Table 3 Segmentation performance evaluated for different liver functions, based on the Dice Similarity Coefficient.

From: Automatic segmentation of liver structures in multi-phase MRI using variants of nnU-Net and Swin UNETR

 

standard nnU-Net6

ResEnc nnU-Net21

Swin UNETR22

Normal liver function

DSC

Liver parenchyma

0.98 [0.98, 0.98]

0.98 [0.97, 0.98]

0.97 [0.97, 0.98]

Portal vein

0.87 [0.84, 0.89]

0.86 [0.84, 0.89]

0.77 [0.73, 0.81]

Hepatic veins

0.81 [0.79, 0.84]

0.81 [0.79, 0.84]

0.75 [0.70, 0.79]

Lesions

0.56 [0.41, 0.72]

0.44 [0.26, 0.63]

0.47 [0.31, 0.63]

Ascites

–

–

–

Abdominal aorta

0.96 [0.95, 0.97]

0.96 [0.95, 0.97]

0.94 [0.92, 0.96]

Thoracic aorta

0.91 [0.87, 0.95]

0.91 [0.87, 0.96]

0.85 [0.78, 0.92]

Limited hepatic impairment

Liver parenchyma

0.97 [0.96, 0.98]

0.97 [0.96, 0.98]

0.95 [0.94, 0.97]

Portal vein

0.86 [0.85, 0.87]

0.85 [0.84, 0.87]

0.76 [0.72, 0.81]

Hepatic veins

0.80 [0.78, 0.82]

0.79 [0.76, 0.81]

0.69 [0.64, 0.73]

Lesions

0.56 [0.44, 0.69]

0.54 [0.41, 0.68]

0.49 [0.39, 0.60]

Ascites

0.35 [0.00, 0.81]

0.31 [0.00, 0.77]

0.17 [0.00, 0.44]

Abdominal aorta

0.96 [0.96, 0.97]

0.96 [0.96, 0.97]

0.95 [0.94, 0.96]

Thoracic aorta

0.93 [0.89, 0.96]

0.92 [0.89, 0.95]

0.87 [0.81, 0.94]

Significant hepatic impairment

Liver parenchyma

0.97 [0.96, 0.98]

0.96 [0.95, 0.98]

0.95 [0.93, 0.97]

Portal vein

0.73 [0.56, 0.89]

0.71 [0.54, 0.88]

0.63 [0.46, 0.81]

Hepatic veins

0.72 [0.67, 0.76]

0.72 [0.68, 0.77]

0.46 [0.31, 0.60]

Lesions

0.57 [0.37, 0.77]

0.53 [0.33, 0.74]

0.42 [0.22, 0.62]

Ascites

0.68 [0.42, 0.89]

0.54 [0.35, 0.73]

0.57 [0.29, 0.84]

Abdominal aorta

0.97 [0.96, 0.98]

0.97 [0.96, 0.97]

0.95 [0.93, 0.96]

Thoracic aorta

0.95 [0.94, 0.96]

0.95 [0.94, 0.96]

0.88 [0.82, 0.94]

  1. The results are demonstrated for the standard nnU-Net6, the ResEnc nnU-Net21 and the Swin UNETR22. Best results are written in bold.
Back to article page