Unsupervised deep learning of electrocardiograms enables scalable human disease profiling
Abstract The 12-lead electrocardiogram (ECG) is inexpensive and widely available. Whether conditions across the human disease landscape can be detected using the ECG is unclear. We developed a deep learning denoising autoencoder and systematically evaluated associations between ECG encodings and ~1,...
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-01-01
|
| Series: | npj Digital Medicine |
| Online Access: | https://doi.org/10.1038/s41746-024-01418-9 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841544302484783104 |
|---|---|
| author | Sam F. Friedman Shaan Khurshid Rachael A. Venn Xin Wang Nate Diamant Paolo Di Achille Lu-Chen Weng Seung Hoan Choi Christopher Reeder James P. Pirruccello Pulkit Singh Emily S. Lau Anthony Philippakis Christopher D. Anderson Mahnaz Maddah Puneet Batra Patrick T. Ellinor Jennifer E. Ho Steven A. Lubitz |
| author_facet | Sam F. Friedman Shaan Khurshid Rachael A. Venn Xin Wang Nate Diamant Paolo Di Achille Lu-Chen Weng Seung Hoan Choi Christopher Reeder James P. Pirruccello Pulkit Singh Emily S. Lau Anthony Philippakis Christopher D. Anderson Mahnaz Maddah Puneet Batra Patrick T. Ellinor Jennifer E. Ho Steven A. Lubitz |
| author_sort | Sam F. Friedman |
| collection | DOAJ |
| description | Abstract The 12-lead electrocardiogram (ECG) is inexpensive and widely available. Whether conditions across the human disease landscape can be detected using the ECG is unclear. We developed a deep learning denoising autoencoder and systematically evaluated associations between ECG encodings and ~1,600 Phecode-based diseases in three datasets separate from model development, and meta-analyzed the results. The latent space ECG model identified associations with 645 prevalent and 606 incident Phecodes. Associations were most enriched in the circulatory (n = 140, 82% of category-specific Phecodes), respiratory (n = 53, 62%) and endocrine/metabolic (n = 73, 45%) categories, with additional associations across the phenome. The strongest ECG association was with hypertension (p < 2.2×10-308). The ECG latent space model demonstrated more associations than models using standard ECG intervals, and offered favorable discrimination of prevalent disease compared to models comprising age, sex, and race. We further demonstrate how latent space models can be used to generate disease-specific ECG waveforms and facilitate individual disease profiling. |
| format | Article |
| id | doaj-art-a06932aedc4e4d5ab67fc4215739cd9c |
| institution | Kabale University |
| issn | 2398-6352 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | npj Digital Medicine |
| spelling | doaj-art-a06932aedc4e4d5ab67fc4215739cd9c2025-01-12T12:40:57ZengNature Portfolionpj Digital Medicine2398-63522025-01-018111310.1038/s41746-024-01418-9Unsupervised deep learning of electrocardiograms enables scalable human disease profilingSam F. Friedman0Shaan Khurshid1Rachael A. Venn2Xin Wang3Nate Diamant4Paolo Di Achille5Lu-Chen Weng6Seung Hoan Choi7Christopher Reeder8James P. Pirruccello9Pulkit Singh10Emily S. Lau11Anthony Philippakis12Christopher D. Anderson13Mahnaz Maddah14Puneet Batra15Patrick T. Ellinor16Jennifer E. Ho17Steven A. Lubitz18Data Sciences Platform, The Broad Institute of MIT and HarvardCardiovascular Research Center, Massachusetts General HospitalCardiovascular Research Center, Massachusetts General HospitalCardiovascular Research Center, Massachusetts General HospitalData Sciences Platform, The Broad Institute of MIT and HarvardData Sciences Platform, The Broad Institute of MIT and HarvardCardiovascular Research Center, Massachusetts General HospitalCardiovascular Disease Initiative, The Broad Institute of MIT and HarvardData Sciences Platform, The Broad Institute of MIT and HarvardCardiovascular Disease Initiative, The Broad Institute of MIT and HarvardData Sciences Platform, The Broad Institute of MIT and HarvardCardiovascular Research Center, Massachusetts General HospitalGoogle VenturesDepartment of Neurology, Brigham and Women’s HospitalData Sciences Platform, The Broad Institute of MIT and HarvardData Sciences Platform, The Broad Institute of MIT and HarvardCardiovascular Research Center, Massachusetts General HospitalCardiovascular Disease Initiative, The Broad Institute of MIT and HarvardCardiovascular Research Center, Massachusetts General HospitalAbstract The 12-lead electrocardiogram (ECG) is inexpensive and widely available. Whether conditions across the human disease landscape can be detected using the ECG is unclear. We developed a deep learning denoising autoencoder and systematically evaluated associations between ECG encodings and ~1,600 Phecode-based diseases in three datasets separate from model development, and meta-analyzed the results. The latent space ECG model identified associations with 645 prevalent and 606 incident Phecodes. Associations were most enriched in the circulatory (n = 140, 82% of category-specific Phecodes), respiratory (n = 53, 62%) and endocrine/metabolic (n = 73, 45%) categories, with additional associations across the phenome. The strongest ECG association was with hypertension (p < 2.2×10-308). The ECG latent space model demonstrated more associations than models using standard ECG intervals, and offered favorable discrimination of prevalent disease compared to models comprising age, sex, and race. We further demonstrate how latent space models can be used to generate disease-specific ECG waveforms and facilitate individual disease profiling.https://doi.org/10.1038/s41746-024-01418-9 |
| spellingShingle | Sam F. Friedman Shaan Khurshid Rachael A. Venn Xin Wang Nate Diamant Paolo Di Achille Lu-Chen Weng Seung Hoan Choi Christopher Reeder James P. Pirruccello Pulkit Singh Emily S. Lau Anthony Philippakis Christopher D. Anderson Mahnaz Maddah Puneet Batra Patrick T. Ellinor Jennifer E. Ho Steven A. Lubitz Unsupervised deep learning of electrocardiograms enables scalable human disease profiling npj Digital Medicine |
| title | Unsupervised deep learning of electrocardiograms enables scalable human disease profiling |
| title_full | Unsupervised deep learning of electrocardiograms enables scalable human disease profiling |
| title_fullStr | Unsupervised deep learning of electrocardiograms enables scalable human disease profiling |
| title_full_unstemmed | Unsupervised deep learning of electrocardiograms enables scalable human disease profiling |
| title_short | Unsupervised deep learning of electrocardiograms enables scalable human disease profiling |
| title_sort | unsupervised deep learning of electrocardiograms enables scalable human disease profiling |
| url | https://doi.org/10.1038/s41746-024-01418-9 |
| work_keys_str_mv | AT samffriedman unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT shaankhurshid unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT rachaelavenn unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT xinwang unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT natediamant unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT paolodiachille unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT luchenweng unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT seunghoanchoi unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT christopherreeder unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT jamesppirruccello unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT pulkitsingh unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT emilyslau unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT anthonyphilippakis unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT christopherdanderson unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT mahnazmaddah unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT puneetbatra unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT patricktellinor unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT jennifereho unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling AT stevenalubitz unsuperviseddeeplearningofelectrocardiogramsenablesscalablehumandiseaseprofiling |