Explainable artificial intelligence framework for electrocardiography analysis

Patent metadata

Field	Value
Publication number	US-11324457-B2
Application number	US-201916253942-A
Country	US
Kind code	B2
Filing date	Jan 22, 2019
Priority date	Jan 22, 2019
Publication date	May 10, 2022
Grant date	May 10, 2022

Abstract

There is included an apparatus and system including an intra-heartbeat (HB) extraction code configured to extract intra-HB features from electrocardiography (ECG) signals, and an inter-HB extraction code configured to extract inter-HB features from the ECG signals, and at least one attention mechanism code configured to control at least one of the intra-HB extraction code and inter-HB extraction code based on at least one attention mechanism.

First claim

What is claimed is:

1. An apparatus comprising:
at least one memory configured to store computer program code;
at least one hardware processor configured to access said computer program code and operate as instructed by said computer program code, said computer program code including:
intra-heartbeat (HB) extraction code configured to cause the at least one hardware processor to extract intra-HB features from electrocardiography (ECG) signals;
inter-HB extraction code configured to cause the at least one hardware processor to extract inter-HB features from the ECG signals;
attention mechanism code configured to cause the at least one hardware processor to control the extraction of the inter-HB and intra-HB features based on at least one attention mechanism; and
ECG analysis code configured to cause the at least one hardware processor to obtain and statistically process the intra-HB features and the inter-HB features,

wherein the ECG analysis code is further configured to cause the at least one hardware processor to output at least an outlier alarm based on a result of statistically processing the intra-HB features and the inter-HB features, and
wherein the intra-HB extraction code is further configured to cause the at least one hardware processor to extract the intra-HB features in parallel with extraction of the inter-HB features.

2. The apparatus according to claim 1 , wherein the computer program code further includes extraction pool code configured to cause the at least one hardware processor to extract at least one extraction model from an extraction pool and apply the at least one extraction model to extraction of at least one of the intra-HB features and the inter-HB features by a corresponding one of the intra-HB extraction code and the inter-HB extraction code.

3. The apparatus according to claim 1 , wherein the computer program code further includes second inter-HB extraction code configured to cause the at least one processor to extract second inter-HB features from the ECG signals in parallel with both of the extraction of the intra-HB features and the extraction of the inter-HB features.

4. The apparatus according to claim 3 , wherein the at least one attention mechanism code is configured to cause the at least one hardware processor to control the intra-HB extraction based on the at least one attention mechanism, and
wherein the computer program code further includes:
second attention mechanism code configured to cause the at least one hardware processor to control the inter-HB extraction based on a second attention mechanism; and
third attention mechanism code configured to cause the at least one hardware processor to control the second inter-HB extraction based on a third attention mechanism.

5. The apparatus according to claim 1 , wherein the computer program code further includes task specific pool code configured to cause the at least one hardware processor to extract at least one task specific model from a task specific pool and apply the at least one task specific model to statistically process the intra-HB features and the inter-HB features.

6. The apparatus according to claim 5 , wherein the ECG analysis code is further configured to cause the at least one hardware processor to output further at least one of a classification result and a predicted diagnosis based on the result of statistically processing the intra-HB features and the inter-HB features.

7. The apparatus according to claim 6 , wherein statistically processing the intra-HB features and the inter-HB features comprises at least one of batch normalization and instance normalization based on the at least one task specific model.

8. The apparatus according to claim 7 , wherein computer program code further includes feedback link code configured to cause the at least one hardware processor to feedback an output of the ECG analysis code to the at least one attention mechanism, and
wherein the at least one attention mechanism code is configured to cause the at least one hardware processor to update the attention mechanism based on the output.

9. A method performed by at least one computer processor comprising:
extracting intra-HB features from electrocardiography (ECG) signals;
extracting, inter-HB features from the ECG signals;
controlling at least one of the intra-HB extraction and inter-HB extraction based on at least one attention mechanism;
obtaining and statistically processing the intra-HB features and the inter-HB features; and
outputting at least an outlier alarm based on a result of statistically processing the intra-HB features and the inter-HB features,
wherein extracting the intra-HB features is in parallel with extraction of the inter-HB features.

10. The method according to claim 9 , further comprising:
extracting at least one extraction model from an extraction pool and applying the at least one extraction model to extraction of at least one of the intra-HB features and the inter-HB features.

11. The method according to claim 9 , further comprising:
extracting second inter-HB features from the ECG signals in parallel with both of the extraction of the intra-HB features and the extraction of the inter-HB.

12. The method according to claim 11 , further comprising:
controlling the intra-HB extraction based on the at least one attention mechanism;
controlling the inter-HB extraction based a second attention mechanism; and
controlling the second inter-HB extraction based on a third attention mechanism.

13. The method according to claim 9 , further comprising:
extracting at least one task specific model from a task specific pool; and
applying the at least one task specific model to statistically process the intra-HB features and the inter-HB features.

14. The method according to claim 13 , further comprising:
outputting at least one of a classification result and a predicted diagnosis based on the result of statistically processing the intra-HB features and the inter-HB features.

15. The method according to claim 14 , wherein statistically processing the intra-HB features and the inter-HB features comprises at least one of batch normalization and instance normalization based on the at least one task specific model.

16. A non-transitory computer readable medium storing a program causing a computer to execute a process, the process comprising:
extracting intra-HB features from electrocardiography (ECG) signals;
extracting inter-HB features from the ECG signals;
controlling at least one of the intra-HB extraction and inter-HB extraction based on at least one attention mechanism;
obtaining and statistically processing the intra-HB features and the inter-HB features; and
outputting at least an outlier alarm based on a result of statistically processing the intra-HB features and the inter-HB features,
wherein extracting the intra-HB features is in parallel with extraction of the inter-HB features.

Assignees

• Tencent America LLC

Inventors

• Wang Kun
• Du Nan
• Xie Yusheng
• Tu min
• Zhang Shangqing
• Li Yaliang
• Yang Tao
• Tang Hui
• Han Lianyi
• Fan Wei

Classifications

• G06F2218/12

  Classification; Matching · CPC title

• G06F2218/08

  Feature extraction · CPC title

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06N3/045

  Combinations of networks · CPC title