Table 3 Segmentation Statistics for CDS and FL on Validation Sets

From: An international study presenting a federated learning AI platform for pediatric brain tumors

Metric

DSCs for CDS

DSCs for FL

IoU score for CDS

IoU score for FL

EP

0.8378 ± 0.0842

0.8044 ± 0.1051

0.7291 ± 0.1144

0.6848 ± 0.1380

DIPG

0.8866 ± 0.0544

0.8622 ± 0.0861

0.8003 ± 0.0826

0.7663 ± 0.1129

MB

0.7782 ± 0.2275

0.7800 ± 0.2151

0.6782 ± 0.2261

0.6775 ± 0.2207

PA

0.8233 ± 0.2126

0.8194 ± 0.2087

0.7386 ± 0.2204

0.7312 ± 0.2139

  1. The table compares the segmentation performance between CDS and FL on validation sets. A two-sided t-test for the DSC distributions with t-statistic (degrees of freedom), p value, effect size, and the 95% confidence interval for each class: 1) EP: 4.874 (54), 9.998e-06, 0.5, [0.01, 0.04], 2) DIPG: 3.511 (115), 0.0006, 0.37, [0.01,0.04], MB: −0.284 (146), 0.7771, 0.06, [−0.01,0.02], PA: 0.513 (89), 0.6090, 0.22, [0.00, 0.02].
Back to article page