Validation of a deep-learning semantic segmentation approach to fully automate MRI-based left-ventricular deformation analysis in cardiotoxicity

2021 ◽  
Vol 94 (1120) ◽  
pp. 20201101
Author(s):  
Julia Karr ◽  
Michael Cohen ◽  
Samuel A McQuiston ◽  
Teja Poorsala ◽  
Christopher Malozzi
Keyword(s):  
Breast Cancer ◽  
Deep Learning ◽  
Repeated Measures ◽  
Performance Metrics ◽  
Intraclass Correlation ◽  
Myocardial Strain ◽  
Strain Analysis ◽  
Left Ventricular ◽  
Cancer Data ◽  
Left Ventricular Deformation

Objective: Left-ventricular (LV) strain measurements with the Displacement Encoding with Stimulated Echoes (DENSE) MRI sequence provide accurate estimates of cardiotoxicity damage related to chemotherapy for breast cancer. This study investigated an automated and supervised deep convolutional neural network (DCNN) model for LV chamber quantification before strain analysis in DENSE images. Methods: The DeepLabV3 +DCNN with three versions of ResNet-50 backbone was designed to conduct chamber quantification on 42 female breast cancer data sets. The convolutional layers in the three ResNet-50 backbones were varied as non-atrous, atrous and modified, atrous with accuracy improvements like using Laplacian of Gaussian filters. Parameters such as LV end-diastolic diameter (LVEDD) and ejection fraction (LVEF) were quantified, and myocardial strains analyzed with the Radial Point Interpolation Method (RPIM). Myocardial classification was validated with the performance metrics of accuracy, Dice, average perpendicular distance (APD) and others. Repeated measures ANOVA and intraclass correlation (ICC) with Cronbach’s α (C-Alpha) tests were conducted between the three DCNNs and a vendor tool on chamber quantification and myocardial strain analysis. Results: Validation results in the same test-set for myocardial classification were accuracy = 97%, Dice = 0.92, APD = 1.2 mm with the modified ResNet-50, and accuracy = 95%, Dice = 0.90, APD = 1.7 mm with the atrous ResNet-50. The ICC results between the modified ResNet-50, atrous ResNet-50 and vendor-tool were C-Alpha = 0.97 for LVEF (55±7%, 54±7%, 54±7%, p = 0.6), and C-Alpha = 0.87 for LVEDD (4.6 ± 0.3 cm, 4.6 ± 0.3 cm, 4.6 ± 0.4 cm, p = 0.7). Conclusion: Similar performance metrics and equivalent parameters obtained from comparisons between the atrous networks and vendor tool show that segmentation with the modified, atrous DCNN is applicable for automated LV chamber quantification and subsequent strain analysis in cardiotoxicity. Advances in knowledge: A novel deep-learning technique for segmenting DENSE images was developed and validated for LV chamber quantification and strain analysis in cardiotoxicity detection.

Abstract P345: A MRI Displacement-based Deep-learning Semantic Segmentation Tool For Left-ventricular Longitudinal Strain Analysis In Cardiotoxicity

2021 ◽  
Vol 129 (Suppl_1) ◽  
Author(s):  
Julia Kar ◽  
Michael V Cohen ◽  
Teja Poorsala ◽  
Samuel A McQuiston ◽  
Cheri Revere ◽  
...  
Keyword(s):  
Breast Cancer ◽  
Deep Learning ◽  
Cancer Patients ◽  
Healthy Subjects ◽  
Longitudinal Strain ◽  
Strain Analysis ◽  
Semantic Segmentation ◽  
Left Ventricular ◽  
Phase Unwrapping ◽  
Breast Cancer Patients

