Table 29 Extended benchmarking with transformer-based and lightweight models.

From: A federated learning-based privacy-preserving image processing framework for brain tumor detection from CT scans

Model

Accuracy

Precision

F1-score

Specificity

Sensitivity

NPV

MCC

FPR

FNR

CNN (baseline)

0.912

0.8991

0.8905

0.9203

0.8707

0.9154

0.8644

0.0797

0.1293

SVM

0.8945

0.879

0.8653

0.9024

0.8412

0.9007

0.8388

0.0976

0.1588

dResU-Net

0.9367

0.9182

0.9121

0.9431

0.8998

0.934

0.896

0.0569

0.1002

EfficientNet-B3

0.9488

0.927

0.9199

0.9512

0.9107

0.9462

0.9104

0.0488

0.0893

MobileNetV3

0.9352

0.9064

0.9021

0.936

0.8891

0.9288

0.8917

0.064

0.1109

TransUNet

0.9541

0.9367

0.9262

0.9577

0.9193

0.9524

0.9241

0.0423

0.0807

SwinUNet

0.9608

0.9445

0.9368

0.9633

0.9284

0.96

0.9337

0.0367

0.0716

Proposed (HGBOA + ResNet + CapsuleNet)

0.9907

0.9854

0.9955

0.9879

0.9982

0.9985

0.9762

0.0159

0.0091

Back to article page