Skip to main content

Table 7 Classification performance versus model complexity

From: An adaptive multi-modal hybrid model for classifying thyroid nodules by combining ultrasound and infrared thermal images

Modality

Model

F1

F2

Paramsa

FLOPsb

Single(US)

ResNet

0.6945

0.8140

11.17 Mc

1.96 Gd

Single(IRT)

ResNet

0.6438

0.7618

11.17 M

1.96 G

Single(US)

ViT

0.6889

0.7854

85.65 M

16.86 G

Single(IRT)

ViT

0.7191

0.7441

85.65 M

16.86 G

Two(US+IRT)

ResNet w/ AMCE

0.9358

0.9428

87.58 M

16.47 G

Two(US+IRT)

ViT w/ AMCE

0.9446

0.9383

236.54 M

46.28 G

Two(US+IRT)

AmmH(Hybrid w/ AMCE)

0.9717

0.9738

121.30 M

23.77 G

  1. a The number of parameters that need to be trained during the model training, which is used to measure the space complexity of a model.
  2. b The number of floating-point operations, which is used to measure the time complexity of a model.
  3. c M: Millions.
  4. d G: 10\(^9\)
Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us