Global longitudinal strain (GLS) computed in the left-ventricle (LV) is an established metric for detecting cardiotoxicity in breast cancer patients treated with antineoplastic agents. The purpose of this study was to develop a novel, MRI-based, deep-learning semantic segmentation tool that automates the phase-unwrapping for LV displacement computation in GLS. Strain analysis via phase-unwrapping was conducted on 30 breast cancer patients and 30 healthy females acquired with the Displacement Encoding with Stimulated Echoes (DENSE) sequence. A ResNet-50 deep convolutional neural network (DCNN) architecture for automated phase-unwrapping, a previously validated ResNet-50 DCNN for chamber quantification and the Radial Point Interpolation Method were used for GLS computation (Figure 1). The DCNN's performance was measured with F1 and Dice scores, and validated in comparison to the robust transport of intensity equation (RTIE) and quality guided phase-unwrapping (QGPU) conventional algorithms. The three techniques were compared by intraclass correlation coefficient with Cronbach’s alpha (C-alpha) index. Classification accuracy with the DCNN was F1 score of 0.92 and Dice score of 0.89. The GLS results from RTIE, QGPU and DCNN were -16.0 ± 2%, -16.1 ± 3% and -15.9 ± 3% (C-alpha = 0.89) for patients and -18.9 ± 3%, -19.0 ± 4% and -18.9 ± 3% (C-alpha = 0.92) for healthy subjects. Comparable validation results from the three techniques show the feasibility of a DCNN-based approach to LV displacement and GLS analysis. The dissimilarities between patients and healthy subjects demonstrate that DCNN-based GLS computation may detect LV abnormalities related to cardiotoxicity.

scholarly journals Direct comparison of multilayer left ventricular global longitudinal strain using CMR feature tracking and speckle tracking echocardiography

2021 ◽  
Vol 21 (1) ◽  
Author(s):  
Saikrishna Ananthapadmanabhan ◽  
Giau Vo ◽  
Tuan Nguyen ◽  
Hany Dimitri ◽  
James Otton
Keyword(s):  
Feature Tracking ◽  
Speckle Tracking ◽  
Longitudinal Strain ◽  
Speckle Tracking Echocardiography ◽  
Imaging Modality ◽  
Global Longitudinal Strain ◽  
Intraclass Correlation ◽  
Strain Analysis ◽  
Left Ventricular ◽  
Multilayer Strain

Abstract Background Cardiac magnetic resonance feature tracking (CMR-FT) and speckle tracking echocardiography (STE) are well-established strain imaging modalities. Multilayer strain measurement permits independent assessment of endocardial and epicardial strain. This novel and layer specific approach to evaluating myocardial deformation parameters may provide greater insight into cardiac contractility when compared to whole-layer strain analysis. The aim of this study is to validate CMR-FT as a tool for multilayer strain analysis by providing a direct comparison between multilayer global longitudinal strain (GLS) values between CMR-FT and STE. Methods We studied 100 patients who had an acute myocardial infarction (AMI), who underwent CMR imaging and echocardiogram at baseline and follow-up (48 ± 13 days). Dedicated tissue tracking software was used to analyse single- and multi-layer GLS values for CMR-FT and STE. Results Correlation coefficients for CMR-FT and STE were 0.685, 0.687, and 0.660 for endocardial, epicardial, and whole-layer GLS respectively (all p < 0.001). Bland Altman analysis showed good inter-modality agreement with minimal bias. The absolute limits of agreement in our study were 6.4, 5.9, and 5.5 for endocardial, whole-layer, and epicardial GLS respectively. Absolute biases were 1.79, 0.80, and 0.98 respectively. Intraclass correlation coefficient (ICC) values showed moderate agreement with values of 0.626, 0.632, and 0.671 respectively (all p < 0.001). Conclusion There is good inter-modality agreement between CMR-FT and STE for whole-layer, endocardial, and epicardial GLS, and although values should not be used interchangeably our study demonstrates that CMR-FT is a viable imaging modality for multilayer strain

Abstract 224: A Deep Learning Approach to Left-Ventricular Chamber Quantification for Fully Automated Three Dimensional Strain Analysis in Cardiotoxicity

