Myocardial Function Imaging in Echocardiography Using Deep Learning

Deep Learning-Derived Myocardial Strain

BACKGROUND

Echocardiographic strain measurements require extensive operator experience and have significant intervendor variability. Creating an automated, open-source, vendor-agnostic method to retrospectively measure global longitudinal strain (GLS) from standard echocardiography B-mode images would greatly improve post hoc research applications and may streamline patient analyses.

OBJECTIVES

This study was seeking to develop an automated deep learning strain (DLS) analysis pipeline and validate its performance across multiple applications and populations.

METHODS

Interobserver/-vendor variation of traditional GLS, and simulated effects of variation in contour on speckle-tracking measurements were assessed. The DLS pipeline was designed to take semantic segmentation results from EchoNet-Dynamic and derive longitudinal strain by calculating change in the length of the left ventricular endocardial contour. DLS was evaluated for agreement with GLS on a large external dataset and applied across a range of conditions that result in cardiac hypertrophy.

RESULTS

In patients scanned by 2 sonographers using 2 vendors, GLS had an intraclass correlation of 0.29 (95% CI: -0.01 to 0.53, P = 0.03) between vendor measurements and 0.63 (95% CI: 0.48-0.74, P < 0.001) between sonographers. With minor changes in initial input contour, step-wise pixel shifts resulted in a mean absolute error of 3.48% and proportional strain difference of 13.52% by a 6-pixel shift. In external validation, DLS maintained moderate agreement with 2-dimensional GLS (intraclass correlation coefficient [ICC]: 0.56, P = 0.002) with a bias of -3.31% (limits of agreement: -11.65% to 5.02%). The DLS method showed differences (P < 0.0001) between populations with cardiac hypertrophy and had moderate agreement in a patient population of advanced cardiac amyloidosis: ICC was 0.64 (95% CI: 0.53-0.72), P < 0.001, with a bias of 0.57%, limits of agreement of -4.87% to 6.01% vs 2-dimensional GLS.

CONCLUSIONS

The open-source DLS provides lower variation than human measurements and similar quantitative results. The method is rapid, consistent, vendor-agnostic, publicly released, and applicable across a wide range of imaging qualities.

A Motion-Aware DNN Model with Edge Focus Loss and Quality Control for Short-Axis Left Ventricle Segmentation of Cine MR Sequences

Accurate segmentation of the left ventricle myocardium is the key step of automatic assessment of cardiac function. However, the current methods mainly focus on the end-diastolic and the end-systolic frames in cine MR sequences and lack the attention to myocardial motion in the cardiac cycle. Additionally, due to the lack of fine segmentation tools, the simplified approach, excluding papillary muscles and trabeculae from myocardium, is applied in clinical practice. To solve these problems, we propose a motion-aware DNN model with edge focus loss and quality control in this paper. Specifically, the bidirectional ConvLSTM layer and a new motion attention layer are proposed to encode motion-aware feature maps, and an edge focus loss function is proposed to train the model to generate the fine segmentation results. Additionally, a quality control method is proposed to filter out the abnormal segmentations before subsequent analyses. Compared with state-of-the-art segmentation models on the public dataset and the in-house dataset, the proposed method has obtained high segmentation accuracy. On the 17-segment model, the proposed method has obtained the highest Pearson correlation coefficient at 14 of 17 segments, and the mean PCC of 85%. The experimental results highlight the segmentation accuracy of the proposed method as well as its availability to substitute for the manually annotated boundaries for the automatic assessment of cardiac function.

Automatic Detection and Tracking of Anatomical Landmarks in Transesophageal Echocardiography for Quantification of Left Ventricular Function

OBJECTIVE

Evaluation of left ventricular (LV) function in critical care patients is useful for guidance of therapy and early detection of LV dysfunction, but the tools currently available are too time-consuming. To resolve this issue, we previously proposed a method for the continuous and automatic quantification of global LV function in critical care patients based on the detection and tracking of anatomical landmarks on transesophageal heart ultrasound. In the present study, our aim was to improve the performance of mitral annulus detection in transesophageal echocardiography (TEE).

METHODS

We investigated several state-of-the-art networks for both the detection and tracking of the mitral annulus in TEE. We integrated the networks into a pipeline for automatic assessment of LV function through estimation of the mitral annular plane systolic excursion (MAPSE), called autoMAPSE. TEE recordings from a total of 245 patients were collected from St. Olav's University Hospital and used to train and test the respective networks. We evaluated the agreement between autoMAPSE estimates and manual references annotated by expert echocardiographers in 30 Echolab patients and 50 critical care patients. Furthermore, we proposed a prototype of autoMAPSE for clinical integration and tested it in critical care patients in the intensive care unit.

