A deep learning based method for left ventricular strain measurements: repeatability and accuracy compared to experienced echocardiographers

Abstract Background Speckle tracking echocardiography (STE) provides quantification of left ventricular (LV) deformation and is useful in the assessment of LV function. STE is increasingly being used clinically, and every effort to simplify and standardize STE is important. Manual outlining of regio...

Full description

Saved in:
Bibliographic Details
Main Authors: Magnus Rogstadkjernet, Sigurd Z. Zha, Lars G. Klæboe, Camilla K. Larsen, John M. Aalen, Esther Scheirlynck, Bjørn-Jostein Singstad, Steven Droogmans, Bernard Cosyns, Otto A. Smiseth, Kristina H. Haugaa, Thor Edvardsen, Eigil Samset, Pål H. Brekke
Format: Article
Language:English
Published: BMC 2024-11-01
Series:BMC Medical Imaging
Subjects:
Online Access:https://doi.org/10.1186/s12880-024-01470-7
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846164756194394112
author Magnus Rogstadkjernet
Sigurd Z. Zha
Lars G. Klæboe
Camilla K. Larsen
John M. Aalen
Esther Scheirlynck
Bjørn-Jostein Singstad
Steven Droogmans
Bernard Cosyns
Otto A. Smiseth
Kristina H. Haugaa
Thor Edvardsen
Eigil Samset
Pål H. Brekke
author_facet Magnus Rogstadkjernet
Sigurd Z. Zha
Lars G. Klæboe
Camilla K. Larsen
John M. Aalen
Esther Scheirlynck
Bjørn-Jostein Singstad
Steven Droogmans
Bernard Cosyns
Otto A. Smiseth
Kristina H. Haugaa
Thor Edvardsen
Eigil Samset
Pål H. Brekke
author_sort Magnus Rogstadkjernet
collection DOAJ
description Abstract Background Speckle tracking echocardiography (STE) provides quantification of left ventricular (LV) deformation and is useful in the assessment of LV function. STE is increasingly being used clinically, and every effort to simplify and standardize STE is important. Manual outlining of regions of interest (ROIs) is labor intensive and may influence assessment of strain values. Purpose We hypothesized that a deep learning (DL) model, trained on clinical echocardiographic exams, can be combined with a readily available echocardiographic analysis software, to automate strain calculation with comparable fidelity to trained cardiologists. Methods Data consisted of still frame echocardiographic images with cardiologist-defined ROIs from 672 clinical echocardiographic exams from a university hospital outpatient clinic. Exams included patients with ischemic heart disease, heart failure, valvular disease, and conduction abnormalities, and some healthy subjects. An EfficientNetB1-based architecture was employed, and different techniques and properties including data set size, data quality, augmentations, and transfer learning were evaluated. DL predicted ROIs were reintroduced into commercially available echocardiographic analysis software to automatically calculate strain values. Results DL-automated strain calculations had an average absolute difference of 0.75 (95% CI 0.58–0.92) for global longitudinal strain (GLS), and 1.16 (95% CI 1.03–1.29) for single-projection longitudinal strain (LS), compared to operators. A Bland–Altman plot revealed no obvious bias, though there were fewer outliers in the lower average LS ranges. Techniques and data properties yielded no significant increase/decrease in performance. Conclusion The study demonstrates that DL-assisted, automated strain measurements are feasible, and provide results within interobserver variation. Employing DL in echocardiographic analyses could further facilitate adoption of STE parameters in clinical practice and research, and improve reproducibility.
format Article
id doaj-art-5d494022ecb341ab8768554221743f05
institution Kabale University
issn 1471-2342
language English
publishDate 2024-11-01
publisher BMC
record_format Article
series BMC Medical Imaging
spelling doaj-art-5d494022ecb341ab8768554221743f052024-11-17T12:53:59ZengBMCBMC Medical Imaging1471-23422024-11-0124111010.1186/s12880-024-01470-7A deep learning based method for left ventricular strain measurements: repeatability and accuracy compared to experienced echocardiographersMagnus Rogstadkjernet0Sigurd Z. Zha1Lars G. Klæboe2Camilla K. Larsen3John M. Aalen4Esther Scheirlynck5Bjørn-Jostein Singstad6Steven Droogmans7Bernard Cosyns8Otto A. Smiseth9Kristina H. Haugaa10Thor Edvardsen11Eigil Samset12Pål H. Brekke13Institute for Clinical Medicine, University of OsloInstitute for Clinical Medicine, University of OsloProCardio Center for Innovation, Department of Cardiology, Oslo University HospitalInstitute for Surgical Research, Oslo University Hospital and University of OsloInstitute for Surgical Research, Oslo University Hospital and University of OsloCentrum Voor Hart-en Vaatziekten, Universitair Ziekenhuis Brussel-Vrije Universiteit BrusselProCardio Center for Innovation, Department of Cardiology, Oslo University HospitalDepartment of Cardiology, Universitair Ziekenhuis BrusselCentrum Voor Hart- en Vaatziekten, Universitair Ziekenhuis BrusselInstitute for Clinical Medicine, University of OsloProCardio Center for Innovation, Department of Cardiology, Oslo University HospitalDepartment of Cardiology, Oslo University HospitalProCardio Center for Innovation, Department of Cardiology, Oslo University HospitalDepartment of Cardiology, Oslo University HospitalAbstract Background Speckle tracking echocardiography (STE) provides quantification of left ventricular (LV) deformation and is useful in the assessment of LV function. STE is increasingly being used clinically, and every effort to simplify and standardize STE is important. Manual outlining of regions