2020 ◽  
Vol 127 (Suppl_1) ◽  
Author(s):  
Bryant M Baldwin ◽  
Shane Joseph ◽  
Xiaodong Zhong ◽  
Ranya Kakish ◽  
Cherie Revere ◽  
...  
Keyword(s):  
Deep Learning ◽  
Interpolation Method ◽  
Three Dimensional ◽  
Strain Analysis ◽  
Correlation Coefficients ◽  
Semantic Segmentation ◽  
Left Ventricular ◽  
Breast Cancer Patients ◽  
Left Ventricular Chamber ◽  
Longitudinal Strains

This study investigated MRI and semantic segmentation-based deep-learning (SSDL) automation for left-ventricular chamber quantifications (LVCQ) and low longitudinal strain (LLS) determination, thus eliminating user-bias by providing an automated tool to detect cardiotoxicity (CT) in breast cancer patients treated with antineoplastic agents. Displacement Encoding with Stimulated Echoes-based (DENSE) myocardial images from 26 patients were analyzed with the tool’s Convolution Neural Network with underlying Resnet-50 architecture. Quantifications based on the SSDL tool’s output were for LV end-diastolic diameter (LVEDD), ejection fraction (LVEF), and mass (LVM) (see figure for phase sequence). LLS was analyzed with Radial Point Interpolation Method (RPIM) with DENSE phase-based displacements. LVCQs were validated by comparison to measurements obtained with an existing semi-automated vendor tool (VT) and strains by 2 independent users employing Bland-Altman analysis (BAA) and interclass correlation coefficients estimated with Cronbach’s Alpha (C-Alpha) index. F1 score for classification accuracy was 0.92. LVCQs determined by SSDL and VT were 4.6 ± 0.5 vs 4.6 ± 0.7 cm (C-Alpha = 0.93 and BAA = 0.5 ± 0.5 cm) for LVEDD, 58 ± 5 vs 58 ± 6 % (0.90, 1 ± 5%) for LVEF, 119 ± 17 vs 121 ± 14 g (0.93, 5 ± 8 g) for LV mass, while LLS was 14 ± 4 vs 14 ± 3 % (0.86, 0.2 ± 6%). Hence, equivalent LV dimensions, mass and strains measured by VT and DENSE imaging validate our unique automated analytic tool. Longitudinal strains in patients can then be analyzed without user bias to detect abnormalities for the indication of cardiotoxicity and the need for therapeutic intervention even if LVEF is not affected.

The quality of breast cancer quality measures.

2018 ◽  
Vol 36 (30_suppl) ◽  
pp. 108-108
Author(s):  
Whitney Lane ◽  
Christel Rushing ◽  
Daniel Nussbaum ◽  
Dan G. Blazer ◽  
Rachel Adams Greenup
Keyword(s):  
Breast Cancer ◽  
Patient Preference ◽  
Cancer Care ◽  
Performance Metrics ◽  
Prophylactic Mastectomy ◽  
Patient Choice ◽  
Breast Cancer Care ◽  
Cancer Data ◽  
Breast Care

108 Background: To assess quality in breast cancer care, standardized metrics are needed. Many accepted breast performance metrics are based on evidence-based practice; however, most fail to reflect patient choice in treatment decisions. Given the focus on patient-centered breast care, we sought to determine how compliance with established quality metrics correlates with receipt of breast cancer care impacted by patient preference. Methods: American College of Surgeons (ACS) National Cancer Data Base facilities were designated compliant or non-compliant based on Commission of Cancer (CoC) breast metrics MASTRT, BCSRT and HT*, which all improve survival. Compliant facilities met the expected performance rate (EPR) for all three metrics, while non-compliant facilities failed to meet the EPR for any. Rates of breast conserving surgery (BCS) for early stage cancer, immediate breast reconstruction (IBR), and contralateral prophylactic mastectomy (CPM) are proposed metrics that are impacted by patient preference. For these, quality is defined as high rates of BCS, high rates of IBR, and low rates of CPM. Multivariable logistic regression models were used to estimate the association between facility level rates on these measures and the probability of treatment at a CoC compliant facility. Results: 729 facilities were included in the analysis. Based on the CoC measures, 79 (10.8%) were considered compliant and 650 (89.2%) non-compliant. Rates of BCS and IBR did not differ between compliant and non-compliant facilities; however, women treated at compliant facilities were more likely to undergo CPM (26.3% vs 21.4%; p = 0.02). In a multivariate model treatment at compliant facilities was associated with higher rates of BCS, IBR, and CPM; however, the predictive value of these metrics was minimal (Estimated OR range: 1.01-1.03). Conclusions: Rates of preference driven therapies do not differentiate CoC compliant and non-compliant hospitals. The quality of a hospital’s breast care is likely poorly measured by metrics that are influenced by, but cannot account for patient values. *MASTRT (RT≤1yr of diagnosis in women with ≥4 +lymph nodes); BCSRT (RT ≤1yr of diagnosis for women ≤70 receiving BCS); HT (hormone therapy recommended ≤1yr of diagnosis for HR-positive breast cancer)

