Interrater Reliability and Accuracy of CISS versus Contrasted T1 Vibe for the Presence of Optic Canal Invasion in Tuberculum Sella Meningiomas

Background: A recent study has shown that constructive interference steady state (CISS) is better correlated with intraoperative cavernous sinus invasion than post contrast T1 volume-interpolated breath-hold examination (VIBE). However, we are unsure if that result can be extrapolated to other important anatomic locations, such as the optic canal. Our current study was to determine if CISS has better interrater agreement and is more accurate at predicting optic canal invasion than post contrast T1 VIBE.

Materials and Methods: Presurgical diagnostic MRI studies containing both CISS and post-contrast T1 sequences from 27 patients (54 optic canals) were analyzed retrospectively. Five neuroradiologists blinded to the radiology reports and surgical results were recruited. Each optic canal was divided into four visual quadrants and was evaluated at two anatomic locations, the tip of the anterior clinoid and the optic strut. Each quadrant was considered positive for tumor invasion if there is tumor invasion at either of the two anatomic locations. The reference standard at quadrant level was each patient's presurgical Humphrey visual test result. Each optic canal was considered positive for tumor invasion if any of the four quadrants at either of the two anatomic locations was positive. The reference standard at canal level was both presurgical Humphrey visual test result and intraoperative observation of gross optic canal invasion by tumor.

Results: There was good inter-rater agreement on the presence/absence of tumor involvement at a particular quadrant, 0.635 for CISS and 0.643 for post-contrast T1 (95% CI for difference: −0.086, 0.010). At the quadrant level with the Humphrey visual test as the standard reference, the readers’ mean sensitivity and specificity of predicting tumor invasion were 0.724 and 0.742 with CISS and 0.768 and 0.692 with postcontrast T1, respectively. At the canal level with the Humphrey visual test result as the standard reference, the readers’ mean sensitivity and specificity were 0.775 and 0.632 with CISS and 0.758 and 0.640 with postcontrast T1, respectively. At canal level with the intraoperative findings as the standard reference, the readers’ mean sensitivity and specificity were 0.643 and 0.438 with CISS and 0.643 and 0.454 with postcontrast T1, respectively. There is no statistical difference between the two MRI sequences at either level in terms of reader accuracy with either Humphrey visual test or intraoperative findings as the standard reference.

Conclusions: There was good interreader agreement on optic nerve tumor involvement for both CISS and postcontrast T1. Both CISS and postcontrast T1 sequences have good accuracy at predicting optic canal tumor invasion at both visual quadrants and optic canal levels without statistical difference in accuracy between the two MRI sequences.