RESULTS

Compared with manual references, we achieved a mean difference of 0.8 (95% limits of agreement: -2.9 to 4.7) mm in Echolab patients, with a feasibility of 85.7%. In critical care patients, we reached a mean difference of 0.6 (95% limits of agreement: -2.3 to 3.5) mm and a feasibility of 88.1%. The clinical prototype of autoMAPSE achieved real-time performance.

CONCLUSION

Automatic quantification of LV function had high feasibility in clinical settings. The agreement with manual references was comparable to inter-observer variability of clinical experts.

Atrial Septal Defect Detection in Children Based on Ultrasound Video Using Multiple Instances Learning

Thoracic echocardiography (TTE) can provide sufficient cardiac structure information, evaluate hemodynamics and cardiac function, and is an effective method for atrial septal defect (ASD) examination. This paper aims to study a deep learning method based on cardiac ultrasound video to assist in ASD diagnosis. We chose four standard views in pediatric cardiac ultrasound to identify atrial septal defects; the four standard views were as follows: subcostal sagittal view of the atrium septum (subSAS), apical four-chamber view (A4C), the low parasternal four-chamber view (LPS4C), and parasternal short-axis view of large artery (PSAX). We enlist data from 300 children patients as part of a double-blind experiment for five-fold cross-validation to verify the performance of our model. In addition, data from 30 children patients (15 positives and 15 negatives) are collected for clinician testing and compared to our model test results (these 30 samples do not participate in model training). In our model, we present a block random selection, maximal agreement decision, and frame sampling strategy for training and testing respectively, resNet18 and r3D networks are used to extract the frame features and aggregate them to build a rich video-level representation. We validate our model using our private dataset by five cross-validation. For ASD detection, we achieve 89.33 ± 3.13 AUC, 84.95 ± 3.88 accuracy, 85.70 ± 4.91 sensitivity, 81.51 ± 8.15 specificity, and 81.99 ± 5.30 F1 score. The proposed model is a multiple instances learning-based deep learning model for video atrial septal defect detection which effectively improves ASD detection accuracy when compared to the performances of previous networks and clinical doctors.

Automatic 3D left atrial strain extraction framework on cardiac computed tomography

BACKGROUND AND OBJECTIVE

Strain analysis provides insights into myocardial function and cardiac condition evaluation. However, the anatomical characteristics of left atrium (LA) inherently limit LA strain analysis when using echocardiography. Cardiac computed tomography (CT) with its superior spatial resolution, has become critical for in-depth evaluation of LA function. Recent studies have explored the feasibility of CT-derived strain; however, they relied on manually selected regions of interest (ROIs) and mainly focused on left ventricle (LV). This study aimed to propose a first-of-its-kind fully automatic deep learning (DL)-based framework for three-dimensional (3D) LA strain extraction on cardiac CT.

METHODS

A total of 111 patients undergoing ECG-gated contrast-enhanced CT for evaluating subclinical atrial fibrillation (AF) were enrolled in this study. We developed a 3D strain extraction framework on cardiac CT images, containing a 2.5D GN-U-Net network for LA segmentation, axis-oriented 3D view extraction, and LA strain measure. The segmentation accuracy was evaluated using Dice similarity coefficient (DSC). The model-extracted LA volumes and emptying fraction (EF) were compared with ground-truth measurements using intraclass correlation coefficient (ICC), correlation coefficient (r), and Bland-Altman plot (B-A). The automatically extracted LA strains were evaluated against the LA strains measured from 2D echocardiograms. We utilized this framework to gauge the effect of AF burden on LA strain, employing the atrial high rate episode (AHRE) burden as the measurement parameter.

RESULTS

The GN-U-Net LA segmentation network achieved a DSC score of 0.9603 on the test set. The framework-extracted LA estimates demonstrated excellent ICCs of 0.949 (95 % CI: 0.93-0.97) for minimal LA volume, 0.904 (95 % CI: 0.86-0.93) for maximal LA volume, and 0.902 (95 % CI: 0.86-0.93) for EF, compared with expert measurements. The framework-extracted LA strains demonstrated moderate agreement with the LA strains based on 2D echocardiography (ICCs >0.703). Patients with AHRE > 6 min had significantly lower global strain and LAEF, as extracted by the framework than those with AHRE ≤ 6 min.

CONCLUSION

The promising results highlighted the feasibility and clinical usefulness of automatically extracting 3D LA strain from CT images using a DL-based framework. This tool could provide a 3D-based alternative to echocardiography for assessing LA function.

