Skip to main content
Insights into Imaging

Table 1 Detailed performance of the Res-Unet model in the validation and internal test dataset

From: Multi-channel deep learning model-based myocardial spatial–temporal morphology feature on cardiac MRI cine images diagnoses the cause of LVH

 

Cine MRI

Validation dataset (N = 48)

Internal test dataset (N = 63)

sequence

Dice (Mean ± SD)

HD (mm, Mean ± SD)

Dice (Mean ± SD)

HD (mm, Mean ± SD)

Per-slice level

2CH

0.934 ± 0.033

2.919 ± 1.740

0.921 ± 0.124

3.111 ± 2.506

4CH

0.933 ± 0.044

2.975 ± 2.254

0.944 ± 0.037

2.441 ± 2.009

SAX

0.941 ± 0.040

2.470 ± 2.046

0.944 ± 0.043

2.087 ± 1.141

Per-case level

2CH

0.935 ± 0.028

3.735 ± 1.476

0.921 ± 0.122

4.021 ± 2.077

4CH

0.934 ± 0.040

4.170 ± 2.286

0.945 ± 0.031

3.531 ± 1.670

SAX

0.941 ± 0.031

4.510 ± 5.357

0.945 ± 0.031

2.884 ± 1.322

  1. 2CH—two-chamber, 4CH—four-chamber, SAX—short axis, HD—Hausdorff distance, SD—standard deviation
Back to article page