Table 5 Accuracies of the classification on internal and external validation datasets, for 6 segmentations (4 automatic, 2 manual).

From: Prediction of therapeutic intensity level from automatic multiclass segmentation of traumatic brain injury lesions on CT-scans

Classification model

Applied segmentation

Internal validation accuracy

External validation accuracy

4-class classification model

CNN1 (BLAST-CT)

50% (3/6)

67% (8/12)

CNN2

67% (4/6)

67% (8/12)

Manual4

67% (4/6)

67% (8/12)

7-class classification model

CNN3

83% (5/6)

75% (9/12)

CNN4

83% (5/6)

83% (10/12)

Manual7

83% (5/6)

67% (8/12)