scholarly journals High interstudy repeatability of automatic deep learnt biventricular CMR measurements

2021 ◽  
Vol 22 (Supplement_2) ◽  
Author(s):  
S Alabed ◽  
K Karunasaagarar ◽  
F Alandejani ◽  
P Garg ◽  
J Uthoff ◽  
...  
Keyword(s):  
Pulmonary Hypertension ◽  
Deep Learning ◽  
Automatic Segmentation ◽  
Intraclass Correlation ◽  
Therapy Response ◽  
Left Ventricular ◽  
Dice Similarity Coefficient ◽  
Learning Approaches ◽  
Fully Automatic ◽  
Automatic Contouring

Abstract Funding Acknowledgements Type of funding sources: Foundation. Main funding source(s): Wellcome Trust (UK), NIHR (UK) Introduction Cardiac magnetic resonance (CMR) measurements have significant diagnostic and prognostic value. Accurate and repeatable measurements are essential to assess disease severity, evaluate therapy response and monitor disease progression. Deep learning approaches have shown promise for automatic left ventricular (LV) segmentation on CMR, however fully automatic right ventricular (RV) segmentation remains challenging. We aimed to develop a biventricular automatic contouring model and evaluate the interstudy repeatability of the model in a prospectively recruited cohort. Methods A deep learning CMR contouring model was developed in a retrospective multi-vendor (Siemens and General Electric), multi-pathology cohort of patients, predominantly with heart failure, pulmonary hypertension and lung diseases (n = 400, ASPIRE registry). Biventricular segmentations were made on all CMR studies across cardiac phases. To test the accuracy of the automatic segmentation, 30 ASPIRE CMRs were segmented independently by two CMR experts. Each segmentation was compared to the automatic contouring with agreement assessed using the Dice similarity coefficient (DSC).  A prospective validation cohort of 46 subjects (10 healthy volunteers and 36 patients with pulmonary hypertension) were recruited to assess interstudy agreement of automatic and manual CMR assessments. Two CMR studies were performed during separate sessions on the same day. Interstudy repeatability was assessed using intraclass correlation coefficient (ICC) and Bland-Altman plots.  Results DSC showed high agreement (figure 1) comparing automatic and expert CMR readers, with minimal bias towards either CMR expert. The scan-scan repeatability CMR measurements were higher for all automatic RV measurements (ICC 0.89 to 0.98) compared to manual RV measurements (0.78 to 0.98). LV automatic and manual measurements were similarly repeatable (figure 2). Bland-Altman plots showed strong agreement with small mean differences between the scan-scan measurements (figure 2). Conclusion Fully automatic biventricular short-axis segmentations are comparable with expert manual segmentations, and have shown excellent interstudy repeatability.

scholarly journals Heart transplantation and antibody-mediated rejection: role of myocardial strain as an early marker of cardiac dysfunction in patients with anti-HLA antibody

