To investigate the application of artificial intelligence-assisted compressed sensing (Acs) technology in the upper abdominal fat-suppressed T2-weighted imaging (T2WI) sequence.

30 patients underwent MRI examination of the upper abdomen in the Third Affiliated Hospital of Sun Yat-sen University from June 2022 to October 2022 were recruited in this study. The informed consents of all patients were obtained and the local ethical committee approval was received. Among them, 21 patients were male and 9 female, aged 31-76 years with a median age of 55 years. All patients received MRI by conventional parallel imaging (PI) and Acs sequences, respectively. Axial position, gating and triggering, and frequency-selective saturation techniques (AX-T2WI-FS-RT) were adopted in the PI group, and Acs technique (AX-T2WI-FS-BH-Acs) was employed in the Acs group. Objective assessment indexes included image signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR) and scanning time, etc. Comparison between two groups was performed by paired t test. The image quality score and the number of detected lesions between two groups were compared by Wilcoxon test. The consistency in the image quality scores between two physicians was evaluated by Kappa test.

The average SNR in the Acs group was 24.3±8.2, significantly higher than 11.7±4.4 in the PI group (t=13.00, P<0.05). The CNR in the Acs group was 4.2±2.3, significantly higher than 2.2±1.3 in the PI group (t=9.20, P<0.05). The scanning time in the Acs group was66 s, significantly shorter than 156 s in the PI group. The median image quality score in the Acs group was 4.4(4.3, 4.6), significantly higher than 4.1(3.4, 4.4) in the PI group (Z=3.98, P<0.05). For patients with respiratory disorders in the Acs group, the image quality score was 4.6(4.3, 4.7), significantly higher than 3.4(3.1, 3.4) in the PI group (Z=3.80, P<0.05). The number of detected lesions in two groups was 1(0, 3) and 1(0, 2), and no significant difference was observed (Z=0.50, P>0.05). The consistency in the image quality score was high between two physicians (κ=0.96).

Compared with the AX-T2WI-FS-RT sequence of the upper abdomen in the PI group, Acs technique using T2WI-FS-BH-Acs sequence can significantly shorten the scanning time and improve the image quality, especially for patients with respiratory disorders, without sacrificing the detection rate of lesions.