Dynamic Contrast-Enhanced pituitary MRI with higher temporal resolution using deep learning Reconstruction: A feasibility study

초록

Objective To assess the image quality and diagnostic performance of pituitary MRI with higher temporal resolution using deep learning–based reconstruction dynamic contrast-enhanced imaging (DLR-DCE) compared with conventional dynamic contrast-enhanced imaging (C-DCE) for evaluation of sellar lesions. Methods In this retrospective study, 43 consecutive patients underwent a DCE protocol comprising six sequential phases of DLR-DCE followed immediately by a single phase of C-DCE between April and August 2022. DLR-DCE reduced scan time (40 to 27 s per phase) and gap (0.4 to 0 mm), while increasing slices per phase (6 to 8). Two neuroradiologists independently evaluated image quality, diagnosed sellar lesions, and assessed cavernous sinus invasion using the last phase of DLR-DCE and subsequent C-DCE images. Results DLR-DCE demonstrated significantly improved delineation of the normal gland and cavernous segment of the oculomotor nerve for both readers (all P ≤ 0.034). Lesion conspicuity was improved for one reader ( P = 0.001). DLR-DCE yielded significantly higher lesion and pituitary gland SNR values, pituitary gland–to–lesion CNR value than C-DCE in both readers ( P ≤ 0.002). The diagnostic accuracy for cavernous sinus invasion was numerically higher with DLR-DCE; however, the difference was not significant for either reader (all P = 0.250). The diagnostic performance for sellar lesions was comparable between the two techniques (all P ≥ 0.05). Conclusion In routine clinical practice, DLR-DCE pituitary MRI achieves higher temporal resolution and improved spatial sampling without compromising image quality or diagnostic performance compared with C-DCE. © 2026 Elsevier B.V.

키워드