Personalized LSTM Models for ECG Lead Transformations Led to Fewer Diagnostic Errors Than Generalized Models: Deriving 12-Lead ECG from Lead II, V2, and V6
Background and Objective: The prevalence of chronic cardiovascular diseases (CVDs) has risen globally, nearly doubling from 1990 to 2019. ECG is a simple, non-invasive measurement that can help identify CVDs at an early and treatable stage. A multi-lead ECG, up to 15 leads in a wearable form factor,...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-01-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/23/3/1389 |
| _version_ | 1797623192699797504 |
|---|---|
| author | Prashanth Shyam Kumar Mouli Ramasamy Kamala Ramya Kallur Pratyush Rai Vijay K. Varadan |
| author_facet | Prashanth Shyam Kumar Mouli Ramasamy Kamala Ramya Kallur Pratyush Rai Vijay K. Varadan |
| author_sort | Prashanth Shyam Kumar |
| collection | DOAJ |
| description | Background and Objective: The prevalence of chronic cardiovascular diseases (CVDs) has risen globally, nearly doubling from 1990 to 2019. ECG is a simple, non-invasive measurement that can help identify CVDs at an early and treatable stage. A multi-lead ECG, up to 15 leads in a wearable form factor, is desirable. We seek to derive multiple ECG leads from a select subset of leads so that the number of electrodes can be reduced in line with a patient-friendly wearable device. We further compare personalized derivations to generalized derivations. Methods: Long-Short Term Memory (LSTM) networks using Lead II, V2, and V6 as input are trained to obtain generalized models using Bayesian Optimization for hyperparameter tuning for all patients and personalized models for each patient by applying transfer learning to the generalized models. We compare quantitatively using error metrics Root Mean Square Error (RMSE), R<sup>2</sup>, and Pearson correlation (ρ). We compare qualitatively by matching ECG interpretations of board-certified cardiologists. Results: ECG interpretations from personalized models, when corrected for an intra-observer variance, were identical to the original ECGs, whereas generalized models led to errors. Mean performance values for generalized and personalized models were (RMSE-74.31 µV, R<sup>2</sup>-72.05, ρ-0.88) and (RMSE-26.27 µV, R<sup>2</sup>-96.38, ρ-0.98), respectively. Conclusions: Diagnostic accuracy based on derived ECG is the most critical validation of ECG derivation methods. Personalized transformation should be sought to derive ECGs. Performing a personalized calibration step to wearable ECG systems and LSTM networks could yield ambulatory 15-lead ECGs with accuracy comparable to clinical ECGs. |
| first_indexed | 2024-03-11T09:25:15Z |
| format | Article |
| id | doaj.art-6c93c7f151fd424db8e517f5dd7fff82 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-11T09:25:15Z |
| publishDate | 2023-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-6c93c7f151fd424db8e517f5dd7fff822023-11-16T18:00:20ZengMDPI AGSensors1424-82202023-01-01233138910.3390/s23031389Personalized LSTM Models for ECG Lead Transformations Led to Fewer Diagnostic Errors Than Generalized Models: Deriving 12-Lead ECG from Lead II, V2, and V6Prashanth Shyam Kumar0Mouli Ramasamy1Kamala Ramya Kallur2Pratyush Rai3Vijay K. Varadan4The Department of Engineering Science and Mechanics, The Pennsylvania State University, 212 Earth-Engineering Sciences Bldg, University Park, PA 16802, USAThe Department of Engineering Science and Mechanics, The Pennsylvania State University, 212 Earth-Engineering Sciences Bldg, University Park, PA 16802, USAGeisinger Medical Center, 100 North Academy Avenue, Danville, PA 17822, USAThe Department of Biomedical Engineering, The University of Arkansas, 4183 Bell Engineering Center, Fayetteville, AR 72701, USAThe Department of Engineering Science and Mechanics, The Pennsylvania State University, 212 Earth-Engineering Sciences Bldg, University Park, PA 16802, USABackground and Objective: The prevalence of chronic cardiovascular diseases (CVDs) has risen globally, nearly