Synopsis

Keywords: Digestive, Image Reconstruction

Higher resolution and better image quality of 3D FIESTA sequence is used for evaluating bowel inflammation in CD; however, we found 2D SSFSE_DLwith significantly higher SNR and CNR of ileum and terminal ileum than 3D FIESTA and, of course, conventional 2D SSFSE. 2D SSFSE_DL shortened scan time and also offered better image quality than 3D FIESTA in assessment of bowel inflammation in CD. The SSFSE sequence alone has been considered sufficient in lesion detection[2]; SSFSE with DLR has great potential in elevation of detection rate without extra imaging time.

Introduction and Purpose

Crohn's disease（CD） is a chronic inflammatory disease that can affect any portion of the gastrointestinal tract. Endoscopy is commonly used to diagnose CD as the reference standard while magnetic resonance imaging has the advantage of being non-invasive and able to investigate small bowel involvement that cannot be reached by endoscopy. Previous studies have shown that fast imaging employing steady state acquisition(FIESTA) sequence was superior to single-shot fast spin echo(SSFSE) sequence in showing linear ulcers and pseudopolyps[1]. However, the scanning time of three-dimensional FIESTA(3D FIESTA)sequence is longer than that of two-dimensional SSFSE(2D SSFSE) sequence but provides higher resolution and better image quality. Deep learning reconstruction (DLR) is excellent at noise reduction without additional acquisition and post-processing time and has recently been introduced for clinical use. Our study aimed to explore the feasibility of 2D SSFSE with DLR (SSFSE_DL) for evaluating bowel inflammation in CD using 3D FIESTA sequence as reference.

Materials and Methods

Patients：Forty-seven patients with histological diagnosis of CD confirmed by endoscopy (36 male/11 female, age: 14–56 years) were prospectively recruited in this study. Imaging：All patients fasted for 6–8 hours prior to the scheduled exam time, and ingested 2000 mL of oral contrast agent (water with polyethylene glycol electrolytes powder dissolved) over 1 hour before magnetic resonance enterography (MRE) scan. All MRE scans were performed on a 3.0 T clinical MRI scanner (Signa Architect, GE Healthcare). Parameters of 3D FIESTA and 2D SSFSE sequences were shown in Table 1. 2D SSFSE_DL and 2D SSFSE without DLR (SSFSE_O) were automatically reconstructed with commercial DL-based reconstruction algorithm (AIRTM Recon DL). Assessment: Two radiologists made diagnosis in consensus and independently drew regions of interest on the bowel-segment of the suspected lesions to calculate signal-to-noise-ratio (SNR) and contrast to-noise-ratio (CNR). Representative MR images are shown in Figure 1. Statistical Tests: According to data normality and equality of variance, paired Wilcoxon signed-rank tests or paired t-test were performed to evaluate the statistical significance of differences in SNR and CNR between 2D SSFSE_DL and SSFSE_O, and between 2D SSFSE_DL and 3D FIESTA. P value <0.05 was considered to be statistically significant.

Results

SSFSE_DL showed significantly higher SNR in jejunum (125.24±106.84 vs. 69.41±42.79, P=0.026), ileum (167.87±110.42 vs. 98.18±49.51, P=0.001) and terminal ileum (190.34±80.70 vs. 123.24±48.10, P＜0.0001) than SSFSE_O as well as in ileum (94.85±45.88, P=0.02) and terminal ileum (88.81±33.57, P<0.0001) than 3D FIESTA. SSFSE_DL showed significantly higher CNR in jejunum (123.64±107.00 vs. 68.08±42.87, P=0.027), ileum (166.24±110.61 vs. 96.73±49.67, P=0.001) and terminal ileum (188.60±80.71 vs. 121.85±48.15, P＜0.0001) than SSFSE_O as well as ileum(93.29±45.87, P=0.02) and in terminal ileum (87.16±33.43, P<0.0001) than 3D FIESTA. There was no significant difference of SNR and CNR in jejunum between 2D SSFSE_DL and 3D FIESTA (P = 0.477) (Figure 2).

Discussion and conclusion

2D SSFSE_DL showed overall better image quality, has shorter scanning time and consistent diagnosis performance on assessing bowel inflammation in CD with 3D FIESTA. The previous research showed that SSFSE has superior lesion detection rate and imaging ability on motion target to DWI. The SSFSE sequence alone might be sufficient in lesion detection[2]. With DLR SSFSE images, inter-observer agreement on the lesion detection rate elevated. The inline DLR SSFSE shortened scan time and increase patient compliance.

Acknowledgements

No acknowledgement found.