Open Access

Table 6

Metrics over the test set obtained by the ensembles, grouping the individual models by AA (a) and by encoder (b).

Combination Dice Precision Recall
All base 0.8973 0.8631 0.9537
All visual 0.9026 0.8857 0.9401
All sloan 0.8997 0.8737 0.9479
All color 0.9006 0.8790 0.9438
All mass 0.9097 0.8941 0.9424
All 0.9054 0.8823 0.9484

(a)
Combination Dice Precision Recall
All ResNet18 0.9029 0.8823 0.9437
All ResNet50 0.9008 0.8781 0.9441
All EfficientNetB1 0.8985 0.8674 0.9515
All EfficientNetB2 0.8951 0.8677 0.9467
All EfficientNetB6 0.8972 0.8725 0.9445
All DenseNet161 0.9049 0.8927 0.9353
All DenseNet201 0.9035 0.8829 0.9432

(b)

Notes. In Table 6a, each row collects the results of the ensemble formed by the seven models (one for each encoder) that have been trained with the corresponding AA. Last row shows the ensemble made up of the 35 trained models. In Table 6b, each row shows the results of the ensemble formed by the five models (one for each AA) that have been trained with the same encoder. Color coding corresponds to the one used in Table 1.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.