Table 4 The performance of the proposed method compared to previous methods on the MIT-BIH dataset.

From: A novel hybrid CNN-transformer model for arrhythmia detection without R-peak identification using stockwell transform

Methods

Acc(%)

N

S

V

Q

F

F1

Se

PPV

F1

Se

PPV

F1

Se

PPV

F1

Se

PPV

F1

Se

PPV

Lin et al.33

99.2

.962

.973

.951

.908

.905

.911

.982

.985

.979

–

–

–

.991

.983

1.00

Oliveira et al.34

95.3

.974

.971

.978

.916

.761

.566

.950

.930

.950

–

–

–

–

–

–

Chen et al.35

93.1

.969

.984

.954

.334

.295

.384

.773

.708

.851

–

–

–

–

–

–

Kung et al.36

98.6

–

–

–

.815

.754

.887

.970

.967

.974

–

–

–

–

–

–

Ince et al.37

98.3

–

–

–

.581

.635

.537

.860

.846

.874

–

–

–

–

–

–

Shi et al.38

94.2

.971

.953

.989

.627

.907

.479

.885

.929

.845

–

–

–

–

–

–

Xie et al.39

96.5

.900

.943

.877

.848

.797

.906

.966

.971

.962

–

–

–

.816

.908

.741

Zhai et al.40

96.1

.899

.879

.920

.754

.768

.740

.931

.938

.924

–

–

–

.700

.624

.796

Feiyan et al.32

99.6

.999

.999

.999

.973

.980

.966

.990

.991

.989

–

–

–

.924

.907

.942

Proposed

99.6

.997

.988

.991

.997

.997

.997

.993

.994

.991

1.00

1.00

1.00

1.00

1.00

1.00

  1. Each class (N, S, V, Q, F) includes F1-score, Sensitivity (Se), and PPV. Significant values are in bold.
Back to article page