Multiple Sclerosis, Fat-Saturated T2-Weighted MRI, FLAIR MRI, Agreement
Objectives: To evaluate the correlation and the agreement of FLAIR FS and FLAIR MRI for the assessment of MS lesions. To assess difference in the demonstration of the number of lesions
Diagnostic agreement and test comparison study. Brain MRI studies of patients with confirmed clinical diagnosis of MS performed during the period from 2013 to 2014 were selected by an independent investigator. These were performed according to the institutional protocol, in an HD23 General Electric 1.5 Tesla MRI scanner (GE Medical Systems, Milwaukee, Wisc. USA). Prior estimation of sample size, twelve pairs of FLAIR FS and FLAIR images were randomly distributed without data. One neuroradiologist and one radiology resident reviewed the images in Barco Nio 3MP displays (E-3620 MA) (BOAT, Brussels, Belgium). They counted the number of cortical, juxtacortical, periventricular, deep white matter and posterior fossa lesions. Concordance correlation coefficient (CCC) of Lin, and paired t-test statistics were used.
Results: The Pearson correlation was r = 0.98, p = < .001. Lin´s CCC was pc = 0.83 (95% CI, 0.67 - 0.98), bias correction factor Cb = 0.89 and slope = 0.75. The mean difference between sequences was 6.58 SD 7.67, Bland Altman 95% limits of agreement from -21.61 to 8.45. For radiologist A the mean difference between number of lesion observed in FLAIR-FS and FLAIR was 6.58 (mean FLAIR 17, mean FLAIR FS 23.58, 95% CI: 1.71-11.46, p < 0.0063). For radiologist B the mean difference between number of lesion observed in FLAIR-FS and FLAIR was 7.92 (mean FLAIR 21, mean FLAIR FS 29.33, 95% CI: 1.76-14.07, p < 0.0082).
Conclusions: There was a positive and high correlation between the number of lesions observed in FLAIR FS and FLAIR. The Lin coefficient showed that agreement was poor, because more number of lesions were observed in FLAIR FS, this was statistically significant and demonstrates the usefulness of this new sequence in the diagnosis and monitoring of MS patients.