Table 4 Performance metrics of individual snapshots and snapshot ensembles for multi-labeled pediatric thoracic disease classification across three random seeds.

From: Integrating snapshot ensemble learning into masked autoencoders for efficient self-supervised pretraining in medical imaging

Seed number

Epochs

AUC

AUPRC

Sensitivity

Precision

F1-score

42

200

0.752

0.614

0.737

0.549

0.605

400

0.759

0.622

0.726

0.572

0.607

600

0.751

0.620

0.745

0.551

0.614

800

0.750

0.614

0.744

0.541

0.608

Snapshot ensemble

0.759

0.623

0.736

0.552

0.610

1004

200

0.747

0.610

0.764

0.541

0.617

400

0.757

0.621

0.726

0.564

0.607

600

0.761

0.625

0.725

0.573

0.603

800

0.761

0.624

0.708

0.568

0.594

Snapshot ensemble

0.765

0.627

0.739

0.572

0.616

2023

200

0.751

0.613

0.752

0.545

0.618

400

0.757

0.615

0.756

0.552

0.617

600

0.752

0.615

0.747

0.546

0.615

800

0.758

0.617

0.747

0.544

0.615

Snapshot ensemble

0.760

0.618

0.751

0.549

0.619

Mean of three ensembles

0.761 (0.003)

0.623 (0.005)

0.742 (0.008)

0.558 (0.013)

0.615 (0.005)