion tensor imaging(DTI)based structural network were performed to evaluated the differences between the subgroups generated from the cluster analysis.Results:Subjects with mild WMH(Fazekas 1)and low DPVS alone were constantly obtained from the cluster analysis under 0%,5%,10%and 15%noise threshold.Subjects with low DPVS scores and high MARS rating scores were also obtained by the cluster analysis under 5%and 10%noise threshold,along with subjects with high DPVS scores and low MARS rating.However,no significant differences of cognitive performances were found between these two groups,indicating such imaging subgroups were clinical identical.No cluster was left after the noise threshold reached the level of 20%.TBSS and DTI network results were obtained to evaluated the differences between the subgroups generated by the cluster analysis.Under the 0%noise threshold,multiple network parameters and TBSS skeletonized FA values were found to be significantly different between the groups,along with cognitive rating scales such as TMT and SCWT scores.However,no correlations between the cognitive performances and DTI network parameters were found between the groups under 5%and 10%noise threshold,indicating the topological changes of DTI network were not influenced by the processing speed and executive function and vice versa.The interaction between TBSS skeletonized FA values and TMT-B were found between the subgroups generated by cluster analysis under 5%and 10%threshold,suggesting that these two groups could still be potential subgroups of age-related CSVD despite the fact that further study is crucially need to find the clinical significance of such clustering.Conclusions:using cluster analysis to generate subgroups of age-related CSVD might be a promising frontier of future research.However,further study is still needed to obtain the clinical significance of such clustering.