Left Ventricular Segmentation, Warping, and Myocardial Registration for Automated Strain Measurement

The left ventricular global longitudinal strain (LVGLS) is a crucial prognostic indicator. However, inconsistencies in measurements due to the speckle tracking algorithm and manual adjustments have hindered its standardization and democratization. To solve this issue, we proposed a fully automated strain measurement by artificial intelligence-assisted LV segmentation contours. The LV segmentation model was trained from echocardiograms of 368 adults (11,125 frames). We compared the registration-like effects of dynamic time warping (DTW) with speckle tracking on a synthetic echocardiographic dataset in experiment-1. In experiment-2, we enrolled 80 patients to compare the DTW method with commercially available software. In experiment-3, we combined the segmentation model and DTW method to create the artificial intelligence (AI)-DTW method, which was then tested on 40 patients with general LV morphology, 20 with dilated cardiomyopathy (DCMP), and 20 with transthyretin-associated cardiac amyloidosis (ATTR-CA), 20 with severe aortic stenosis (AS), and 20 with severe mitral regurgitation (MR). Experiments-1 and -2 revealed that the DTW method is consistent with dedicated software. In experiment-3, the AI-DTW strain method showed comparable results for general LV morphology (bias - 0.137 ± 0.398%), DCMP (- 0.397 ± 0.607%), ATTR-CA (0.095 ± 0.581%), AS (0.334 ± 0.358%), and MR (0.237 ± 0.490%). Moreover, the strain curves showed a high correlation in their characteristics, with R-squared values of 0.8879-0.9452 for those LV morphology in experiment-3. Measuring LVGLS through dynamic warping of segmentation contour is a feasible method compared to traditional tracking techniques. This approach has the potential to decrease the need for manual demarcation and make LVGLS measurements more efficient and user-friendly for daily practice.

Automated Segmentation and Quantification of the Right Ventricle in 2-D Echocardiography

OBJECTIVE

The right ventricle receives less attention than its left counterpart in echocardiography research, practice and development of automated solutions. In the work described here, we sought to determine that the deep learning methods for automated segmentation of the left ventricle in 2-D echocardiograms are also valid for the right ventricle. Additionally, here we describe and explore a keypoint detection approach to segmentation that guards against erratic behavior often displayed by segmentation models.

METHODS

We used a data set of echo images focused on the right ventricle from 250 participants to train and evaluate several deep learning models for segmentation and keypoint detection. We propose a compact architecture (U-Net KP) employing the latter approach. The architecture is designed to balance high speed with accuracy and robustness.

RESULTS

All featured models achieved segmentation accuracy close to the inter-observer variability. When computing the metrics of right ventricular systolic function from contour predictions of U-Net KP, we obtained the bias and 95% limits of agreement of 0.8 ± 10.8% for the right ventricular fractional area change measurements, -0.04 ± 0.54 cm for the tricuspid annular plane systolic excursion measurements and 0.2 ± 6.6% for the right ventricular free wall strain measurements. These results were also comparable to the semi-automatically derived inter-observer discrepancies of 0.4 ± 11.8%, -0.37 ± 0.58 cm and -1.0 ± 7.7% for the aforementioned metrics, respectively.

CONCLUSION

Given the appropriate data, automated segmentation and quantification of the right ventricle in 2-D echocardiography are feasible with existing methods. However, keypoint detection architectures may offer higher robustness and information density for the same computational cost.

External validation of a deep learning algorithm for automated echocardiographic strain measurements

Aims

Echocardiographic strain imaging reflects myocardial deformation and is a sensitive measure of cardiac function and wall-motion abnormalities. Deep learning (DL) algorithms could automate the interpretation of echocardiographic strain imaging.

Methods and results

We developed and trained an automated DL-based algorithm for left ventricular (LV) strain measurements in an internal dataset. Global longitudinal strain (GLS) was validated externally in (i) a real-world Taiwanese cohort of participants with and without heart failure (HF), (ii) a core-lab measured dataset from the multinational prevalence of microvascular dysfunction-HF and preserved ejection fraction (PROMIS-HFpEF) study, and regional strain in (iii) the HMC-QU-MI study of patients with suspected myocardial infarction. Outcomes included measures of agreement [bias, mean absolute difference (MAD), root-mean-squared-error (RMSE), and Pearson's correlation (R)] and area under the curve (AUC) to identify HF and regional wall-motion abnormalities. The DL workflow successfully analysed 3741 (89%) studies in the Taiwanese cohort, 176 (96%) in PROMIS-HFpEF, and 158 (98%) in HMC-QU-MI. Automated GLS showed good agreement with manual measurements (mean ± SD): -18.9 ± 4.5% vs. -18.2 ± 4.4%, respectively, bias 0.68 ± 2.52%, MAD 2.0 ± 1.67, RMSE = 2.61, R = 0.84 in the Taiwanese cohort; and -15.4 ± 4.1% vs. -15.9 ± 3.6%, respectively, bias -0.65 ± 2.71%, MAD 2.19 ± 1.71, RMSE = 2.78, R = 0.76 in PROMIS-HFpEF. In the Taiwanese cohort, automated GLS accurately identified patients with HF (AUC = 0.89 for total HF and AUC = 0.98 for HF with reduced ejection fraction). In HMC-QU-MI, automated regional strain identified regional wall-motion abnormalities with an average AUC = 0.80.

