Diagnostic Performance of Fast Brain MRI Compared with Routine Clinical MRI in Patients with Glioma Grades 3 and 4: A Pilot Study.

Abstract

Background and purpose: EPIMix is a fast brain MRI technique not previously investigated in patients with grade 3 and 4 gliomas. This pilot study aimed to investigate the diagnostic performance of EPIMix in the radiological treatment evaluation of adult patients with grade 3 and 4 gliomas compared with routine clinical MRI (rcMRI).

Materials and methods: Patients with grade 3 and 4 gliomas investigated with rcMRI and EPIMix were retrospectively included in the study. Three readers (R1-R3) participated in the radiological assessment applying the Response Assessment for Neuro-Oncology (RANO 2.0) criteria, of whom two (R1 and R2) independently evaluated EPIMix and later rcMRI by measuring contrast-enhancing and non-contrast-enhancing tumor regions at each follow-up. For cases with discrepant evaluations, an unblinded side-by-side (EPIMix and rcMRI) reading was performed together with a third reader (R3). Comparisons between methods (EPIMix versus rcMRI) were performed using the weighted Cohen κ. The sensitivity and specificity to progressive disease (PD) on a follow-up scan were calculated for EPIMix compared with rcMRI with receiver operating characteristic curves (ROC) to assess the area under the curve (AUC).

Results: Of 35 patients (mean age, 53 years; 31% women), a total of 93 MRIs encompassing 58 follow-up investigations showed PD at a blinded reading in 33% of EPIMix (19/58, R1-2), while in 31% (18/58 exams, R1), and 34% (20/58 exams, R2) of rcMRI. An almost perfect agreement for tumor category assessment was found between EPIMix and rcMRI (EPIMixR1 versus rcMRIR1 κ = 0.96; EPIMixR2 versus rcMRIR2 κ = 0.89). The sensitivity for EPIMix to detect PD was 1.00 (0.81-1.00) for R1 and 0.90 (0.68-0.99) for R2, while the specificity was 0.97 (0.86-1.00) for R1 and R2. The AUC for PD was 0.99 for R1 (EPIMixR1 versus rcMRIR1) and 0.94 for R2 (EPIMixR2 versus rcMRIR2). The P value of the DeLong test AUCR1 versus AUCR2 was P = .20 (R1-R2).

Conclusions: In this pilot study, EPIMix was used as a fast MRI alternative for treatment evaluation of patients with glioma grades 3 and 4, with high but slightly lower diagnostic performance than rcMRI.