Accelerated Reduced Field of View T2-Weighted Imaging of Pancreaticobiliary Disorders Using Deep Learning-Based Reconstruction: Reduction of Acquisition Time and Improvement of Image Quality.

BACKGROUND: Reduced field-of-view (rFOV) T2WI improves in-plane spatial resolution. Deep learning-based reconstruction (DLR) has emerged as enhancing image quality. We compared examination time, image quality, and lesion detection rates between pancreaticobiliary rFOV T2WI with and without DLR. METHODS: 198 patients who underwent pancreaticobiliary rFOV T2WI were included. The protocol included rFOV T2WI with 2 NEX (rFOV T2WIN2) and 1 NEX (rFOV T2WIN1). DLR was applied to generate corresponding datasets: rFOV T2WIN2-DLR and rFOV T2WIN1-DLR datasets. Three observers evaluated the noise, respiratory motion artifacts (RMA), overall image quality (OIQ), and diagnostic confidence (DC). Two observers measured the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR). Image quality metrics were compared using either ANOVA or Friedman test. Lesion detection rates were tested using the Cochran's Q test and McNemar test. RESULTS: The noise, RMA, OIQ, and DC scores of rFOV T2WIN1-DLR were notably higher than those of rFOV T2WIN1. The noise, OIQ, and DC scores of rFOV T2WIN2 were notably higher than those of rFOV T2WIN1 (P < .001). The SNR and CNR of rFOV T2WI with DLR were notably higher than those of rFOV T2WI without DLR (P < .05). The rFOV T2WIN1-DLR sequence yielded a higher lesion detection rate (92.5%; 491/531) compared to rFOV T2WIN1 (76.5%; 406/531). CONCLUSIONS: Pancreaticobiliary rFOV T2WI with DLR is feasible and yields superior image quality compared to rFOV T2WI without DLR. A 37.1% to 74.4% reduction in acquisition time is achievable without increasing image noise or compromising overall image quality and lesion detection rate. RELEVANCE STATEMENT: The perfect balance of image quality, scanning time, and lesion detection rate can be achieved by using DLR in pancreaticobiliary rFOV T2WI.