The impact of 3D real-IR delayed post gadolinium MRI parameterisation on the diagnostic performance and optimal descriptor selection in Ménière's disease

ObjectivesTo compare the performance and optimal combination of MRI descriptors used for the diagnosis of M & eacute;ni & egrave;re's disease (MD) between a real-IR sequence with "zero-point" endolymph (ZPE), and an optimised real-IR sequence with negative signal endolymph (NSE).Materials and methodsThis retrospective single-centre cross-sectional study evaluated delayed post-gadolinium ZPE and NSE real-IR MRI in consecutive patients with M & eacute;ni & egrave;riform symptoms (8/2020-10/2023). Two observers assessed 14 MRI descriptors. "Definite MD" (2015 criteria) and "all MD" ears (wider clinical criteria) were compared to controls. Cohen's kappa and risk ratios (RR) were evaluated for each descriptor. Forward stepwise logistic regression established which combination of descriptors best predicted MD.ResultsThe study included 132 patients (57 men; mean age 57.7 +/- 13.6), with 87 "all MD" (56 "definite") and 39 control ears. The NSE sequence demonstrated increased perilymph SNR, and improved both diagnostic performance and reliability for 9/14 descriptors. However, ZPE demonstrated superior diagnostic performance for the best descriptor of "saccule absent, large as or confluent with the utricle" (RR 6.571, ZPE; 6.300, NSE) and that of "asymmetric perilymphatic enhancement" (RR 3.628, ZPE; 2.903, NSE). Both sequences combined these two descriptors in the optimal predictive model for "definite MD", with "grade 2 cochlear hydrops" also significant for NSE. ZPE and NSE descriptor combinations both correctly classified 95.8% of ears. The ZPE descriptor combination performed better for "all MD" (ZPE, AUC-ROC 0.914; NSE, AUC-ROC 0.893).ConclusionParameter optimisation with NSE Real-IR influenced the optimal selection of MRI descriptors but did not improve their diagnostic performance in definite MD.Key PointsQuestionDelayed post-gadolinium ZPE (FLAIR) and NSE (REAL-IR) sequences are widely applied for diagnosing MD, but their relative benefits remain unclear.FindingsOptimised NSE sequences improve perilymphatic depiction and influence the selection of the optimal MRI descriptors, but do not improve diagnostic performance.Clinical relevanceRadiologists may continue to apply either ZPE or NSE sequences since they offer similar diagnostic abilities, but the choice of the sequence will influence which MRI features should be evaluated to support the diagnosis of MD.Key PointsQuestionDelayed post-gadolinium ZPE (FLAIR) and NSE (REAL-IR) sequences are widely applied for diagnosing MD, but their relative benefits remain unclear.FindingsOptimised NSE sequences improve perilymphatic depiction and influence the selection of the optimal MRI descriptors, but do not improve diagnostic performance.Clinical relevanceRadiologists may continue to apply either ZPE or NSE sequences since they offer similar diagnostic abilities, but the choice of the sequence will influence which MRI features should be evaluated to support the diagnosis of MD.Key PointsQuestionDelayed post-gadolinium ZPE (FLAIR) and NSE (REAL-IR) sequences are widely applied for diagnosing MD, but their relative benefits remain unclear.FindingsOptimised NSE sequences improve perilymphatic depiction and influence the selection of the optimal MRI descriptors, but do not improve diagnostic performance.Clinical relevanceRadiologists may continue to apply either ZPE or NSE sequences since they offer similar diagnostic abilities, but the choice of the sequence will influence which MRI features should be evaluated to support the diagnosis of MD.