Objective:To explore the application value of deep learning reconstruction(DLR)of magnetic resonance(MR)on the image quality and lesion diagnostic efficacy of three-dimensional liver acceleration volume acquisition(3D-LAVA)sequences.Methods:A retrospective analysis was conducted on the image data of magnetic resonance imaging(MRI)of 45 patients who underwent MR contrast-enhanced examinations on abdomen at West China Hospital of Sichuan University from June 2023 to September 2023.For the venous phase of the LAVA sequence,conventional reconstruction and DLR were performed respectively to obtain conventional LAVA images and DLR-LAVA images at the venous phase.Two radiologists conducted subjective evaluation for the overall image quality,image clarity,artifacts,and diagnostic confidence of the two groups of images by using a 5-point scale.And then,the signal-to-noise ratio(SNR)of the images of liver,spleen,and lesion,as well as the contrast-to-noise ratio(CNR)of the liver,spleen to fat,muscle,and lesions were objectively evaluated.The two radiologists measured respectively the maximum and minimum diameters of the largest layer of the largest lesion,and compared the diagnostic efficacy of conventional LAVA and DLR-LAVA for lesions.The diagnostic efficacies of conventional LAVA and DLR-LAVA were respectively assessed by using the receiver operating characteristic(ROC)curve.Results:The consistency of the two radiologists'subjective evaluations for the image quality of the conventional LAVA and DLR-LAVA sequences was moderate and above moderate,with Kappa values ranging from 0.553 to 0.902,and P<0.001.Both radiologists found that the subjective image quality,image clarity,artifacts,and diagnostic confidence of the DLR-LAVA sequence were higher than those of the conventional LAVA sequence(Z Physician 1=4.24,5.49,3.50,4.47,Z Physician 2=4.15,3.12,3.77,4.26,P<0.05).The SNR values of the liver,spleen,and lesions in the DLR-LAVA images were all higher than those in the conventional LAVA sequence(t=-20.45,-18.58,-5.51,P<0.001).The CNR values between liver and fat,between liver and muscle,between liver and spleen,between spleen and fat,between spleen and muscle,and between lesion and liver in DLR-LAVA sequence were higher than those in the conventional LAVA sequence(t=-20.49,-19.94,-18.05,-16.74,P<0.001).There were no statistically significant differences in the measured values of the maximum and minimum diameters of the largest lesions between the conventional LAVA and DLR-LAVA sequences(P>0.05).The area under curve(AUC)values of physician 1 who used respectively the conventional LAVA and DLR-LAVA sequences were 0.897 and 0.916 in diagnosing lesions,while those of physician 2 were respectively 0.851 and 0.890.Conclusion:DLR can improve the image quality of the LAVA sequence for liver,and has a higher diagnostic value for lesions.

Objective:To explore the application value of deep learning reconstruction(DLR)of magnetic resonance(MR)on the image quality and lesion diagnostic efficacy of three-dimensional liver acceleration volume acquisition(3D-LAVA)sequences.Methods:A retrospective analysis was conducted on the image data of magnetic resonance imaging(MRI)of 45 patients who underwent MR contrast-enhanced examinations on abdomen at West China Hospital of Sichuan University from June 2023 to September 2023.For the venous phase of the LAVA sequence,conventional reconstruction and DLR were performed respectively to obtain conventional LAVA images and DLR-LAVA images at the venous phase.Two radiologists conducted subjective evaluation for the overall image quality,image clarity,artifacts,and diagnostic confidence of the two groups of images by using a 5-point scale.And then,the signal-to-noise ratio(SNR)of the images of liver,spleen,and lesion,as well as the contrast-to-noise ratio(CNR)of the liver,spleen to fat,muscle,and lesions were objectively evaluated.The two radiologists measured respectively the maximum and minimum diameters of the largest layer of the largest lesion,and compared the diagnostic efficacy of conventional LAVA and DLR-LAVA for lesions.The diagnostic efficacies of conventional LAVA and DLR-LAVA were respectively assessed by using the receiver operating characteristic(ROC)curve.Results:The consistency of the two radiologists'subjective evaluations for the image quality of the conventional LAVA and DLR-LAVA sequences was moderate and above moderate,with Kappa values ranging from 0.553 to 0.902,and P<0.001.Both radiologists found that the subjective image quality,image clarity,artifacts,and diagnostic confidence of the DLR-LAVA sequence were higher than those of the conventional LAVA sequence(Z Physician 1=4.24,5.49,3.50,4.47,Z Physician 2=4.15,3.12,3.77,4.26,P<0.05).The SNR values of the liver,spleen,and lesions in the DLR-LAVA images were all higher than those in the conventional LAVA sequence(t=-20.45,-18.58,-5.51,P<0.001).The CNR values between liver and fat,between liver and muscle,between liver and spleen,between spleen and fat,between spleen and muscle,and between lesion and liver in DLR-LAVA sequence were higher than those in the conventional LAVA sequence(t=-20.49,-19.94,-18.05,-16.74,P<0.001).There were no statistically significant differences in the measured values of the maximum and minimum diameters of the largest lesions between the conventional LAVA and DLR-LAVA sequences(P>0.05).The area under curve(AUC)values of physician 1 who used respectively the conventional LAVA and DLR-LAVA sequences were 0.897 and 0.916 in diagnosing lesions,while those of physician 2 were respectively 0.851 and 0.890.Conclusion:DLR can improve the image quality of the LAVA sequence for liver,and has a higher diagnostic value for lesions.