Few-shot electrocardiogram (ECG) signal classification method based on improved siamese network

1. A few-shot electrocardiogram (ECG) signal classification method based on an improved Siamese network, comprising the following steps:

a) acquiring n original ECG signals to form an original ECG signal set D, D={(x 1 , y 1 ), (x 2 , y 2 ), . . . , (x i , y i ), . . . , (x n , y n )}, wherein x i denotes an i-th original ECG signal, and y i denotes a class label corresponding to the i-th original ECG signal x i , i∈{1, . . . , n};

b) preprocessing the original ECG signal set D to remove noise in the n original ECG signals, thereby acquiring a clean ECG signal set D′, D′={(x′ 1 , y 1 ), (x′ 2 , y 2 ), . . . , (x′ i , y i ), . . . , (x′ n , y n )}, wherein x′ i denotes an i-th clean ECG signal;

c) normalizing the i-th clean ECG signal x′ i to acquire a normalized ECG signal x″ i ; and performing zero-padding in an end of a sequence of the normalized ECG signal x″ i if a length of the sequence of the normalized ECG signal x″ i is less than L max , wherein the length of the sequence of the normalized ECG signal x″ i is equal to L max , and a normalized ECG signal set D″ is acquired, D″={(x″ 1 , y 1 ), (x″ 2 , y 2 ), . . . , (x″ i , y i ), . . . , (x″ n , y n )};

d) creating a sample pair set P based on the normalized ECG signal set D″,

P

=

{

(

(

x

1

″

,

x

2

″

)

,

Y

′

)

,

(

(

x

2

″

,

x

3

″

)

,

Y

′

)

,

…

,

(

(

x

i

-

1

″

,

x

i

″

)

,

Y

′

)

,

(

(

x

i

″

,

x

i

+

1

″

)

,

Y

′

)

,

…

,

(

(

x

n

-

2

″

,

x

n

-

1

″

)

,

Y

′

)

,

(

(

x

n

-

1

″

,

x

n

″

)

,

Y

′

)

}

,

wherein

Y

′

=

{

1

y

i

-

1

=

y

i

0

y

i

-

1

≠

y

i

;

y i−1 denotes a class label corresponding to an (i−1)-th original ECG signal x i−1 ; and there are M sample pairs in the sample pair set P,

M

=

n

×

(

n

-

1

)

2

;

e) constructing a few-shot classification model, and inputting a sample pair ((x′ i , x′ i+1 ),Y′) from the sample pair set P into the few-shot classification model to acquire a similarity score E w (x″ i , x″ i+1 );

f) training, by an adaptive moment estimation (Adam) optimizer, the few-shot classification model through a loss function L to acquire an optimized few-shot classification model;

g) randomly sampling K ECG signals from each of N classes in a Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) dataset to form a support set S support , S support ={(s 1 , a 1 ), (s 2 , a 2 ), . . . , (s i , a i ), . . . , (s NK , a NK )}, wherein S; denotes an i-th ECG signal, and a i denotes a class label corresponding to the i-th ECG signal s i , i∈{1, . . . , NK};

h) randomly sampling Q ECG signals from each of the N classes in the MIT-BIH dataset to form a set S query , S query ={(q 1 , b 1 ), (q 2 , b 2 ), . . . , (q i , b i ), . . . , (q NQ , b NQ )}, wherein q i denotes an i-th ECG signal, and b i denotes a class label corresponding to the i-th ECG signal q i , i∈{1, . . . , NQ};

i) replacing the i-th original ECG signal x i with the i-th ECG signal s i , and repeating the steps b) and c) to acquire an i-th normalized ECG signal s″ i , wherein a normalized support set S″ support is acquired, S″ support ={(s″ 1 , a 1 ), (s″ 2 , a 2 ), . . . , (s″ i , a i ), . . . , (s″ NK , a NK )}; and replacing the i-th original ECG signal x i with the i-th ECG signal q i , and repeating the steps b) and c) to acquire an i-th normalized ECG signal q″ i , a normalized query set S″ query is acquired, S″ query ={(q″ 1 , b 1 ), (q″ 2 , b 2 ), . . . , (q″ i , b i ), . . . , (q″ NQ , b NQ )}; and

j) inputting the i-th normalized ECG signal s″ i and the i-th normalized ECG signal q″ i into the optimized few-shot classification model to acquire a classification result.

2. The few-shot ECG signal classification method based on the improved Siamese network according to claim 1 , wherein the step a) comprises: acquiring the n original ECG signals from a University of California Riverside (UCR) dataset.

3. The few-shot ECG signal classification method based on the improved Siamese network according to claim 1 , wherein the step b) comprises: denoising, by a first median filter and a second median filter in sequence, the i-th original ECG signal x i to acquire the i-th clean ECG signal x′ i .

4. The few-shot ECG signal classification method based on the improved Siamese network according to claim 3 , wherein the first median filter has a width of 300 ms, and the second median filter has a width of 600 ms.

5. The few-shot ECG signal classification method based on the improved Siamese network according to claim 1 , wherein L max =187.

