Cortical lesion detection using FLAWS in multiple sclerosis
AUTHORS: Müller J, Rahmanzadeh R, Tsagkas C, Barakovic M, Lu PJ, Yaldizli O, Weigel M, Beaumont J, La Rosa F, Bach Cuadra M, Kappos L, Kuhle J, Gambarota G, Granziera C
ECTRIMS 2021 - Multiple Sclerosis Journal, 27(2_SUPPL): 465, October 2021
Introduction: Cortical lesions (CL) are common in multiple sclerosis (MS), but their visualization is challenging on conventional MRI. Specialized T1-weighted sequences such as Magnetization Prepared 2 Rapid Acquisition Gradient Echoes (MP2RAGE) are able to improve cortical lesion detection. FLuid And White matter Suppression (FLAWS) provides an exquisite sensitivity to the grey matter signal and therefore may further facilitate cortical lesion identification.
Objective: To compare the manual detection rate of CL when using MP2RAGE or FLAWS.
Methods: We included 30 relapsing-remitting MS patients who underwent magnetic resonance imaging with isotropic spatial resolution of 1mm at 3T, including MP2RAGE and FLAWS. CL were segmented manually and classified as intracortical (ICL) or leukocortical lesions (LCL) by consensus of three trained raters on (i) MP2RAGE uniform/flat images and (ii) minimum FLAWS-contrast images that suppress both white matter and cerebro-spinal fluid signals (grey-matter-specific). Lesion volumes were calculated using an in-house algorithm. Number of CL, ICL, LCL and lesions volumes were compared using Wilcoxon signed rank tests. Comparison of proportions was performed using Friedman test. In an exploratory analysis, individual CL counts of the three raters were compared and interrater variability was quantified using Fleiss ϰ.
Results: CLs were identified in 20/30 patients on MP2RAGE and 24/30 patients on FLAWS (66.6% and 80%, respectively; p=0.178). In total, 391 CL were identified, hereof 201 on MP2RAGE (47 ICL, 154 LCL, median CLs per patient [range]: 3 [0-56]) and 190 on FLAWS (48 ICL, 142 LCL, median CLs per patient [range]: 4 [0-36]; all p>0.05). Proportions of ICL and LCL were similar between MP2RAGE (23.4 and 76.6%) and FLAWS (25.3 and 74.7%; both p>0.05). Per patient, total lesion loads of CL, ICL and LCL were comparable between sequences, but volumes of segmented CL were larger on FLAWS (median [IQR]: 33 µl ) than on MP2RAGE (24 µl ; p=0.027). Interrater agreement of lesion counts was moderate for CL (MP2RAGE ϰ=0.436; FLAWS ϰ=0.467; both p<0.001), LCL (MP2RAGE ϰ=0.426, FLAWS ϰ=0.463; both p<0.001) and ICL on MP2RAGE (ϰ=0.514, p<0.001), but substantial for ICL on FLAWS (ϰ=0.661, p<0.001).
Conclusion: Size of identified cortical lesions was larger on FLAWS than on MP2RAGE while overall detection rate of cortical lesions was similar when performed on MP2RAGE or FLAWS.