Minireviews
Copyright ©The Author(s) 2021.
World J Gastroenterol. Oct 28, 2021; 27(40): 6825-6843
Published online Oct 28, 2021. doi: 10.3748/wjg.v27.i40.6825
Table 1 Applications of artificial intelligence segmentation methods to liver magnetic resonance imaging
Ref.
Task
Method
MR image
DICE
Mole et al[37], 2020Segment liver from T1 mapping technique to aid surgical planning3D U-NetT1 map0.97
Winther et al[27], 2020Segment liver from Gd-EOB-DTPA-enhanced MRI for volume calculations3D U-NetGadoxetic acid-enhanced MRI0.96 ± 1.9
Liu et al[30], 2020Segment liver for automated liver iron quantification2D U-NetT2* map0.86 ± 0.01
Wang et al[43], 2019Segment Liver across multiple imaging modalities and techniques2D U-NetT1- and T2*- weightedT1-w: 0.95 ± 0.03
T2-w: 0.92 ± 0.05
Cunha et al[46], 2020Segment liver to classify if adequate contrast uptake has occurred in contrast enhanced scans2D U-NetPre- and post-contrast T1- weighted, and T2- weightedNot reported
Chen et al[31], 2020Segment multiple organs in abdominal scans, to aid radiotherapy planning2D Dense U-NetT1-weightedLiver: 0.96 ± 0.009
Bousabarah et al[36], 2020Segment and delineate HCCs2D U-NetGadoxetic acid-enhanced MRILiver: 0.91 ± 0.01
Tumour: 0.68 ± 0.03
Ivashchenko et al[41], 2019Segment liver, vasculature and biliary tree4D k-mean clusteringGadoxetic acid-enhanced MRILiver: 0.95 ± 0.01
Irving et al[44], 2017Segment liver with vessel exclusion to assist in liver assessment2D U-NetT1 map0.95
Yang et al[45], 2019Segment liver across multiple domains via domain transferCycle GAN and 2D U-NetGadoxetic acid-enhanced MRI0.891 ± 0.040
Christ et al[39], 2017Segment liver and tumours within, in CT and MRITwo sequential 2D U-NetsDiffusion-weightedLiver: 0.87
Tumour: 0.697
Fu et al[35], 2018Segment multiple organs in abdominal scans, to aid radiotherapy planningThree Dense CNNsT2/T1-weightedLiver: 0.953 ± 0.007
Valindria et al[33], 2018Segment multiple organs in multi-modal (MR,CT) scansResNet Encoder DecoderT2-weightedLiver: 0.914
Masoumi et al[42], 2012Segment the liverWatershed (non-AI) + ANNAbdominal MRI0.94 (IoU not DICE)
Jansen et al[40], 2019Segment liver and metastasesCNNDCE-MR and diffusion-weightedLiver: 0.95