2021 ◽  
Vol 42 (Supplement_1) ◽  
Author(s):  
C Sciaccaluga ◽  
B.M Natali ◽  
G.E Mandoli ◽  
N Sisti ◽  
F.M Righini ◽  
...  
Keyword(s):  
Cardiac Dysfunction ◽  
Longitudinal Strain ◽  
Global Longitudinal Strain ◽  
Myocardial Strain ◽  
Strain Analysis ◽  
Left Ventricular ◽  
Hla Antibodies ◽  
Antibody Mediated Rejection ◽  
Right Ventricular Strain ◽  
Hla Antibody

Abstract Background Antibody-mediated rejection of the transplanted heart is still currently diagnosed by endomyocardial biopsy whereas clinical elements, anti-Human Leukocite Antigens (HLA) antibody and graft dysfunction represents supplementary components. Purpose The aim of the study was to identify though a non-invasive imaging technique, such as advanced transthoracic echocardiography, early signs of altered cardiac function in patients with anti-HLA antibodies and no histological signs of antibody-mediated rejection. Methods The study population included 117 heart transplanted patients, in whom both acute and chronic rejection was excluded. They were divided into two groups “HLA+`' (45 patients) and “HLA−” (72 patients), based on the presence and the absence of circulating anti-HLA antibodies, respectively. The echocardiographic exam was performed within one week from the biopsy, including Speckle Tracking analysis. Results Deceleration Time of E wave was the strongest traditional echocardiographic parameter which correlated with circulating anti-HLA antibodies (165±39,5 vs 196,5±25; p&lt;0.001). Regarding strain analysis, both left ventricular global longitudinal strain (−16,1±3,4 vs −19,8±2,0; p&lt;0.001) and right ventricular strain (−17,2±0,7 vs −20,6±0,5; p=0.0002) differed significantly between the two subgroups (Figure 1). On the other hand, neither peak atrial longitudinal strain nor peak atrial contraction strain showed a significant correlation with anti-HLA antibodies. Conclusion The presence of circulating anti-HLA antibodies seems to be correlated with a mild cardiac dysfunction, even in the absence of antibody-mediated rejection. This subtle dysfunction is not completely detectable by standard echocardiographic parameters, whereas strain analysis has showed promising results since it revealed more clearly an impaired function of both ventricles in heart transplanted HLA+ patients, with potentially important clinical repercussion. FUNDunding Acknowledgement Type of funding sources: None. Figure 1

P5253Does regional myocardial strain by cardiac magnetic resonance feature tracking reflect scar in ischemic heart disease?

2019 ◽  
Vol 40 (Supplement_1) ◽  
Author(s):  
K Stathogiannis ◽  
V Mor-Avi ◽  
R Lang ◽  
A R Patel
Keyword(s):  
Cardiac Magnetic Resonance ◽  
Magnetic Resonance ◽  
Feature Tracking ◽  
Circumferential Strain ◽  
Myocardial Strain ◽  
Strain Analysis ◽  
Left Ventricular ◽  
Lv Function ◽  
Regional Strain ◽  
The Relationship