Conclusion

DL algorithms can interpret echocardiographic strain images with similar accuracy as conventional measurements. These results highlight the potential of DL algorithms to democratize the use of cardiac strain measurements and reduce time-spent and costs for echo labs globally.

Ultrafast Cardiac Imaging Using Deep Learning for Speckle-Tracking Echocardiography

High-quality ultrafast ultrasound imaging is based on coherent compounding from multiple transmissions of plane waves (PW) or diverging waves (DW). However, compounding results in reduced frame rate, as well as destructive interferences from high-velocity tissue motion if motion compensation (MoCo) is not considered. While many studies have recently shown the interest of deep learning for the reconstruction of high-quality static images from PW or DW, its ability to achieve such performance while maintaining the capability of tracking cardiac motion has yet to be assessed. In this article, we addressed such issue by deploying a complex-weighted convolutional neural network (CNN) for image reconstruction and a state-of-the-art speckle-tracking method. The evaluation of this approach was first performed by designing an adapted simulation framework, which provides specific reference data, i.e., high-quality, motion artifact-free cardiac images. The obtained results showed that, while using only three DWs as input, the CNN-based approach yielded an image quality and a motion accuracy equivalent to those obtained by compounding 31 DWs free of motion artifacts. The performance was then further evaluated on nonsimulated, experimental in vitro data, using a spinning disk phantom. This experiment demonstrated that our approach yielded high-quality image reconstruction and motion estimation, under a large range of velocities and outperforms a state-of-the-art MoCo-based approach at high velocities. Our method was finally assessed on in vivo datasets and showed consistent improvement in image quality and motion estimation compared to standard compounding. This demonstrates the feasibility and effectiveness of deep learning reconstruction for ultrafast speckle-tracking echocardiography.

Deep learning for automated left ventricular outflow tract diameter measurements in 2D echocardiography

BACKGROUND

Measurement of the left ventricular outflow tract diameter (LVOTd) in echocardiography is a common source of error when used to calculate the stroke volume. The aim of this study is to assess whether a deep learning (DL) model, trained on a clinical echocardiographic dataset, can perform automatic LVOTd measurements on par with expert cardiologists.

METHODS

Data consisted of 649 consecutive transthoracic echocardiographic examinations of patients with coronary artery disease admitted to a university hospital. 1304 LVOTd measurements in the parasternal long axis (PLAX) and zoomed parasternal long axis views (ZPLAX) were collected, with each patient having 1-6 measurements per examination. Data quality control was performed by an expert cardiologist, and spatial geometry data was preserved for each LVOTd measurement to convert DL predictions into metric units. A convolutional neural network based on the U-Net was used as the DL model.

RESULTS

The mean absolute LVOTd error was 1.04 (95% confidence interval [CI] 0.90-1.19) mm for DL predictions on the test set. The mean relative LVOTd errors across all data subgroups ranged from 3.8 to 5.1% for the test set. Generally, the DL model had superior performance on the ZPLAX view compared to the PLAX view. DL model precision for patients with repeated LVOTd measurements had a mean coefficient of variation of 2.2 (95% CI 1.6-2.7) %, which was comparable to the clinicians for the test set.

CONCLUSION

DL for automatic LVOTd measurements in PLAX and ZPLAX views is feasible when trained on a limited clinical dataset. While the DL predicted LVOTd measurements were within the expected range of clinical inter-observer variability, the robustness of the DL model requires validation on independent datasets. Future experiments using temporal information and anatomical constraints could improve valvular identification and reduce outliers, which are challenges that must be addressed before clinical utilization.

Challenges and solutions of echocardiography generalization for deep learning: a study in patients with constrictive pericarditis

Purpose

The inherent characteristics of transthoracic echocardiography (TTE) images such as low signal-to-noise ratio and acquisition variations can limit the direct use of TTE images in the development and generalization of deep learning models. As such, we propose an innovative automated framework to address the common challenges in the process of echocardiography deep learning model generalization on the challenging task of constrictive pericarditis (CP) and cardiac amyloidosis (CA) differentiation.

Approach

Patients with a confirmed diagnosis of CP or CA and normal cases from Mayo Clinic Rochester and Arizona were identified to extract baseline demographics and the apical 4 chamber view from TTE studies. We proposed an innovative preprocessing and image generalization framework to process the images for training the ResNet50, ResNeXt101, and EfficientNetB2 models. Ablation studies were conducted to justify the effect of each proposed processing step in the final classification performance.

