Automated segmentation of the left-ventricle from MRI with a fully convolutional network to investigate CTRCD in breast cancer patients

Purpose: The goal of this study was to develop a fully convolutional network (FCN) tool to automatedly segment the left-ventricular (LV) myocardium in displacement encoding with stimulated echoes MRI. The segmentation results are used for LV chamber quantification and strain analyses in breast cancer patients susceptible to cancer therapy-related cardiac dysfunction (CTRCD). Approach: A DeepLabV3+ FCN with a ResNet-101 backbone was custom-designed to conduct chamber quantification on 45 female breast cancer datasets (23 training, 11 validation, and 11 test sets). LV structural parameters and LV ejection fraction (LVEF) were measured, and myocardial strains estimated with the radial point interpolation method. Myocardial classification validation was against quantization-based ground-truth with computations of accuracy, Dice score, average perpendicular distance (APD), Hausdorff-distance, and others. Additional validations were conducted with equivalence tests and Cronbach's alpha (C - α ) intraclass correlation coefficients between the FCN and a vendor tool on chamber quantification and myocardial strain computations. Results: Myocardial classification results against ground-truth were Dice = 0.89 , APD = 2.4    mm , and accuracy = 97 % for the validation set and Dice = 0.90 , APD = 2.5    mm , and accuracy = 97 % for the test set. The confidence intervals (CI) and two one-sided t-test results of equivalence tests between the FCN and vendor-tool were CI = - 1.36 % to 2.42%, p-value < 0.001 for LVEF (58 ± 5 % versus 57 ± 6 % ), and CI = - 0.71 % to 0.63%, p-value < 0.001 for longitudinal strain (- 15 ± 2 % versus - 15 ± 3 % ). Conclusions: The validation results were found equivalent to the vendor tool-based parameter estimates, which show that accurate LV chamber quantification followed by strain analysis for CTRCD investigation can be achieved with our proposed FCN methodology.

Can global longitudinal strain (GLS) with magnetic resonance prognosticate early cancer therapy-related cardiac dysfunction (CTRCD) in breast cancer patients, a prospective study?

PURPOSE

To determine if Artificial Intelligence-based computation of global longitudinal strain (GLS) from left ventricular (LV) MRI is an early prognostic factor of cancer therapy-related cardiac dysfunction (CTRCD) in breast cancer patients. The main hypothesis based on the patients receiving antineoplastic chemotherapy treatment was CTRCD risk analysis with GLS that was independent of LV ejection fraction (LVEF).

METHODS

Displacement Encoding with Stimulated Echoes (DENSE) MRI was acquired on 32 breast cancer patients at baseline and 3- and 6-month follow-ups after chemotherapy. Two DeepLabV3+ Fully Convolutional Networks (FCNs) were deployed to automate image segmentation for LV chamber quantification and phase-unwrapping for 3D strains, computed with the Radial Point Interpolation Method. CTRCD risk (cardiotoxicity and adverse cardiac events) was analyzed with Cox Proportional Hazards (PH) models with clinical and contractile prognostic factors.

RESULTS

GLS worsened from baseline to the 3- and 6-month follow-ups (-19.1 ± 2.1%, -16.0 ± 3.1%, -16.1 ± 3.0%; P < 0.001). Univariable Cox regression showed the 3-month GLS significantly associated as an agonist (hazard ratio [HR]-per-SD: 2.1; 95% CI: 1.4-3.1; P < 0.001) and LVEF as a protector (HR-per-SD: 0.8; 95% CI: 0.7-0.9; P = 0.001) for CTRCD occurrence. Bivariable regression showed the 3-month GLS (HR-per-SD: 2.0; 95% CI: 1.2-3.4; P = 0.01) as a CTRCD prognostic factor independent of other covariates, including LVEF (HR-per-SD: 1.0; 95% CI: 0.9-1.2; P = 0.9).

CONCLUSIONS

The end-point analyses proved the hypothesis that GLS is an early, independent prognosticator of incident CTRCD risk. This novel GLS-guided approach to CTRCD risk analysis could improve antineoplastic treatment with further validation in a larger clinical trial.

