Utility of Single-Shot Fast Spin-Echo T2-Weighted MRI With Deep Learning Reconstruction for Assessing Myometrial Invasion in Endometrial Cancer.
Authors
Affiliations (1)
Affiliations (1)
- Department of Radiology, The University of Osaka Graduate School of Medicine, Osaka, Japan.
Abstract
To evaluate the clinical feasibility of single-shot fast spin-echo T2-weighted MRI with deep learning reconstruction (SSFSE-DL-T2WI) for assessing myometrial invasion in endometrial cancer, compared with standard T2WI. This retrospective study included 72 patients with endometrial cancer who underwent preoperative contrast-enhanced (CE) MRI. Standard T2WI was acquired using a para-axial fast spin-echo (FSE) sequence and a para-sagittal periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) sequence (mean acquisition time: 383 s). Para-axial and para-sagittal SSFSE-DL-T2WI were obtained with a mean acquisition time of 47 seconds. Signal-to-noise ratio (SNR) and tumor contrast were measured by one radiologist, whereas image quality was assessed by 2 radiologists. Comparisons between standard T2WI and SSFSE-DL-T2WI were performed using paired t tests or Wilcoxon signed-rank tests. Myometrial invasion was evaluated using T2WI alone and in combination with CE-T1WI. Diagnostic performance was evaluated using pathologic myometrial invasion in the surgical specimens as the reference standard and was compared between the 2 T2WI methods using receiver operating characteristic (ROC) analysis. SSFSE-DL-T2WI demonstrated significantly higher SNR but lower tumor-to-junctional zone contrast compared with FSE-T2WI (P<0.01). No significant differences in SNR or contrast were observed between SSFSE-DL-T2WI and PROPELLER-T2WI. Both readers rated SSFSE-DL-T2WI significantly better than standard T2WI with respect to noise, artifacts, and tumor visibility (P<0.01). The areas under the ROC curve (AUCs) for detecting any and deep myometrial invasion did not differ significantly between SSFSE-DL-T2WI (mean: 0.77 and 0.86, respectively) and standard T2WI (mean: 0.72 and 0.80, respectively). Similarly, when combined with CE-T1WI, AUCs were comparable between SSFSE-DL-T2WI (mean: 0.83 and 0.91, respectively) and standard T2WI (mean: 0.80 and 0.89, respectively). SSFSE-DL-T2WI provides superior image quality and comparable diagnostic accuracy for assessing myometrial invasion in endometrial cancer, while requiring substantially shorter acquisition time compared with standard T2WI.