Related Experiment Video
Updated: Jun 19, 2025
Transthoracic Speckle Tracking Echocardiography for the Quantitative Assessment of Left Ventricular Myocardial Deformation
Published on: October 20, 2016
Fully automatic estimation of global left ventricular systolic function using deep learning in transoesophageal
Erik Andreas Rye Berg1,2,3, Anders Austlid Taskén4, Trym Nordal5
1Centre for Innovative Ultrasound Solutions, Department of Circulation and Medical Imaging, Faculty of Medicine and Health Science, Norwegian University of Science and Technology, Prinsesse Kristinas gate 3, Trondheim 7030, Norway.
A new deep learning method for automatic mitral annular plane systolic excursion (auto-MAPSE) estimation in transesophageal echocardiography (TOE) is fast, feasible, and reliable for monitoring cardiac function in heart disease patients.
Area of Science:
- Cardiology
- Medical Imaging
- Artificial Intelligence
Background:
- Accurate monitoring of cardiac function is crucial during major surgery and intensive care.
- Transesophageal echocardiography (TOE) is a key imaging modality for assessing cardiac performance.
- Current methods for measuring mitral annular plane systolic excursion (MAPSE) can be time-consuming and subjective.
Purpose of the Study:
- To develop and clinically validate a fully automatic estimation of MAPSE (auto-MAPSE) using deep learning in TOE.
- To assess the feasibility, agreement, and inter-rater reliability of auto-MAPSE compared to manual measurements.
- To evaluate the potential of auto-MAPSE for real-time monitoring of left ventricular function.
Main Methods:
- A deep-learning-based auto-MAPSE method was developed and trained on 105 patient TOE recordings.
- Feasibility, agreement, and inter-rater reliability were assessed in 80 consecutive patients with heart disease.
- Comparisons were made against manual MAPSE reference measurements, with and without electrocardiogram (ECG) tracings.
Main Results:
- Auto-MAPSE demonstrated high feasibility (>90% overall, ≥95% in at least two walls).
- Agreement with manual reference was comparable to manual inter-observer agreement (bias -0.5 to -0.2 mm).
- Inter-rater reliability (Intra-class correlation coefficient) for auto-MAPSE was high (0.88-0.90).
Conclusions:
- Auto-MAPSE is a fast, highly feasible, and reproducible method for estimating systolic excursion.
- The performance of auto-MAPSE is comparable to manual assessments and manual inter-observer variability.
- Auto-MAPSE holds significant potential to enhance real-time monitoring of left ventricular function in clinical settings.
