	US 12,279,875 B2
	Electrocardiogram (ECG) signal detection and positioning method based on weakly supervised learning
Minglei Shu, Jinan (CN); Pengyao Xu, Jinan (CN); Shuwang Zhou, Jinan (CN); Zhaoyang Liu, Jinan (CN); and Kaiwei Li, Jinan (CN)
Assigned to QILU UNIVERSITY OF TECHNOLOGY (SHANDONG ACADEMY OF SCIENCES), Jinan (CN); and SHANDONG COMPUTER SCIENCE CENTER (NATIONAL SUPERCOMPUTING CENTER IN JINAN), Jinan (CN)
Filed by Qilu University of Technology (Shandong Academy of Sciences), Jinan (CN); and SHANDONG COMPUTER SCIENCE CENTER (NATIONAL SUPERCOMPUTING CENTER IN JINAN), Jinan (CN)
Filed on Dec. 14, 2023, as Appl. No. 18/539,331.
Claims priority of application No. 2023104375867 (CN), filed on Apr. 23, 2023.
Prior Publication US 2024/0350066 A1, Oct. 24, 2024
Int. Cl. A61B 5/318 (2021.01); A61B 5/00 (2006.01); A61B 5/308 (2021.01); G06N 3/0895 (2023.01)

CPC A61B 5/318 (2021.01) [A61B 5/308 (2021.01); A61B 5/7203 (2013.01); G06N 3/0895 (2023.01)]

6 Claims

1. An electrocardiograph (ECG) detection and positioning method based on weakly supervised learning, comprising the following steps:

a-1) operating a plurality of ECG devices to obtain n pieces of ECG signal data of different lengths;

a-2) collecting the n pieces of ECG signal data of different lengths from the plurality of ECG devices to obtain an original ECG dataset X₀, wherein X₀={x₁, x₂, . . . , x_j, . . . , x_n}, x_jrepresents a j^thoriginal ECG signal record, and j∈{1, 2, . . . , n};

b) preprocessing the original ECG dataset X₀to obtain a preprocessed ECG dataset X₀′, wherein X₀′={x₁′, x₂′, . . . , x_j′, . . . , x_n′}, and x_j′ represents a j^thpreprocessed original ECG signal record;

c) establishing a deep learning model, and inputting the preprocessed ECG dataset into the deep learning model to obtain a globally recorded category prediction value y^preand a predicted abnormal signal location set L; and

d) training the deep learning model;

wherein the step c) comprises the following substeps:

c-1) constituting the deep learning model by a multi-scale feature extraction module, a self-attention encoding module, and a classification and positioning module;

c-2) constituting the multi-scale feature extraction module of the deep learning model by a start convolutional block, a first multi-scale feature extraction block, and a second multi-scale feature extraction block, wherein the start convolutional block of the multi-scale feature extraction module is sequentially constituted by a first start convolutional layer, a batch normalization (BN) layer, a PRELU activation function layer, and a second start convolutional layer; and inputting the j^thpreprocessed original ECG signal record x_j′ into the start convolutional block of the multi-scale feature extraction module to obtain a feature map X₁;

c-3) constituting the first multi-scale feature extraction block of the multi-scale feature extraction module by a first BN layer, a first PRELU activation function layer, a channel local attention (CLA) layer, a first convolutional layer, a second BN layer, a second PRELU activation function layer, a dropout layer, a second convolutional layer, and a maximum pooling layer, wherein the CLA layer is constituted by a first branch and a second branch, the first branch of the CLA layer is sequentially constituted by a global average pooling layer, a first fully connected layer, a Swish activation function layer, a second fully connected layer, and a Sigmoid activation function layer, and the second branch of the CLA layer is sequentially constituted by a global maximum pooling layer, a maximum pooling layer, a first fully connected layer, a Swish activation function layer, a second fully connected layer, a Sigmoid activation function layer, and an upsampling layer; sequentially inputting the feature map X₁into the first BN layer and the first PRELU activation function layer of the first multi-scale feature extraction block to obtain a feature map X_{1_1}; inputting the feature map X_{1_1}into the first branch of the CLA layer of the first multi-scale feature extraction block to obtain a feature map X_{1_1}^1_2; inputting the feature map X_{1_1}into the second branch of the CLA layer of the first multi-scale feature extraction block to obtain a feature map X_{1_1}^2_3; multiplying the feature map X_{1_1}and the feature map X_{1_1}^1_2to obtain a first product, and then multiplying the first product and the feature map X_{1_1}^2_3to obtain a feature map X_{1_2}; sequentially inputting the feature map X_{1_2}into the first convolutional layer, the second BN layer, the second PRELU activation function layer, the dropout layer, and the second convolutional layer of the first multi-scale feature extraction block to obtain a feature map X_{1_3}; and adding up the feature map X_{1_3}and the feature map X₁to obtain a first addition result, and then inputting the first addition result into the maximum pooling layer of the first multi-scale feature extraction block to obtain a feature map X₂;