Direct left-ventricular global longitudinal strain (GLS) computation with a fully convolutional network

This study's purpose was to develop a direct MRI-based, deep-learning semantic segmentation approach for computing global longitudinal strain (GLS), a known metric for detecting left-ventricular (LV) cardiotoxicity in breast cancer. Displacement Encoding with Stimulated Echoes cardiac image phases acquired from 30 breast cancer patients and 30 healthy females were unwrapped via a DeepLabV3 + fully convolutional network (FCN). Myocardial strains were directly computed from the unwrapped phases with the Radial Point Interpolation Method. FCN-unwrapped phases of a phantom's rotating gel were validated against quality-guided phase-unwrapping (QGPU) and robust transport of intensity equation (RTIE) phase-unwrapping. FCN performance on unwrapping human LV data was measured with F1 and Dice scores versus QGPU ground-truth. The reliability of FCN-based strains was assessed against RTIE-based strains with Cronbach's alpha (C-α) intraclass correlation coefficient. Mean squared error (MSE) of unwrapping the phantom experiment data at 0 dB signal-to-noise ratio were 1.6, 2.7 and 6.1 with FCN, QGPU and RTIE techniques. Human data classification accuracies were F1 = 0.95 (Dice = 0.96) with FCN and F1 = 0.94 (Dice = 0.95) with RTIE. GLS results from FCN and RTIE were -16 ± 3% vs. -16 ± 3% (C-α = 0.9) for patients and -20 ± 3% vs. -20 ± 3% (C-α = 0.9) for healthy subjects. The low MSE from the phantom validation demonstrates accuracy of phase-unwrapping with the FCN and comparable human subject results versus RTIE demonstrate GLS analysis accuracy. A deep-learning methodology for phase-unwrapping in medical images and GLS computation was developed and validated in a heterogeneous cohort.

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

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.

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

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.

Fully automated and comprehensive MRI-based left-ventricular contractility analysis in post-chemotherapy breast cancer patients

OBJECTIVE

This study investigated the occurrence of cardiotoxicity-related left-ventricular (LV) contractile dysfunction in breast cancer patients following treatment with antineoplastic chemotherapy agents.

METHODS

A validated and automated MRI-based LV contractility analysis tool consisting of quantization-based boundary detection, unwrapping of image phases and the meshfree Radial Point Interpolation Method was used toward measuring LV chamber quantifications (LVCQ), three-dimensional strains and torsions in patients and healthy subjects. Data were acquired with the Displacement Encoding with Stimulated Echoes (DENSE) sequence on 21 female patients and 21 age-matched healthy females. Estimates of patient LVCQs from DENSE acquisitions were validated in comparison to similar steady-state free precession measurements and their strain results validated via Bland-Altman interobserver agreements. The occurrence of LV abnormalities was investigated via significant differences in contractility measurements (LVCQs, strains and torsions) between patients and healthy subjects.

RESULTS

Repeated measures analysis showed similarities between LVCQ measurements from DENSE and steady-state free precession, including cardiac output (4.7 ± 0.4 L, 4.6 ± 0.4 L, p = 0.8), and LV ejection fractions (59±6%, 58±5%, p = 0.2). Differences found between patients and healthy subjects included enlarged basal diameter (5.0 ± 0.5 cm vs 4.4 ± 0.5 cm, p < 0.01), apical torsion (6.0 ± 1.1° vs 9.7 ± 1.4°, p < 0.001) and global longitudinal strain (-0.15 ± 0.02 vs. -0.21 ± 0.04, p < 0.001), but not LV ejection fraction (59±6% vs. 63±6%, p = 0.1).

CONCLUSION

The results from the statistical analysis reveal the possibility of LV abnormalities in the post-chemotherapy patients via enlarged basal diameter and reduced longitudinal strain and torsion, in comparison to healthy subjects.

ADVANCES IN KNOWLEDGE

This study shows that subclinical LV abnormalities in post-chemotherapy breast cancer patients can be detected with an automated technique for the comprehensive analysis of contractile parameters.

Abstract 174: Investigating Subclinical Cardiotoxicity in Long-term Breast Cancer Survivors Following Chemotherapy With DENSE-based 3D Strain Analysis

This study investigated if left ventricular (LV) mechanical contractile parameters in addition to LV ejection fraction (LVEF) indicated cardiac remodeling and fibrosis in patients exposed to cardiotoxic chemotherapy agents (CCA). Cardiac deformation data were obtained using single-scan acquisitions with the Displacement Encoding with Stimulated Echoes (DENSE) MRI sequence. Contractile analysis consisted of automated myocardial boundary detection and unwrapping 3D DENSE phase images for intra-myocardial displacements, and followed by analyzing torsion and 3D strains with the meshfree Radial Point Interpolation Method (RPIM). Data were acquired on 13 CCA-exposed patients who were undergoing chemotherapy and/or care for cardiac complications. DENSE LVEF measurements in patients were validated against steady-state free precession (SSFP) MRI data, all contractility computations were compared to healthy subjects and Bland-Altman agreements established between strains computed by independent observers. A significant difference was not found between DENSE and SSFP LVEF computations (52 ± 11% vs 48 ± 15%, p=0.33). Significant differences were seen with enlarged LV diameters in patients versus healthy subjects (6.0 ± 1.1 cm vs 4.9 ± 0.7 cm, p<0.001) and similarly for torsion and longitudinal strain but not in LVEF results (p > 0.05). Bland-Altman agreements were 0.01 ± 0.06 for longitudinal strain and 0.1 ± 1.9 ° for torsion. Statistical analysis confirm changed LV size and function in patients and indicate remodeling that is otherwise not demonstrated by LVEF measurements, which may precede clinically apparent heart failure development and worsened morbidity.