Results

The models were initially trained and validated on 720 unique TTE studies from Mayo Rochester and further validated on 225 studies from Mayo Arizona. With our proposed generalization framework, EfficientNetB2 generalized the best with an average area under the curve (AUC) of 0.96 (± 0.01 ) and 0.83 (± 0.03 ) on the Rochester and Arizona test sets, respectively.

Conclusions

Leveraging the proposed generalization techniques, we successfully developed an echocardiography-based deep learning model that can accurately differentiate CP from CA and normal cases and applied the model to images from two sites. The proposed framework can be further extended for the development of echocardiography-based deep learning models.

Non-invasive monitoring of cardiac function through Ballistocardiogram: an algorithm integrating short-time Fourier transform and ensemble empirical mode decomposition

The Ballistocardiogram (BCG) is a vibration signal that is generated by the displacement of the entire body due to the injection of blood during each heartbeat. It has been extensively utilized to monitor heart rate. The morphological features of the BCG signal serve as effective indicators for the identification of atrial fibrillation and heart failure, holding great significance for BCG signal analysis. The IJK-complex identification allows for the estimation of inter-beat intervals (IBI) and enables a more detailed analysis of BCG amplitude and interval waves. This study presents a novel algorithm for identifying the IJK-complex in BCG signals, which is an improvement over most existing algorithms that only perform IBI estimation. The proposed algorithm employs a short-time Fourier transform and summation across frequencies to initially estimate the occurrence of the J wave using peak finding, followed by Ensemble Empirical Mode Decomposition and a regional search to precisely identify the J wave. The algorithm's ability to detect the morphological features of BCG signals and estimate heart rates was validated through experiments conducted on 10 healthy subjects and 2 patients with coronary heart disease. In comparison to commonly used methods, the presented scheme ensures accurate heart rate estimation and exhibits superior capability in detecting BCG morphological features. This advancement holds significant value for future applications involving BCG signals.

Deep Learning for Improved Precision and Reproducibility of Left Ventricular Strain in Echocardiography: A Test-Retest Study

AIMS

Assessment of left ventricular (LV) function by echocardiography is hampered by modest test-retest reproducibility. A novel artificial intelligence (AI) method based on deep learning provides fully automated measurements of LV global longitudinal strain (GLS) and may improve the clinical utility of echocardiography by reducing user-related variability. The aim of this study was to assess within-patient test-retest reproducibility of LV GLS measured by the novel AI method in repeated echocardiograms recorded by different echocardiographers and to compare the results to manual measurements.

METHODS

Two test-retest data sets (n = 40 and n = 32) were obtained at separate centers. Repeated recordings were acquired in immediate succession by 2 different echocardiographers at each center. For each data set, 4 readers measured GLS in both recordings using a semiautomatic method to construct test-retest interreader and intrareader scenarios. Agreement, mean absolute difference, and minimal detectable change (MDC) were compared to analyses by AI. In a subset of 10 patients, beat-to-beat variability in 3 cardiac cycles was assessed by 2 readers and AI.

RESULTS

Test-retest variability was lower with AI compared with interreader scenarios (data set I: MDC = 3.7 vs 5.5, mean absolute difference = 1.4 vs 2.1, respectively; data set II: MDC = 3.9 vs 5.2, mean absolute difference = 1.6 vs 1.9, respectively; all P < .05). There was bias in GLS measurements in 13 of 24 test-retest interreader scenarios (largest bias, 3.2 strain units). In contrast, there was no bias in measurements by AI. Beat-to-beat MDCs were 1,5, 2.1, and 2.3 for AI and the 2 readers, respectively. Processing time for analyses of GLS by the AI method was 7.9 ± 2.8 seconds.

CONCLUSION

A fast AI method for automated measurements of LV GLS reduced test-retest variability and removed bias between readers in both test-retest data sets. By improving the precision and reproducibility, AI may increase the clinical utility of echocardiography.

Retrospective Ultrasound Doppler Quantification Using a Single Acquisition in Healthy Adults

OBJECTIVE

Using an experimental tool for retrospective ultrasound Doppler quantification-with high temporal resolution and large spatial coverage-simultaneous flow and tissue measurements were obtained. We compared and validated these experimental values against conventional measurements to determine if the experimental acquisition produced trustworthy tissue and flow velocities.

METHODS

We included 21 healthy volunteers. The only exclusion criterion was the presence of an irregular heartbeat. Two ultrasound examinations were performed for each participant, one using conventional and one using experimental acquisition. The experimental acquisition used multiple plane wave emissions combined with electrocardiography stitching to obtain continuous data with over 3500 frames per second. With two recordings covering a biplane apical view of the left ventricle, we retrospectively extracted selected flow and tissue velocities.