doubling from 1990 to 2019. ECG is a simple, non-invasive measurement that can help identify CVDs at an early and treatable stage. A multi-lead ECG, up to 15 leads in a wearable form factor, is desirable. We seek to derive multiple ECG leads from a select subset of leads so that the number of electrodes can be reduced in line with a patient-friendly wearable device. We further compare personalized derivations to generalized derivations. Methods: Long-Short Term Memory (LSTM) networks using Lead II, V2, and V6 as input are trained to obtain generalized models using Bayesian Optimization for hyperparameter tuning for all patients and personalized models for each patient by applying transfer learning to the generalized models. We compare quantitatively using error metrics Root Mean Square Error (RMSE), R<sup>2</sup>, and Pearson correlation (ρ). We compare qualitatively by matching ECG interpretations of board-certified cardiologists. Results: ECG interpretations from personalized models, when corrected for an intra-observer variance, were identical to the original ECGs, whereas generalized models led to errors. Mean performance values for generalized and personalized models were (RMSE-74.31 µV, R<sup>2</sup>-72.05, ρ-0.88) and (RMSE-26.27 µV, R<sup>2</sup>-96.38, ρ-0.98), respectively. Conclusions: Diagnostic accuracy based on derived ECG is the most critical validation of ECG derivation methods. Personalized transformation should be sought to derive ECGs. Performing a personalized calibration step to wearable ECG systems and LSTM networks could yield ambulatory 15-lead ECGs with accuracy comparable to clinical ECGs.https://www.mdpi.com/1424-8220/23/3/1389ECGLSTM networksBayesian Optimizationpersonalized medicinewearable devices |
| spellingShingle | Prashanth Shyam Kumar Mouli Ramasamy Kamala Ramya Kallur Pratyush Rai Vijay K. Varadan Personalized LSTM Models for ECG Lead Transformations Led to Fewer Diagnostic Errors Than Generalized Models: Deriving 12-Lead ECG from Lead II, V2, and V6 Sensors ECG LSTM networks Bayesian Optimization personalized medicine wearable devices |
| title | Personalized LSTM Models for ECG Lead Transformations Led to Fewer Diagnostic Errors Than Generalized Models: Deriving 12-Lead ECG from Lead II, V2, and V6 |
| title_full | Personalized LSTM Models for ECG Lead Transformations Led to Fewer Diagnostic Errors Than Generalized Models: Deriving 12-Lead ECG from Lead II, V2, and V6 |
| title_fullStr | Personalized LSTM Models for ECG Lead Transformations Led to Fewer Diagnostic Errors Than Generalized Models: Deriving 12-Lead ECG from Lead II, V2, and V6 |
| title_full_unstemmed | Personalized LSTM Models for ECG Lead Transformations Led to Fewer Diagnostic Errors Than Generalized Models: Deriving 12-Lead ECG from Lead II, V2, and V6 |
| title_short | Personalized LSTM Models for ECG Lead Transformations Led to Fewer Diagnostic Errors Than Generalized Models: Deriving 12-Lead ECG from Lead II, V2, and V6 |
| title_sort | personalized lstm models for ecg lead transformations led to fewer diagnostic errors than generalized models deriving 12 lead ecg from lead ii v2 and v6 |
| topic | ECG LSTM networks Bayesian Optimization personalized medicine wearable devices |
| url | https://www.mdpi.com/1424-8220/23/3/1389 |
| work_keys_str_mv | AT prashanthshyamkumar personalizedlstmmodelsforecgleadtransformationsledtofewerdiagnosticerrorsthangeneralizedmodelsderiving12leadecgfromleadiiv2andv6 AT mouliramasamy personalizedlstmmodelsforecgleadtransformationsledtofewerdiagnosticerrorsthangeneralizedmodelsderiving12leadecgfromleadiiv2andv6 AT kamalaramyakallur personalizedlstmmodelsforecgleadtransformationsledtofewerdiagnosticerrorsthangeneralizedmodelsderiving12leadecgfromleadiiv2andv6 AT pratyushrai personalizedlstmmodelsforecgleadtransformationsledtofewerdiagnosticerrorsthangeneralizedmodelsderiving12leadecgfromleadiiv2andv6 AT vijaykvaradan personalizedlstmmodelsforecgleadtransformationsledtofewerdiagnosticerrorsthangeneralizedmodelsderiving12leadecgfromleadiiv2andv6 |