TLDR: Researchers developed autoStrain, an AI-powered system using deep learning (TeeTracker and TeeFlow) to automatically measure segmental longitudinal strain (SLS) in transesophageal echocardiography (TEE). Using a novel synthetic dataset, TeeTracker showed superior accuracy in motion estimation. Clinical validation demonstrated promising alignment with reference measurements, indicating AI’s potential to enhance cardiac function assessment and detect abnormalities in critically ill patients.
Assessing the heart’s function, particularly the regional movement of the left ventricle (LV), is vital for diagnosing and managing conditions like myocardial ischemia. The current methods for measuring Segmental Longitudinal Strain (SLS), which indicates the deformation of the heart muscle, are often manual, time-consuming, and require specialized expertise. This makes them less efficient for continuous monitoring in clinical settings.
Transesophageal echocardiography (TEE) offers superior image clarity compared to standard transthoracic echocardiography (TTE), making it a preferred choice for monitoring perioperative patients. TEE allows for continuous observation and minimizes trauma by passively placing the transducer in the esophagus. However, most existing automated tools for motion estimation in echocardiography are optimized for TTE, not TEE, which presents unique challenges such as increased foreshortening, different speckle patterns, and unique noise characteristics.
A new study introduces an automated pipeline called autoStrain, designed to estimate SLS in TEE using advanced deep learning (DL) methods for motion estimation. The research compares two distinct DL approaches: TeeFlow, which is based on the RAFT optical flow model for dense frame-to-frame predictions, and TeeTracker, which utilizes the CoTracker point trajectory model for sparse, long-sequence predictions.
A significant challenge in developing and evaluating such models is the scarcity of real echocardiographic sequences with ground truth motion data. To address this, the researchers leveraged a unique simulation pipeline (SIMUS) to generate a highly realistic synthetic TEE (synTEE) dataset. This dataset included data from 80 patients with ground truth myocardial motion, allowing for robust training and evaluation of both models. The synTEE dataset was designed to incorporate various degrees of speckle decorrelation and even simulated myocardial infarctions to mimic real-world clinical scenarios.
The evaluation revealed that TeeTracker consistently outperformed TeeFlow in accuracy, achieving a mean distance error in motion estimation of 0.65 ± 0.20 mm on the synTEE test dataset. This suggests that point trajectory estimation is more effective than dense motion estimation for myocardial motion in echocardiography. The study also found that a combined training approach, utilizing all synthetic datasets with varying levels of decorrelation, yielded the best overall performance for the models.
Crucially, the incorporation of simulated ischemia in the synTEE data significantly improved the models’ accuracy in quantifying abnormal deformation. This enhancement highlights autoStrain’s potential to detect and localize myocardial abnormalities, thereby increasing its clinical applicability.
Clinical validation on 16 patients further demonstrated that SLS estimation with the autoStrain pipeline aligned well with clinical references, achieving a mean difference of 1.09% (-8.90% to 11.09%). While the estimation of Global Longitudinal Strain (GLS) remained consistent when moving from synthetic to real datasets, the performance for SLS estimation showed a noticeable decline, indicating that further refinements in synthetic datasets are needed to maintain accuracy for local strain estimates in real-world TEE data.
The findings indicate that integrating AI-driven motion estimation with TEE can significantly enhance the precision and efficiency of cardiac function assessment in clinical settings. This advancement could play a crucial role in improving the perioperative monitoring and treatment of critically ill patients. For more detailed information, you can refer to the full research paper available at arXiv:2511.02210.
Also Read:
- AI and Robotics Team Up for Autonomous Cardiac Ultrasound Scanning
- Detecting Arrhythmia with Enhanced CNN-LSTM and H-Infinity Filters
Future work will focus on expanding the clinical validation datasets to include a broader range of cardiac conditions and patient demographics, as well as evaluating the robustness of the methods across different ultrasound vendors. The ultimate goal is to seamlessly integrate these AI tools into real-time clinical workflows, leading to more precise and timely diagnostic assessments.