RESULTS

Flow and tissue velocities were compared between the two acquisitions. Statistical testing showed a small but significant difference. We also exemplified the possibility of extracting spectral tissue Doppler from different sample volumes in the myocardium within the imaging sector, showing a decrease in the velocities from the base to the apex.

CONCLUSION

This study demonstrates the feasibility of simultaneous, retrospective spectral and color Doppler of both tissue and flow from an experimental acquisition covering a full sector width. The measurements were significantly different between the two acquisitions but were still comparable, as the biases were small compared to clinical practice, and the two acquisitions were not done simultaneously. The experimental acquisition also enabled the study of deformation by simultaneous spectral velocity traces from all regions of the image sector.

Machine learning-enhanced echocardiography for screening coronary artery disease

BACKGROUND

Since myocardial work (MW) and left atrial strain are valuable for screening coronary artery disease (CAD), this study aimed to develop a novel CAD screening approach based on machine learning-enhanced echocardiography.

METHODS

This prospective study used data from patients undergoing coronary angiography, in which the novel echocardiography features were extracted by a machine learning algorithm. A total of 818 patients were enrolled and randomly divided into training (80%) and testing (20%) groups. An additional 115 patients were also enrolled in the validation group.

RESULTS

The superior diagnosis model of CAD was optimized using 59 echocardiographic features in a gradient-boosting classifier. This model showed that the value of the receiver operating characteristic area under the curve (AUC) was 0.852 in the test group and 0.834 in the validation group, with high sensitivity (0.952) and low specificity (0.691), suggesting that this model is very sensitive for detecting CAD, but its low specificity may increase the high false-positive rate. We also determined that the false-positive cases were more susceptible to suffering cardiac events than the true-negative cases.

CONCLUSIONS

Machine learning-enhanced echocardiography can improve CAD detection based on the MW and left atrial strain features. Our developed model is valuable for estimating the pre-test probability of CAD and screening CAD patients in clinical practice.

TRIAL REGISTRATION

Registered as NCT03905200 at ClinicalTrials.gov. Registered on 5 April 2019.

Shear-Wave Particle-Velocity Estimation and Enhancement Using a Multi-Resolution Convolutional Neural Network

OBJECTIVE

Tissue mechanical properties are valuable markers for tissue characterization, aiding in the detection and staging of pathologies. Shear wave elastography (SWE) offers a quantitative assessment of tissue mechanical characteristics based on the SW propagation profile, which is derived from the SW particle motion. Improving the signal-to-noise ratio (SNR) of the SW particle motion would directly enhance the accuracy of the material property estimates such as elasticity or viscosity.

METHODS

In this paper, we present a 3-D multi-resolution convolutional neural network (MRCNN) to perform improved estimation of the SW particle velocity V_z. Additionally, we propose a novel approach to generate training data from real acquisitions, providing high SNR ground truth target data, one-to-one paired to inputs that are corrupted with real-world noise and disturbances.

DISCUSSION

By testing the network on in vitro data acquired from a commercial breast elastography phantom, we show that the MRCNN outperforms Loupas' autocorrelation algorithm with an improved SNR of 4.47 dB for the V_z signals, a two-fold decrease in the standard deviation of the downstream elasticity estimates, and a two-fold increase in the contrast-to-noise ratio of the elasticity maps. The generalizability of the network was further demonstrated with a set of ex vivo porcine liver data.

CONCLUSION

The proposed MRCNN outperforms the standard autocorrelation method, in particular in low SNR regimes.

Developing an Echocardiography-Based, Automatic Deep Learning Framework for the Differentiation of Increased Left Ventricular Wall Thickness Etiologies

AIMS

Increased left ventricular (LV) wall thickness is frequently encountered in transthoracic echocardiography (TTE). While accurate and early diagnosis is clinically important, given the differences in available therapeutic options and prognosis, an extensive workup is often required to establish the diagnosis. We propose the first echo-based, automated deep learning model with a fusion architecture to facilitate the evaluation and diagnosis of increased left ventricular (LV) wall thickness.

