This retrospective study was performed at the radiological department of the Fourth Hospital of Hebei Medical University between Oct 2020 and Oct 2022. The methods and patients enrolled in the optimization of the diagnostic performance of DCE-MRI with GRASP sequence have referred to previously published articles [4, 13]. Inclusion criteria: (1) Outpatients who did not have history of radiotherapy and chemotherapy; (2) received DCE-MRI scans of the upper abdomen with free breathing; (3) received extracellular contrast agent; and (4) fasted for four hours before the examination. Exclusion criteria: (1) Patients with chronic liver diseases (cirrhosis, fatty liver, alcoholic liver, etc.) and (2) Patients with inconsistent sequence parameters. A total of 64 consecutive patients were included in this study before (16 men, 14 women, age: 54.9 ± 17.0) and after (18 men, 16 women, age: 58.6 ± 12.6) the optimization of the acquiring parameters. This study was approved by the institutional review board, with no requirement for individual informed consent.

The GRASP technique used in this study combined a continuously acquired radial k-space trajectory with golden-angle sampling (~111.25°) and sparse parallel reconstruction employing compressed sensing, which offers simultaneous high spatial and high temporal resolution as well as motion robustness to DCE-MRI [14, 15]. DCE-MRI of abdomen with the GRASP sequence was performed at a 3T MR scanner (MAGNETOM Vida, Siemens Healthcare, Erlangen, Germany) with all patients in supine position. The acquired images were divided into two groups (before vs after optimization). The optimization scheme follows two principles: (1) reduce the impact on images from unpredictable and irregulate motions during examination and (2) adjust the sequence parameters to increase the number of radial views in each partition. Patients in optimization group were informed to remain calm and try to breathe easily before examination. To reduce the streak artifacts, arms were moved overhead with support. If applicable, a modest weight of sandbag would be placed upon patients’ abdomen to reduce the motion status in breathing. In order to increase the number of radial views, the scanning parameters of GRASP sequence were adjusted based on the following rules: (1) extend the total scanning time; (2) increase the slice thickness; (3) reduce the number of slices; (4) reduce the oversampling; and (5) shorten the time of repetition (TR). Details of the sequence parameters before and after optimization are presented in Table 1. In our institution, pre-contrast MRI was obtained for 23 seconds followed by GRASP scan during free breathing. For dynamic MRI, the conventional exam protocol consisted of a transversal T1-weighted (T1w) dixon breath-holding (BH) sequence (slice thickness = 3.5 mm, 15 s), a transversal fat-saturated (FS) T2w turbo spin echo (TSE) sequences (slice thickness = 6 mm, 2 min 20 s), a transversal diffusion-weighted imaging (DWI) (slice thickness = 6 mm, 2 min 32 s), GRASP sequence (6 min 1 s), and a post-contrast transversal T1w FS-BH sequence (slice thickness = 3.5 mm, 15 s). We optimized the sequence to transversal T1w dixon BH sequence, GRASP sequence, transversal T2w FS-TSE sequence, transversal DWI sequence, and post-contrast transversal T1w FS-BH sequence. In the optimized scheme, the DCE-MRI using the GRASP sequence was obtained continuously during free breathing for 5 minutes and 58 seconds (over 2500 radial spokes). The scan was performed during simultaneous injection of a standard dose of gadoteric acid (0.1 mmol/kg) administered at a rate of 2.5 mL/s.

Two radiologists (XX and XX) manually selected the best structured images at the unenhanced phase, early arterial phase, and late arterial phase among the reconstructed series for both groups of patients. The abdominal aorta was identified after contrast arrival when it had the highest signal intensity, whereas the portal vein was most clearly visible during the portal venous phase. Respectively, for each phase, three slices which best identified the structures of the main portal vein, left branch, and right branch of the hepatic portal vein were selected. Hence, 9 datasets from each patient were reviewed by readers. For the assessment of image quality, signal-to-noise ratio (SNR) of the left lobe and right lobe of the liver, defined as the ratio of the liver signal intensity (SIliver) and the background standard deviation (SD), were calculated as SNR = SI/SD. The contrast-to-noise ratio (CNR) of left lobe and right lobe of the liver, defined as the ratio of the absolute difference between SIliver and spleen signal intensity (SIspleen) to the background standard deviation, were calculated as CNR= (SIliverr−SIspleen)/SD [16]. Without awareness of the scanning parameters, one radiologist (8-years experiences in interpreting gynecological cancers) assessed MR images regarding the severity of radial artifact, the degree of image sharpness and a visual scoring of image quality with a 5-point scale. The quality index was defined as follows: 1 (Extremely poor), 2 (Poor), 3 (Fair), 4 (Good), and 5 (Excellent), where indices 1 and 2 are considered clinically unacceptable and 3 to 5 clinically acceptable [17].

Statistical analyses were performed using SPSS (Version 22; IBM, United States). All measurement values were tested for normal distribution using the Kolmogorov–Smirnov test. Continuous variables were presented as mean ± SD or median (range) depending on the normality of the data. Image quality parameters were compared using the independent t test or Mann–Whitney U test between the two groups. A 2-sided P value below 0.05 was considered as statistically significant.