Journal Article Ahead of print
PubMed: 42454405 DOI: 10.1161/STROKEAHA.125.055101
Quantitative Relationship Between White Matter Hyperintensity Volume and Fazekas Score on Brain MRI.
Stroke
Authors: Yu-Yuan Xu, Francesca M Chappell, Maria Valdes Hernandez, Carmen Arteaga Reyes, Una Clancy, Daniela Jaime Garcia, Stephen Makin, Michael Stringer, Michael Thrippleton, Stewart Wiseman, Yajun Cheng, Dillys Xiaodi Liu, Rosalind Brown, Fergus Doubal, Ian J Deary, Simon Cox, Danielle Kerkhofs, Julie Staals, Walter Backes, Robert van Oostenbrugge, Anna Kopczak, Marco Duering, Joanna Wardlaw
Affiliations
- China National Clinical Research Center for Neurological Diseases, Beijing Tiantan Hospital, Capital Medical University, The University of Edinburgh, United Kingdom. (Y.-Y.X.).
- Institute for Neuroscience and Cardiovascular Research, The University of Edinburgh, United Kingdom. (Y.-Y.X. F.M.C. M.V.H. C.A.R. U.C. D.J.G. S.M. M.S. M.T. S.W. R.B. F.D. J.W.).
- UK Dementia Research Institute, The University of Edinburgh, United Kingdom. (Y.-Y.X. F.M.C. M.V.H. C.A.R. U.C. D.J.G. S.M. M.S. M.T. S.W. R.B. F.D. J.W.).
- Department of Neurology, West China Hospital, Sichuan University, Chengdu (Y.C.).
- Department of Psychiatry and Behavioral Sciences, University of California, San Francisco (D.X.L.).
- Department of Psychology, The University of Edinburgh, United Kingdom. (I.J.D. S.C.).
- Centre for Rural Health, University of Aberdeen, United Kingdom (S.M.).
- Department of Neurology, CARIM School for Cardiovascular Diseases, Maastricht University Medical Center, the Netherlands (D.K. J.S. W.B. R.v.O.).
- Institute for Stroke and Dementia Research, Ludwig-Maximilians Universität, Munich, Germany (A.K. M.D.).
- Medical Image Analysis Center (MIAC AG) and Department of Biomedical Engineering, University of Basel, Switzerland (M.D.).
Abstract
BACKGROUND
The Fazekas score is widely used to grade white matter hyperintensities (WMHs) in cerebral small vessel disease, yet the equivalent volume of each grade is unclear. We quantified the correspondence between Fazekas scores and WMH volume, normalized ratios, and derived conversion equations across populations.
METHODS
We analyzed 1220 participants from 4 mostly United Kingdom-based cohorts representing different cerebral small vessel disease severities: community-dwelling (LBC1936 [Lothian Birth Cohort 1936]), stroke (MSS2 [Mild Stroke Study 2] and MSS3 [Mild Stroke Study 3]), and a cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy-enriched cohort (INVESTIGATE [Imaging Neuro-Vascular, Endothelial & Structural Integrity in Preparation to Treat Small Vessel Diseases]). WMH burden was quantified as absolute WMH volume and as normalized ratios: WMH volume as a percentage of brain volume and WMH volume as a percentage of intracranial volume. Transition zones were defined as the overlap of WMH volume interquartile ranges between adjacent Fazekas scores. Linear, exponential (nonlinear least squares), and log-linear (ln WMH≈Fazekas score and zeros excluded) models were fitted; model fit was assessed using mean squared error, (pseudo-)R 2, and Vuong tests. Periventricular-versus-deep WMH patterns within each total Fazekas score were compared using the Kruskal-Wallis test.
RESULTS
The pooled cohort had a mean age of 69.4±8.9 years; 700 of 1227 records (57.1%) were male. WMH burden increased monotonically with a Fazekas score in all cohorts. Adjacent-grade separation was strongest in LBC1936 (all P≤0.002) but weaker at lower grades in MSS2, MSS3, and INVESTIGATE. The widest transition zones occurred between Fazekas grades of 3 to 4 and 4 to 5, whereas a clear WMH volume gap separated grades 5 and 6. Exponential models generally fit better than linear models, with lower mean squared error and supportive Vuong tests in most cohorts. Log-linear sensitivity analyses showed similar fitted trajectories after back-transformation. WMH burden did not differ consistently by periventricular-versus-deep distribution within Fazekas grades.
CONCLUSIONS
WMH burden followed an exponential trajectory across Fazekas scores, with population-dependent progression. The resulting conversion equations enabled bidirectional translation between visual Fazekas score and quantitative WMH volume, facilitating cross-study comparison, large-scale epidemiology, and might facilitate clinical decision-making in cerebral small vessel disease.
MeSH
White Matter, Brain, Magnetic Resonance Imaging, Reproducibility Of Results, Deep Learning, Cerebrovascular Disorders, Cerebral Small Vessel Diseases