An active learning approach to train a deep learning algorithm for tumor segmentation from brain MR images

Insights into Imaging

Table 2 Metric results for each model iteration selected for the active learning process

Model	Active learning starting point	Sensitivity	Positive predictive value	Dice similarity coefficient	Jaccard similarity coefficient	Modified Hausdorff distance
Model B AL	23 epochs	0.872	0.859	0.854	0.757	4.045
Model C AL1	23 epochs	0.851	0.803	0.813	0.702	4.413
Model C AL2	38 epochs	0.869	0.815	0.830	0.725	4.215
Model C AL3	50 epochs	0.876	0.829	0.841	0.741	4.067