METHODS AND RESULTS

Patients with an established diagnosis of increased LV wall thickness (hypertrophic cardiomyopathy (HCM), cardiac amyloidosis (CA), and hypertensive heart disease (HTN)/others) between 1/2015 and 11/2019 at Mayo Clinic Arizona were identified. The cohort was divided into 80%/10%/10% for training, validation, and testing sets, respectively. Six baseline TTE views were used to optimize a pre-trained InceptionResnetV2 model. Each model output was used to train a meta-learner under a fusion architecture. Model performance was assessed by multiclass area under the receiver operating characteristic curve (AUROC). A total of 586 patients were used for the final analysis (194 HCM, 201 CA, and 191 HTN/others). The mean age was 55.0 years, and 57.8% were male. Among the individual view-dependent models, the apical 4-chamber model had the best performance (AUROC: HCM: 0.94, CA: 0.73, and HTN/other: 0.87). The final fusion model outperformed all the view-dependent models (AUROC: HCM: 0.93, CA: 0.90, and HTN/other: 0.92).

CONCLUSION

The echo-based InceptionResnetV2 fusion model can accurately classify the main etiologies of increased LV wall thickness and can facilitate the process of diagnosis and workup.

Quantifying Valve Regurgitation Using 3-D Doppler Ultrasound Images and Deep Learning

Accurate quantification of cardiac valve regurgitation jets is fundamental for guiding treatment. Cardiac ultrasound is the preferred diagnostic tool, but current methods for measuring the regurgitant volume (RVol) are limited by low accuracy and high interobserver variability. Following recent research, quantitative estimators of orifice size and RVol based on high frame rate 3-D ultrasound have been proposed, but measurement accuracy is limited by the wide point spread function (PSF) relative to the orifice size. The aim of this article was to investigate the use of deep learning to estimate both the orifice size and the RVol. A simulation model was developed to simulate the power-Doppler images of blood flow through orifices with different geometries. A convolutional neural network (CNN) was trained on 30 000 image pairs. The network was used to reconstruct orifices from power-Doppler data, which facilitated estimators for regurgitant orifice areas and flow volumes. We demonstrate that the network improves orifice shape reconstruction, as well as the accuracy of orifice area and flow volume estimation, compared with a previous approach based on thresholding of the power-Doppler signal (THD), and compared with spatially invariant deconvolution (DC). Our approach reduces the area estimation error on simulations: (THD: 13.2 ± 9.9 mm2, DC: 12.8 ± 15.8 mm2, and ours: 3.5 ± 3.2 mm2). In a phantom experiment, our approach reduces both area estimation error (THD: 10.4 ± 8.4 mm2, DC: 10.98 ± 8.17, and ours: 9.9 ± 6.0 mm2) and flow rate estimation error (THD: 20.3 ± 9.9 ml/s, DC: 18.14 ± 13.01 ml/s, and ours: 7.1 ± 10.6 ml/s). We also demonstrate in vivo feasibility for six patients with aortic insufficiency, compared with standard echocardiography and magnetic resonance references.

Motion Estimation by Deep Learning in 2D Echocardiography: Synthetic Dataset and Validation

Motion estimation in echocardiography plays an important role in the characterization of cardiac function, allowing the computation of myocardial deformation indices. However, there exist limitations in clinical practice, particularly with regard to the accuracy and robustness of measurements extracted from images. We therefore propose a novel deep learning solution for motion estimation in echocardiography. Our network corresponds to a modified version of PWC-Net which achieves high performance on ultrasound sequences. In parallel, we designed a novel simulation pipeline allowing the generation of a large amount of realistic B-mode sequences. These synthetic data, together with strategies during training and inference, were used to improve the performance of our deep learning solution, which achieved an average endpoint error of 0.07 ± 0.06 mm per frame and 1.20 ± 0.67 mm between ED and ES on our simulated dataset. The performance of our method was further investigated on 30 patients from a publicly available clinical dataset acquired from a GE system. The method showed promise by achieving a mean absolute error of the global longitudinal strain of 2.5 ± 2.1% and a correlation of 0.77 compared to GLS derived from manual segmentation, much better than one of the most efficient methods in the state-of-the-art (namely the FFT-Xcorr block-matching method). We finally evaluated our method on an auxiliary dataset including 30 patients from another center and acquired with a different system. Comparable results were achieved, illustrating the ability of our method to maintain high performance regardless of the echocardiographic data processed.

Deep learning in ultrasound elastography imaging: A review

It is known that changes in the mechanical properties of tissues are associated with the onset and progression of certain diseases. Ultrasound elastography is a technique to characterize tissue stiffness using ultrasound imaging either by measuring tissue strain using quasi-static elastography or natural organ pulsation elastography, or by tracing a propagated shear wave induced by a source or a natural vibration using dynamic elastography. In recent years, deep learning has begun to emerge in ultrasound elastography research. In this review, several common deep learning frameworks in the computer vision community, such as multilayer perceptron, convolutional neural network, and recurrent neural network are described. Then, recent advances in ultrasound elastography using such deep learning techniques are revisited in terms of algorithm development and clinical diagnosis. Finally, the current challenges and future developments of deep learning in ultrasound elastography are prospected. This article is protected by copyright. All rights reserved.