Abstract Background Cardiac magnetic resonance (CMR) late gadolinium enhancement (LGE) is the gold standard for detection of myocardial scar. We hypothesized that CMR Feature Tracking (FT)-derived regional myocardial strain may reflect the presence of scar and could thus potentially be used instead of LGE imaging. Purpose The aim of this study was to determine the relationship between FT-derived regional myocardial strain and LGE in patients with coronary artery disease (CAD). Methods Seventy-five patients with CAD and typical ischemic LGE patterns on CMR (1.5T) were included (mean age 60±12 years, 70% males). Myocardial strain analysis and LGE identification were performed using dedicated commercial software. Scar was defined by presence of LGE in the same area of the myocardium in both short- and long-axis views. Peak systolic regional longitudinal and circumferential strain (RLS, RCS) values were calculated in the region of interest corresponding to the LGE area and also in a non-LGE myocardial region as a reference in each patient. These comparisons were repeated for a subgroup of 36 patients with left ventricular (LV) ejection fraction (EF) <40% to determine whether the relationship between strain and LGE holds in the presence of reduced LV function, when strain measurements may be altered as a reflection of reduced LVEF itself. Results Both global longitudinal and circumferential strain values were abnormal (−12.8±5.1% and −11.4±4.1%, respectively), reflecting LV dysfunction in this CAD cohort (EF = 40±16%). The magnitude of both RLS and RCS was significantly reduced in areas of LGE, compared to those without LGE: RLS −10.0±5.8% versus −20.4±7.5% (p<0.001); RCS −10.1±5.3±% versus −18.9±7.5%, respectively (p<0.001). Same pattern was noted in the reduced EF subgroup: RLS −8.0±4.7% versus −16.9±6.6% (p<0.001), RCS −7.7±4.3±% versus −16.0±7.9%, respectively (p<0.001). The figure depicts 2 representative cases in long and short axis views, LGE detection and concomitant regional strain analysis. LGE and regional strain analysis. Conclusion Reduced magnitude of regional longitudinal and circumferential strain by CMR-FT correlates with presence of LGE. Pending further validation, this finding may constitute the basis for detection of scar without contrast enhanced imaging, and would result in reduced cost, scan time and risk associated with gadolinium. Acknowledgement/Funding ARP: Research support (software) from Neosoft and Philips

scholarly journals CMR-derived myocardial strain analysis differentiates ischemic and dilated cardiomyopathy - A propensity score-matched study

2021 ◽  
Author(s):  
Julia Vietheer ◽  
Lehmann Lena ◽  
Claudia Unbehaun ◽  
Ullrich Fischer-Rasokat ◽  
Jan Sebastian Wolter ◽  
...  
Keyword(s):  
Propensity Score ◽  
Dilated Cardiomyopathy ◽  
Propensity Score Matching ◽  
Feature Tracking ◽  
Global Longitudinal Strain ◽  
Myocardial Strain ◽  
Radial Strain ◽  
Strain Analysis ◽  
Left Ventricular ◽  
Myocardial Strain Analysis

Abstract Purpose Left ventricular (LV) longitudinal, circumferential, and radial motion can be measured using feature tracking of cardiac magnetic resonance (CMR) images. The aim of our study was to detect differences in LV mechanics between patients with dilated cardiomyopathy (DCM) and ischemic cardiomyopathy (ICM) who were matched using a propensity score-based model. Methods Between April 2017 and October 2019, 1224 patients were included in our CMR registry, among them 141 with ICM and 77 with DCM. Propensity score matching was used to pair patients based on their indexed end-diastolic volume (EDVi), ejection fraction (EF), and septal T1 relaxation time. Feature tracking provided six parameters for global longitudinal, circumferential, and radial strain with corresponding strain rates. Results Propensity score matching yielded 72 patients in each group (DCM mean age 58.6 ± 11.6 years, 15 females; ICM mean age 62.6 ± 13.2 years, 11 females, p = 0.084 and 0.44 respectively; LV-EF 32.2 ± 13.5% vs. 33.8 ± 12.1%, p = 0.356; EDVi 127.2 ± 30.7 ml/m² vs. 121.1 ± 41.8 ml/m², p = 0.251; native T1 values 1165 ± 58 ms vs. 1167 ± 70 ms, p = 0.862). There was no difference in global longitudinal strain between DCM and ICM patients (-10.9 ± 5.5% vs. -11.2 ± 4.7%, p = 0.72), whereas in DCM patients there was a significant reduction in global circumferential strain (-10.0 ± 4.5% vs. -12.2 ± 4.7%, p = 0.002) and radial strain (17.1 ± 8.51 vs. 21.2 ± 9.7%, p = 0.039). Conclusion Our data suggest that ICM and DCM patients have inherently different myocardial mechanics, even if phenotypes are similar. The ability to discriminate these two conditions may aid in developing additional prognostic and therapeutic strategies in the future.

Sign in / Sign up

Export Citation Format

Share Document