Can post-chemotherapy cardiotoxicity be detected in long-term survivors of breast cancer via comprehensive 3D left-ventricular contractility (strain) analysis?

PURPOSE

This study applied a novel and automated contractility analysis tool to investigate possible cardiotoxicity-related left-ventricular (LV) dysfunction in breast cancer patients following treatment with anti-neoplastic chemotherapy agents (CTA). Subclinical dysfunction otherwise undetected via LV ejection fraction (LVEF) was determined.

METHODS

Deformation data were acquired with the Displacement Encoding with Stimulated Echoes (DENSE) MRI sequence on 16 female patients who had CTA-based treatment. The contractility analysis tool consisting of image quantization-based boundary detection and the meshfree Radial Point Interpolation Method was used to compare chamber quantifications, 3D regional strains and torsion between patients and healthy subjects (N = 26 females with N = 14 age-matched). Quantifications of patient LVEFs from DENSE and Steady-State Free Precession (SSFP) acquisitions were compared, Bland-Altman interobserver agreements measured on their strain results and differences in contractile parameters with healthy subjects determined via Student's t-tests.

RESULTS

A significant difference was not found between DENSE and SSFP-based patient LVEFs at 58 ± 7% vs 57 ± 9%, p = 0.6. Bland-Altman agreements were - 0.01 ± 0.05 for longitudinal strain and 0.1 ± 1.3° for torsion. Differences in basal diameter indicating enlargement, 5.2 ± 0.5 cm vs 4.5 ± 0.5 cm, p < 0.01, and torsion, 4.7 ± 1.0° vs 8.1 ± 1.1°, p < 0.001 in the mid-ventricle and 5.9 ± 1.2° vs 10.2 ± 0.9°, p < 0.001 apically, were seen between patients and age-matched healthy subjects and similarly in longitudinal strain, but not in LVEF.

CONCLUSIONS

Results from the statistical analysis reveal the likelihood of LV remodeling in this patient subpopulation otherwise not indicated by LVEF measurements.

Recurrent hepatoblastoma with localization by PET-CT

Hepatoblastoma is the most common primary liver tumor in children, accounting for 79% of pediatric liver malignancies in children younger than 15 years, with most cases reported before the age of 5 years. Localization of primary and recurrent disease is necessary for appropriate clinical decision-making and treatment. We present a case of recurrent hepatoblastoma heralded by rising alpha-fetoprotein levels. After unsuccessful localization by conventional CT and MRI, positron emission tomography CT imaging localized the sites of recurrence.

Adeno-associated virus Rep proteins target DNA sequences to a unique locus in the human genome

We have developed a system for site-specific DNA integration in human cells, mediated by the adeno-associated virus (AAV) Rep proteins. In its normal lysogenic cycle, AAV integrates at a site on human chromosome 19 termed AAVS1. We describe a rapid PCR assay for the detection of integration events at AAVS1 in whole populations of cells. Using this assay, we determined that the AAV Rep proteins, delivered in cis or trans, are required for integration at AAVS1. Only the large forms of the Rep protein, Rep78 and Rep68, promoted site-specific integration. The AAV inverted terminal repeats, present in cis, were not essential for integration at AAVS1, but in cells containing Rep, they increased the efficiency of integration. In the presence of the Rep proteins, the integration of a plasmid containing AAV inverted terminal repeats occurred at high frequency, such that clones containing the plasmid could be isolated without selection. In two of the five clones analyzed by fluorescence in situ hybridization, the plasmid DNA was integrated at AAVS1. In most of the clones, at least one copy of the entire plasmid was integrated in a tandem array. Detailed analysis of the integrated plasmid structure in one clone suggested a complex mechanism producing rearrangements of the flanking genomic DNA, similar to those observed with wild-type AAV.

