Predictive Accuracy of MRI in Differentiation of Cystic Sellar Masses

Introduction: The spectrum of sellar and parasellar region pathology includes a wide range of cystic sellar masses (CSMs), including cystic pituitary adenoma (PA), Rathke’s cleft cysts (RCC), craniopharyngiomas, arachnoid cysts, and xanthogranulomas. CSMs are notable for their propensity for recurrence and are traditionally challenging to distinguish using neuroimaging studies alone. A reliable method to establish a sound differential diagnosis for the treatment of CSMs is critical, as surgical planning and approach selection is highly dependent on pathology. Few studies have assessed the accuracy of preoperative neuroimaging diagnosis of CSMs. We aimed to describe the predictive accuracy of preoperative neuroimaging in diagnosing CSMs.

Methods: Forty-three patients who underwent endoscopic transsphenoidal approaches for CSMs with pathological diagnosis between 2011 and 2018 were included. Preoperative MRIs were retrospectively reviewed by three neuroradiologists blinded to pathological diagnosis and subsequently compared with final pathological diagnosis. The neuroradiologists were asked to provide a single best diagnosis for each case. Subjective neuroradiologist case-by-case commentary was also obtained.

Results: Final pathological diagnoses included 25 RCCs, 15 cystic PAs, 2 arachnoid cysts, and 1 xanthogranuloma; average maximal tumor diameter for these masses was 16.0, 22.1, 26, and 15 mm, respectively. The overall accuracy of predicting the correct diagnosis was 66.7%, with individual accuracy being 62.8, 62.8, and 74.4%. Accuracy by pathology was as follows: RCC (78.7%), cystic PA (57.8%), arachnoid cyst (0%), and xanthogranuloma (0%). At least two out of three neuroradiologists correctly diagnosed 92.0% of RCCs, compared with 53.3% of cystic PAs. Subjective neuroradiology assessment typically narrowed the final diagnosis down to two final choices on differential diagnosis.

Fig. 1 Representative Rathke’s cleft cyst.

Fig. 2 Representative pituitary adenoma.

Fig. 3 Representative arachnoid cyst.

Fig. 4 Single xanthogranuloma case in this series.

Conclusion: Accurate pathology of CSMs remains challenging based on preoperative neuroimaging. RCCs are somewhat easier to predict than cystic PA and rarer CSM subtypes such as arachnoid cyst or xanthogranuloma. Generalizability of these results is limited by small sample sizes and potential interrater differences. Future research by our group will focus on identifying reliable imaging characteristics to accurately differentiate CSMs on MRI.

No conflict of interest has been declared by the author(s).