A Pipeline for the Generation of Synthetic Cardiac Color Doppler

Color Doppler imaging (CDI) is the modality of choice for simultaneous visualization of myocardium and intracavitary flow over a wide scan area. This visualization modality is subject to several sources of error, the main ones being aliasing and clutter. Mitigation of these artifacts is a major concern for better analysis of intracardiac flow. One option to address these issues is through simulations. In this article, we present a numerical framework for generating clinical-like CDI. Synthetic blood vector fields were obtained from a patient-specific computational fluid dynamics CFD model. Realistic texture and clutter artifacts were simulated from real clinical ultrasound cineloops. We simulated several scenarios highlighting the effects of 1) flow acceleration; 2) wall clutter; and 3) transmit wavefronts, on Doppler velocities. As a comparison, an "ideal" color Doppler was also simulated, without these harmful effects. This synthetic dataset is made publicly available and can be used to evaluate the quality of Doppler estimation techniques. Besides, this approach can be seen as a first step toward the generation of comprehensive datasets for training neural networks to improve the quality of Doppler imaging.

Myocardial strain analysis of echocardiography based on deep learning

BACKGROUND

Strain analysis provides more thorough spatiotemporal signatures for myocardial contraction, which is helpful for early detection of cardiac insufficiency. The use of deep learning (DL) to automatically measure myocardial strain from echocardiogram videos has garnered recent attention. However, the development of key techniques including segmentation and motion estimation remains a challenge. In this work, we developed a novel DL-based framework for myocardial segmentation and motion estimation to generate strain measures from echocardiogram videos.

METHODS

Three-dimensional (3D) Convolutional Neural Network (CNN) was developed for myocardial segmentation and optical flow network for motion estimation. The segmentation network was used to define the region of interest (ROI), and the optical flow network was used to estimate the pixel motion in the ROI. We performed a model architecture search to identify the optimal base architecture for motion estimation. The final workflow design and associated hyperparameters are the result of a careful implementation. In addition, we compared the DL model with a traditional speck tracking algorithm on an independent, external clinical data. Each video was double-blind measured by an ultrasound expert and a DL expert using speck tracking echocardiography (STE) and DL method, respectively.

RESULTS

The DL method successfully performed automatic segmentation, motion estimation, and global longitudinal strain (GLS) measurements in all examinations. The 3D segmentation has better spatio-temporal smoothness, average dice correlation reaches 0.82, and the effect of target frame is better than that of previous 2D networks. The best motion estimation network achieved an average end-point error of 0.05 ± 0.03 mm per frame, better than previously reported state-of-the-art. The DL method showed no significant difference relative to the traditional method in GLS measurement, Spearman correlation of 0.90 (p < 0.001) and mean bias -1.2 ± 1.5%.

CONCLUSION

In conclusion, our method exhibits better segmentation and motion estimation performance and demonstrates the feasibility of DL method for automatic strain analysis. The DL approach helps reduce time consumption and human effort, which holds great promise for translational research and precision medicine efforts.

Artificial Intelligence for Automatic Measurement of Left Ventricular Strain in Echocardiography

OBJECTIVES

This study sought to examine if fully automated measurements of global longitudinal strain (GLS) using a novel motion estimation technology based on deep learning and artificial intelligence (AI) are feasible and comparable with a conventional speckle-tracking application.

BACKGROUND

GLS is an important parameter when evaluating left ventricular function. However, analyses of GLS are time consuming and demand expertise, and thus are underused in clinical practice.

METHODS

In this study, 200 patients with a wide range of left ventricle (LV) function were included. Three standard apical cine-loops were analyzed using the AI pipeline. The AI method measured GLS and was compared with a commercially available semiautomatic speckle-tracking software (EchoPAC v202, GE Healthcare, Chicago, Illinois).

RESULTS

The AI method succeeded to both correctly classify all 3 standard apical views and perform timing of cardiac events in 89% of patients. Furthermore, the method successfully performed automatic segmentation, motion estimates, and measurements of GLS in all examinations, across different cardiac pathologies and throughout the spectrum of LV function. GLS was -12.0 ± 4.1% for the AI method and -13.5 ± 5.3% for the reference method. Bias was -1.4 ± 0.3% (95% limits of agreement: 2.3 to -5.1), which is comparable with intervendor studies. The AI method eliminated measurement variability and a complete GLS analysis was processed within 15 s.

CONCLUSIONS

Through the range of LV function this novel AI method succeeds, without any operator input, to automatically identify the 3 standard apical views, perform timing of cardiac events, trace the myocardium, perform motion estimation, and measure GLS. Fully automated measurements based on AI could facilitate the clinical implementation of GLS.