Skip to main content

Table 5 Lung Segmentation Accuracy for the Michigan Medicine Dataset (50% Held-Out) with U-Net Fine Tuning

From: Robust segmentation of lung in chest x-ray: applications in analysis of acute respiratory distress syndrome

 

Adult (n = 50)

Adult Severe ARDS (n = 13)

Pediatric (n = 50)

 

Dice (mean)

Dice (min)

Standard Deviation

Dice (mean)

Dice (min)

Standard Deviation

Dice (mean)

Dice (min)

Standard Deviation

TVAC

0.8608

0.7790

0.0439

0.8301

0.7614

0.0423

0.8467

0.7827

0.0364

U-Net

0.8953

0.4842

0.0975

0.8188

0.0001

0.2685

0.8829

0.4277

0.1403

Random Walker

0.7422

0.3898

0.1238

0.6711

0.3405

0.1448

0.6681

0.2692

0.1675

Active Spline

0.6475

0.2003

0.1743

0.6431

0.3410

0.1270

0.6203

0.2003

0.1814

  1. Data are mean with minimum and standard deviation reported for each algorithm on different patient populations. TVAC = Total Variation-based Active Contour, Dice = Sørensen–Dice coefficient, ARDS = acute respiratory distress syndrome
Back to article page

BMC Medical Imaging

ISSN: 1471-2342

Contact us