of interest (ROIs) is labor intensive and may influence assessment of strain values. Purpose We hypothesized that a deep learning (DL) model, trained on clinical echocardiographic exams, can be combined with a readily available echocardiographic analysis software, to automate strain calculation with comparable fidelity to trained cardiologists. Methods Data consisted of still frame echocardiographic images with cardiologist-defined ROIs from 672 clinical echocardiographic exams from a university hospital outpatient clinic. Exams included patients with ischemic heart disease, heart failure, valvular disease, and conduction abnormalities, and some healthy subjects. An EfficientNetB1-based architecture was employed, and different techniques and properties including data set size, data quality, augmentations, and transfer learning were evaluated. DL predicted ROIs were reintroduced into commercially available echocardiographic analysis software to automatically calculate strain values. Results DL-automated strain calculations had an average absolute difference of 0.75 (95% CI 0.58–0.92) for global longitudinal strain (GLS), and 1.16 (95% CI 1.03–1.29) for single-projection longitudinal strain (LS), compared to operators. A Bland–Altman plot revealed no obvious bias, though there were fewer outliers in the lower average LS ranges. Techniques and data properties yielded no significant increase/decrease in performance. Conclusion The study demonstrates that DL-assisted, automated strain measurements are feasible, and provide results within interobserver variation. Employing DL in echocardiographic analyses could further facilitate adoption of STE parameters in clinical practice and research, and improve reproducibility.https://doi.org/10.1186/s12880-024-01470-7Speckle-tracking echocardiographyStrain rate imagingDeep learningArtificial intelligenceAutomation
spellingShingle Magnus Rogstadkjernet
Sigurd Z. Zha
Lars G. Klæboe
Camilla K. Larsen
John M. Aalen
Esther Scheirlynck
Bjørn-Jostein Singstad
Steven Droogmans
Bernard Cosyns
Otto A. Smiseth
Kristina H. Haugaa
Thor Edvardsen
Eigil Samset
Pål H. Brekke
A deep learning based method for left ventricular strain measurements: repeatability and accuracy compared to experienced echocardiographers
BMC Medical Imaging
Speckle-tracking echocardiography
Strain rate imaging
Deep learning
Artificial intelligence
Automation
title A deep learning based method for left ventricular strain measurements: repeatability and accuracy compared to experienced echocardiographers
title_full A deep learning based method for left ventricular strain measurements: repeatability and accuracy compared to experienced echocardiographers
title_fullStr A deep learning based method for left ventricular strain measurements: repeatability and accuracy compared to experienced echocardiographers
title_full_unstemmed A deep learning based method for left ventricular strain measurements: repeatability and accuracy compared to experienced echocardiographers
title_short A deep learning based method for left ventricular strain measurements: repeatability and accuracy compared to experienced echocardiographers
title_sort deep learning based method for left ventricular strain measurements repeatability and accuracy compared to experienced echocardiographers
topic Speckle-tracking echocardiography
Strain rate imaging
Deep learning
Artificial intelligence
Automation
url https://doi.org/10.1186/s12880-024-01470-7
work_keys_str_mv AT magnusrogstadkjernet adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT sigurdzzha adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT larsgklæboe adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT camillaklarsen adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT johnmaalen adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT estherscheirlynck adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT bjørnjosteinsingstad adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT stevendroogmans adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT bernardcosyns adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT ottoasmiseth adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT kristinahhaugaa adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT thoredvardsen adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT eigilsamset adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT palhbrekke adeeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT magnusrogstadkjernet deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT sigurdzzha deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT larsgklæboe deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT camillaklarsen deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT johnmaalen deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT estherscheirlynck deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT bjørnjosteinsingstad deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT stevendroogmans deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT bernardcosyns deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT ottoasmiseth deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT kristinahhaugaa deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT thoredvardsen deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT eigilsamset deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers
AT palhbrekke deeplearningbasedmethodforleftventricularstrainmeasurementsrepeatabilityandaccuracycomparedtoexperiencedechocardiographers