Application of artificial intelligence-assisted compressed sensing combined with three types of fat-suppressed T₂WI techniques

Abstract

Objective To explore the advantages of artificial intelligence-assisted compressed sensing （ACS） combined with three types of fat-suppressed T2WI techniques in image quality， scanning time and scheme selection. Methods In 30 patients with low back pain who underwent conventional lumbar magnetic resonance imaging （MRI） plain scan， ACS combined with 5 groups of sagittal （SAG） fat-suppressed T2WI sequences were added to original examination sequence. Frequency-selective fat saturation （FS）， water-fat separation （WFI） and short tau inversion recovery （STIR） techniques were employed. ACS-SAG-T2WI-FS （group A， n=30）， Acs-SAG-T2WI-WFI （group B， n=30）， ACS-SAG-STIR （group C n=30）， SAG-T2WI-FS （group D， n=30）， SAG-T2WI-WFI （group E， n=30） and SAG-STIR sequences （group F， n=30） were employed. Objective evaluation indicators of signal-to-noise ratio （SNR） and contrast-to-noise ratio （CNR） and subjective evaluation by two radiologists were carried out to assess the effect. Results The SNR and CNR of the fourth vertebra， the fourth and fifth intervertebral discs and the same spinal cord in group A were higher than those in group D. The SNR and CNR of the fourth vertebra， the fourth and fifth intervertebral discs， and the same spinal cord in group B were higher than those in group E. The SNR and CNR of the fourth vertebra， the fourth and fifth intervertebral discs， and the same spinal cord in group C were higher compared with those in group F （all P < 0.01）. The imaging time in group A was shortened by 13.2% compared with that in group D. The imaging time in group B was reduced by 8.9% compared with that in group E. The imaging time in group C was reduced by 12.4% compared with that in group F （all P < 0.01）. The subjective scores by two radiologists for groups A， B and C were significantly higher than those in groups D， E and F （all P < 0.01；The scores of the two physicians were consistent， Kappa=0.972， P <0.01）. Conclusions The combination of ACS and three types of fat-suppressed T2WI techniques （FS， WFI and STIR） is superior to use of fat-suppressed T2WI techniques alone. ACS-SAG-T2WI-FS sequence is recommended with the shortest imaging time and the highest image quality. ACS-SAG-T2WI-WFI sequence provides multiple groups of phase images， and the time advantage difference is the second choice. ACS-SAG-STIR sequence has the most stable fat-suppression capability， which can be used as the last option.

Keywords

• artificial intelligence-assisted compressed sensing technique|water-fat separation technique|frequency selective saturation method|short inversion time