6. The few-shot ECG signal classification method based on the improved Siamese network according to claim 1 , wherein the step e) comprises:

e-1) constructing the few-shot classification model, comprising an embedding module and a metric module;

e-2) constructing the embedding module of the few-shot classification model, wherein the embedding module comprises a Siamese network formed by a first CMP module and a second CMP module; the first CMP module comprises a convolutional layer, a first rectified linear unit (ReLU) activation function layer, a primary capsule layer of a capsule network, a digital capsule layer of the capsule network, a first fully connected layer, a second ReLU activation function layer, and a second fully connected layer; and the second CMP module comprises a convolutional layer, a first ReLU activation function layer, a primary capsule layer of a capsule network, a digital capsule layer of the capsule network, a first fully connected layer, a second ReLU activation function layer, and a second fully connected layer;

e-3) inputting the i-th normalized ECG signal x″ i into the convolutional layer and the first ReLU activation function layer of the first CMP module in sequence to acquire a feature f 1 1 ; inputting the feature f 1 1 into the primary capsule layer of the capsule network in the first CMP module to acquire a vector f 1 2 ; inputting the vector f 1 2 into the digital capsule layer of the capsule network in the first CMP module to acquire a feature f 1 3 ; inputting the feature f 1 3 into the first fully connected layer and the second ReLU activation function layer of the first CMP module in sequence to acquire a feature f 1 4 ; and inputting the feature f 1 4 into the second fully connected layer of the first CMP module to acquire a feature f(x″ i );

e-4) inputting an (i+1)-th normalized ECG signal x″ i+1 into the convolutional layer and the first ReLU activation function layer of the first CMP module in sequence to acquire a feature f 2 1 ; inputting the feature f 2 1 into the primary capsule layer of the capsule network in the first CMP module to acquire a vector f 2 2 ; inputting the vector f 2 2 into the digital capsule layer of the capsule network in the first CMP module to acquire a feature inputting the feature f 2 3 into the first fully connected layer and the second ReLU activation function layer of the first CMP module in sequence to acquire a feature f 2 4 ; and inputting the feature f 2 4 into the second fully connected layer of the first CMP module to acquire a feature f(x″ i+1 ); and

e-5) inputting the feature f(x″ i ) and the feature f(x″ i+1 ) into the metric module of the few-shot classification model, and calculating the similarity score E w (x″ 1 , x″ i+1 ) by E w (x″ i , x″ i+1 )=∥f(x″ i )−f(x″ i+1 )∥, wherein ∥⋅∥ denotes a Euclidean distance (ED) calculation.

7. The few-shot ECG signal classification method based on the improved Siamese network according to claim 6 , wherein in the step e-2), the convolutional layer of the first CMP module comprises a 3×3 convolution kernel, and the convolutional layer of the second CMP module comprises a 3×3 convolution kernel.

8. The few-shot ECG signal classification method based on the improved Siamese network according to claim 6 , wherein the step j) comprises:

j-1) inputting the i-th normalized ECG signal s″ i of a u-th class into the convolutional layer and the first ReLU activation function layer of the first CMP module in sequence to acquire a feature f 3 1 , u∈{1, . . . , N}; inputting the feature f 3 1 into the primary capsule layer of the capsule network in the first CMP module to acquire a vector f 3 2 ; inputting the vector f 3 2 into the primary capsule layer of the capsule network in the first CMP module to acquire a feature f 3 3 ; inputting the feature f 3 3 into the first fully connected layer and the second ReLU activation function layer of the first CMP module in sequence to acquire a feature f 3 4 ; inputting the feature f 3 4 into the second fully connected layer of the first CMP module to acquire a feature f(s″ i ) u ; and calculating, by a mean( ) function in Python, an average of all K features f(s″ 1 ) u , f(s″ 2 ) u , . . . , f(s″ i ) u , . . . , f(s″ K ) u , of the u-th class to acquire a feature vector μ u ;

j-2) inputting the i-th normalized ECG signal q″ i into the convolutional layer and the first ReLU activation function layer of the first CMP module in sequence to acquire a feature f 4 1 ; inputting the feature f 4 1 into the primary capsule layer of the capsule network in the first CMP module to acquire a vector f 4 2 ; inputting the vector f 4 2 into the primary capsule layer of the capsule network in the first CMP module to acquire a feature f 4 3 ; inputting the feature f 4 3 into the first fully connected layer and the second ReLU activation function layer of the first CMP module in sequence to acquire a feature f 4 4 ; and inputting the feature f 4 4 into the second fully connected layer of the first CMP module to acquire a feature f(q″ i );

j-3) inputting the feature vector μ u and the feature f(q″ i ) into the metric module of the few-shot classification model, and calculating the similarity score E w (μ u , f(q″ i )) by E w (μ u , f(q″ i ))=∥μ u −f(q″ i )∥; and

j-4) calculating a class label ŷ i of the i-th normalized ECG signal q″ i by ŷ i =arg max {E w (μ 1 , f(q″ i )), E w (μ 2 , f(q″ i )), . . . , E w (μ u , f(q″ i )), . . . , E w (μ N , f(q″ i ))}, and combining class labels of all NQ normalized ECG signals to form the classification result.

9. The few-shot ECG signal classification method based on the improved Siamese network according to claim 1 , wherein the step f) comprises: calculating the loss function L by, L=L 1 +αL 2 , wherein

L

1

=

Y

′

⁢

1

2

⁢

(

E

w

(

x

i

″

,

x

i

+

1

″

)

)

2

+

(

1

-

Y

′

)

⁢

{

max

⁡

(

0

,

m

-

E

w

(

x

i

″

,

x

i

+

1

″

)

)

}

2

;

m denotes a hyperparameter, α denotes a hyperparameter; and L 2 denotes a cross entropy loss function.