Recombinant adeno-associated virus mediates a high level of gene transfer but less efficient integration in the K562 human hematopoietic cell line

We tested the ability of a recombinant adeno-associated virus (rAAV) vector to express and integrate exogenous DNA into human hematopoietic cells in the absence of selection. We developed an rAAV vector, AAV-tNGFR, carrying a truncated rat nerve growth factor receptor (tNGFR) cDNA as a cell surface reporter under the control of the Moloney murine leukemia virus (MoMuLV) long terminal repeat. An analogous MoMuLV-based retroviral vector (L-tNGFR) was used in parallel, and gene transfer and expression in human hematopoietic cells were assessed by flow cytometry and DNA analyses. Following gene transfer into K562 cells with AAV-tNGFR at a multiplicity of infection (MOI) of 13 infectious units (IU), 26 to 38% of cells expressed tNGFR on the surface early after transduction, but the proportion of tNGFR expressing cells steadily declined to 3.0 to 3.5% over 1 month of culture. At an MOI of 130 IU, nearly all cells expressed tNGFR immediately posttransduction, but the proportion of cells expressing tNGFR declined to 62% over 2 months of culture. The decline in the proportion of AAV-tNGFR-expressing cells was associated with ongoing losses of vector genomes. In contrast, K562 cells transduced with the retroviral vector L-tNGFR expressed tNGFR in a constant fraction. Integration analyses on clones showed that integration occurred at different sites. Integration frequencies were estimated at about 49% at an MOI of 130 and 2% at an MOI of 1.3. Transduction of primary human CD34+ progenitor cells by AAV-tNGFR was less efficient than with K562 cells and showed a declining percentage of cells expressing tNGFR over 2 weeks of culture. Thus, purified rAAV caused very high gene transfer and expression in human hematopoietic cells early after transduction, which steadily declined during cell passage in the absence of selection. Although the efficiency of integration was low, overall integration was markedly improved at a high MOI. While prolonged episomal persistence may be adequate for gene therapy of nondividing cells, a very high MOI or improvements in basic aspects of AAV-based vectors may be necessary to improve integration frequency in the rapidly dividing hematopoietic cell population.

Gene delivery to skeletal muscle results in sustained expression and systemic delivery of a therapeutic protein

Somatic gene therapy has been proposed as a means to achieve systemic delivery of therapeutic proteins. However, there is limited evidence that current methods of gene delivery can practically achieve this goal. In this study, we demonstrate that, following a single intramuscular administration of a recombinant adeno-associated virus (rAAV) vector containing the beta-galactosidase (AAV-lacZ) gene into adult BALB/c mice, protein expression was detected in myofibers for at least 32 weeks. A single intramuscular administration of an AAV vector containing a gene for human erythropoietin (AAV-Epo) into mice resulted in dose-dependent secretion of erythropoietin and corresponding increases in red blood cell production that persisted for up to 40 weeks. Primary human myotubes transduced in vitro with the AAV-Epo vector also showed dose-dependent production of Epo. These results demonstrate that rAAV vectors are able to transduce skeletal muscle and are capable of achieving sustained expression and systemic delivery of a therapeutic protein following a single intramuscular administration. Gene therapy using AAV vectors may provide a practical strategy for the treatment of inherited and acquired protein deficiencies.

Enhanced synthesis of tumor necrosis factor-inducible proteins, plasminogen activator inhibitor-2, manganese superoxide dismutase, and protein 28/5.6, is selectively triggered by the 55-kDa tumor necrosis factor receptor in human melanoma cells

We have demonstrated that A375 melanoma cells express mRNA for both types of tumor necrosis factor (TNF) receptors and receptor proteins on their plasma membranes. Specific agonist and blocking antibodies to either 55-kDa (TNF-R1) or 75-kDa (TNF-R2) TNF receptors combined with two-dimensional gel analysis were employed to determine which receptor type is responsible for mediating the induction of individual melanoma proteins. Our results indicate that the enhanced synthesis of proteins 21/>7 (M(r)/pI), 28/5.6, and 41/5.7 is selectively induced through TNF-R1. TNF induces these proteins; antagonist antibody to TNF-R1 prevents their induction by TNF, and TNF-R1 agonist induces them in the absence of TNF. Identification of these proteins by immunoblot analysis proved that 21/>7 is manganese superoxide dismutase, protein 28/5.6 is unrelated to 27/28-kDa heat shock protein, and protein 41/5.7 is plasminogen activator inhibitor-2. Furthermore, TNF cytotoxicity for A375 cells is also mediated by TNF-R1. These studies indicate that TNF-R1 is a critical signaling receptor for TNF action on A375 cells and demonstrate the potential use of TNF-R1 antibodies to selectively block or enhance specific effects of TNF on melanoma cells.