Diagnostic accuracy of neuroimaging in pediatric optic chiasm/sellar/suprasellar tumors

Abstract

Preoperative diagnosis for tumors arising in the optic chiasm/sellar/suprasellar region in children is helpful to determine surgical necessity and approach, given the high operative risk in this area. We evaluated the ability to differentiate tumor type by preoperative neuroimaging. Thirty‐eight of 53 tumors were correctly diagnosed by neuroimaging based on final pathologic diagnosis (prediction accuracy 72%). Prediction accuracies were 87% (20/23) for craniopharyngioma, 79% (11/14) for optic pathway glioma, 64% (7/11) for germ cell tumor, and 0% (0/5) for Langerhans cell histiocytosis. Diagnosis of optic chiasm/sellar/suprasellar tumors in children by imaging alone should be considered when biopsy is considered high risk.

Robust deep learning classification of adamantinomatous craniopharyngioma from limited preoperative radiographic images

“…Our optimal models performed on par with the average of human specialists ( ), although ‘Radiologist A’ consistently outperformed our models across the board. As mentioned previously, recent work reported an accuracy of ACP diagnosis of 87% by pediatric neuroradiologists using a complete imaging dataset and clinical history 10 . This performance corroborates the overall generalization capacity of the models presented herein.…”

“…Radiographically, ACP is characterized by heterogeneous solid tissue, cystic regions, and calcification 10 . Recent data indicate that ACP and other tumors of the sellar/suprasellar region may be accurately diagnosed using current radiographic techniques in 64–87% 10 of cases. This sets a high bar for machine-aided diagnoses, but also leaves room for clinically relevant improvement.…”

“…While ACP is the most common sellar/suprasellar pediatric tumor, it is an uncommon entity, representing 2 to 5% of all pediatric brain tumors, with an incidence of approximately 1.9 per million patient-years 10 . In order to facilitate research into this tumor, Advancing Treatment for Pediatric Craniopharyngioma was formed in 2015.…”

Novel Deep Learning Methodology for Automated Classification of Adamantinomatous Craniopharyngioma Using a Small Radiographic Dataset

“…Norris, et al previously demonstrated human experts in a similar scenario achieved a prediction accuracy of 0.87. 10…”

Coexistence of Suprasellar Lesion and Pulmonary Congestion: Two Rare Cases and Literature Review

“…The overwhelmingly lesion in sellar/suprasellar region is pituitary adenomas, however this region can be affected by a variety of non-pituitary lesions, including craniopharyngiomas, cystic nonneoplastic lesions, inflammatory lesions, and so on [1,2]. Given that various diseases occurring in this region, as well as high operative risks, therefore arriving at the right preoperative diagnosis is important for taking appropriate therapeutic decisions, for instance, selecting surgical approach and necessity, ameliorating surgical complications [3]. Implementation of anesthesia to patients suffering sellar/suprasellar lesions is a challenge to every anesthesiologist, due to these lesions frequently accompanying with various neurologic and endocrine dysfunction, depending on properties of lesion, and mass effect on adjacent structures [4].…”