c-4) constituting the second multi-scale feature extraction block of the multi-scale feature extraction module by a first BN layer, a first PRELU activation function layer, a CLA layer, a first convolutional layer, a second BN layer, a second PRELU activation function layer, a dropout layer, a second convolutional layer, and a maximum pooling layer; sequentially inputting the feature map X₂into the first BN layer and the first PRELU activation function layer of the second multi-scale feature extraction block to obtain a feature map X_{2_1}; inputting the feature map X_{2_1}into a first branch of the CLA layer of the second multi-scale feature extraction block to obtain a feature map X_{2_1}^1_2; inputting the feature map X_{2_1}into a second branch of the CLA layer of the second multi-scale feature extraction block to obtain a feature map X_{2_1}^2_3; multiplying the feature map X_{2_1}and the feature map X_{2_1}^1_2to obtain a second product, and then multiplying the second product and the feature map X_{2_1}^2_3to obtain a feature map X_{2_2}; sequentially inputting the feature map X_{2_2}into the first convolutional layer, the second BN layer, the second PRELU activation function layer, the dropout layer, and the second convolutional layer of the second multi-scale feature extraction block to obtain a feature map X_{2_3}; and adding up the feature map X_{2_3}and the feature map X₂to obtain a second addition result, and then inputting the second addition result into the maximum pooling layer of the second multi-scale feature extraction block to obtain a feature map X₃;

c-5) encoding the feature map X₃based on a predefined location to obtain a location information graph X₃^P, and concatenating the feature map X₃and the location information graph X₃^Pto obtain a feature map X₄;

c-6) constituting the self-attention encoding module of the deep learning model by a multi-head self-attention layer, a layer norm layer, and a multi-layer perceptron (MLP) layer, and inputting the feature map X₄into the self-attention encoding module to obtain a feature map X₅;

c-7) constituting the classification and positioning module by a local classification module, a fusion module, and an abnormal location output module;

c-8) constituting the local classification module of the classification and positioning module sequentially by a convolutional layer, a flatten layer, a fully connected layer, and a Softmax function; and inputting the feature map X₅into the local classification module to obtain a predicted value y_c^preof a local location, wherein y_c^pre={y₁^pre, y₂^pre, . . . , y_l^pre, . . . y_l^pre}, y_l^prerepresents a predicted value output by the fully connected layer for an l^thlocation, l∈{1, 2, . . . , m}, m represents a pre quantity of predicted values output by the fully connected layer, y_l^pre=0 indicates a normal ECG signal, and y_l^pre=1 indicates an abnormal ECG signal;

c-9) inputting the predicted value y_c^preof the local location into the fusion module of the classification and positioning module to obtain the globally recorded category prediction value y^pre; and

c-10) inputting the predicted value y_c^preof the local location into the abnormal location output module of the classification and positioning module, wherein the abnormal location output module determines whether each location in the predicted value y_c^preof the local location has an abnormal signal, wherein the abnormal signal reflects an activity state of a heart, and outputs the abnormal signal location set L, wherein L={l, l₂, . . . , l, . . . , l_k}, l_irepresents coordinates of a location of an i^thabnormal signal in the predicted value y_c^preof the local location, i∈{1, 2, . . . , k}, and k represents a quantity of predicted abnormal signals;

wherein in the step c-9), the globally recorded category prediction value y^preis calculated according to a formula y^pre=(1/b²)log((1/m)Σ_l=1^mexp(by_l^pre)), wherein b is a constant.