Table 1 Average overall accuracy of different models

From: A deep learning based smartphone application for early detection of nasopharyngeal carcinoma using endoscopic images

Model

Fold1

Fold2

Fold3

Fold4

Fold5

Average overall accuracy (95% CI)

SwinT

0.9519

0.9547

0.9519

0.9539

0.9449

0.9515 (±0.0039) (0.9481–0.9549)

MaxViT

0.9350

0.9448

0.9358

0.9456

0.9284

0.9379 (±0.0072) (0.9316–0.9443)

PoolF

0.9484

0.9531

0.9503

0.9523

0.9473

0.9503 (±0.0025) (0.9481–0.9524)

CaiT

0.9373

0.9448

0.9440

0.9480

0.9387

0.9426 (±0.0045) (0.9387–0.9465)

ResNet

0.9231

0.9369

0.9239

0.9342

0.9221

0.9280 (±0.0070) (0.9219–0.9341)

DenseNet

0.9393

0.9421

0.9385

0.9452

0.9331

0.9396 (±0.0045) (0.9357–0.9436)

Xception

0.9460

0.9460

0.9456

0.9476

0.9347

0.9440 (±0.0052) (0.9394–0.9486)

ConvNeXt

0.9409

0.9468

0.9495

0.9515

0.9390

0.9455 (±0.0054) (0.9408–0.9503)

Back to article page