Table 3 The mean segmentation accuracy of the model by aneurysm size in the external validation set.

From: Automated detection of intracranial aneurysms using skeleton-based 3D patches, semantic segmentation, and auxiliary classification for overcoming data imbalance in brain TOF-MRA

Aneurysm Size

3D U-Net

3D U-Net with auxiliary loss

nnU-Net

 

1:1

1:2

1:3

1:4

1:5

 < 5 mm

0.778**

0.859*

0.885

0.875

0.870

0.871

0.798**

 < 5–10 mm < 

0.769**

0.859*

0.884

0.874

0.873*

0.875

0.787**

10 mm < 

0.768**

0.861

0.881

0.874

0.873

0.873*

0.788**

  1. For comparisons between groups, t-tests were performed based on a 1:2 ratio, with p-values less than 0.05 denoted by ** and p-values less than 0.5 denoted by *to indicate statistical significance. Results with p-values less than 0.05 are considered statistically significant.
Back to article page