Diagnostic Value of Structural MRI in Spinocerebellar Ataxia Type 3 / 中山大学学报(医学科学版)

ABSTRACT

ObjectiveTo explore the role of structural MRI in the diagnosis of spinocerebellar ataxia type 3 （SCA3） and further evaluate its correlation with disease severity and disease duration. MethodsWe prospectively enrolled 81 genetically diagnosed SCA3 patients ［59 symptomatic （sym-SCA3） and 22 pre-symptomatic （pre-SCA3）］ and 35 age- and sex-matched healthy controls （HCs）. MRI structural images （3D T1 MPRAGE） and clinical data of all subjects were collected. Three observers with different radiological experience measured the width of the superior， middle and inferior cerebellar peduncle （SCP， MCP and ICP）， the anterior-posterior diameters of the pons and spinal cord at the levels of the foramen magnum and upper edge of the 3rd-5th cervical vertebra. One observer performed the measurements again 2 months later to assess for the intra- and inter-observer reliability， respectively. One-way ANOVA， rank-sum test， ROC curve and Random Forest were used to evaluate the diagnostic value of the above metrics for SCA3， and the correlation between the metrics and clinical variables was analyzed. ResultsNot depending on the radiological experience， the metrics based on morphological MRI showed high intra- and inter-observer reliability， among which bilateral superior and middle cerebellar peduncles performed best. The diameters of bilateral SCP， MCP， ICP， pons and spinal cord （except spinal cord at the level of the upper edge of the 5th cervical vertebra） decreased successively in HCs， pre-SCA3 and sym-SCA3 with a statistical difference （P<0.017）. ROC analysis revealed that the left MCP had the highest diagnostic value for pre-SCA3 （AUC=0.911）， with sensitivity， specificity and a cut-off value of 85.7%， 95.5% and 10.15 mm， respectively. In contrast， the right SCP had the highest diagnostic value for sym-SCA3 （AUC=0.999）， with sensitivity， specificity and a cut-off value of 100%， 98.3% and 2.62 mm， respectively. The Random Forest model based on the above metrics also had high diagnostic efficiency （AUC= 0.970， specificity=93.1%）， and the left MCP contributed the most. Correlation analysis showed that the above metrics had a significantly or moderately negative correlation with the Scale for the Assessment and Rating of Ataxia （SARA） and disease duration （P<0.05）. ConclusionNot depending on radiological experience， measurements of brain structure based on morphological MRI are reliable， which can help diagnose SCA3 and predict disease severity and duration. The left MCP and the right SCP perform best for predicting pre-SCA3 and sym-SCA3， respectively. Therefore， the structural MRI is recommended for assisting the clinical diagnosis of SCA3.