Segmentation of left ventricle in late gadolinium enhanced MRI through 2D‐4D registration for infarct localization in 3D patient‐specific left ventricular model

Impact of late gadolinium enhancement image acquisition resolution on neural network based automatic scar segmentation

BACKGROUND

Automatic myocardial scar segmentation from late gadolinium enhancement (LGE) images using neural networks promises an alternative to time-consuming and observer-dependent semi-automatic approaches. However, alterations in data acquisition, reconstruction as well as post-processing may compromise network performance. The objective of the present work was to systematically assess network performance degradation due to a mismatch of point-spread function between training and testing data.

METHODS

Thirty-six high-resolution (0.7×0.7×2.0 mm3) LGE k-space datasets were acquired post-mortem in porcine models of myocardial infarction. The in-plane point-spread function and hence in-plane resolution Δx was retrospectively degraded using k-space lowpass filtering, while field-of-view and matrix size were kept constant. Manual segmentation of the left ventricle (LV) and healthy remote myocardium was performed to quantify location and area (% of myocardium) of scar by thresholding (≥ SD5 above remote). Three standard U-Nets were trained on training resolutions Δxtrain = 0.7, 1.2 and 1.7 mm to predict endo- and epicardial borders of LV myocardium and scar. The scar prediction of the three networks for varying test resolutions (Δxtest = 0.7 to 1.7 mm) was compared against the reference SD5 thresholding at 0.7 mm. Finally, a fourth network trained on a combination of resolutions (Δxtrain = 0.7 to 1.7 mm) was tested.

RESULTS

The prediction of relative scar areas showed the highest precision when the resolution of the test data was identical to or close to the resolution used during training. The median fractional scar errors and precisions (IQR) from networks trained and tested on the same resolution were 0.0 percentage points (p.p.) (1.24 - 1.45), and - 0.5 - 0.0 p.p. (2.00 - 3.25) for networks trained and tested on the most differing resolutions, respectively. Deploying the network trained on multiple resolutions resulted in reduced resolution dependency with median scar errors and IQRs of 0.0 p.p. (1.24 - 1.69) for all investigated test resolutions.

CONCLUSION

A mismatch of the imaging point-spread function between training and test data can lead to degradation of scar segmentation when using current U-Net architectures as demonstrated on LGE porcine myocardial infarction data. Training networks on multi-resolution data can alleviate the resolution dependency.

Assessing Interobserver Variability in the Delineation of Structures in Radiation Oncology: A Systematic Review

PURPOSE

The delineation of target volumes and organs at risk is the main source of uncertainty in radiation therapy. Numerous interobserver variability (IOV) studies have been conducted, often with unclear methodology and nonstandardized reporting. We aimed to identify the parameters chosen in conducting delineation IOV studies and assess their performances and limits.

METHODS AND MATERIALS

We conducted a systematic literature review to highlight major points of heterogeneity and missing data in IOV studies published between 2018 and 2021. For the main used metrics, we did in silico analyses to assess their limits in specific clinical situations.

RESULTS

All disease sites were represented in the 66 studies examined. Organs at risk were studied independently of tumor site in 29% of reviewed IOV studies. In 65% of studies, statistical analyses were performed. No gold standard (GS; ie, reference) was defined in 36% of studies. A single expert was considered as the GS in 21% of studies, without testing intraobserver variability. All studies reported both absolute and relative indices, including the Dice similarity coefficient (DSC) in 68% and the Hausdorff distance (HD) in 42%. Limitations were shown in silico for small structures when using the DSC and dependence on irregular shapes when using the HD. Variations in DSC values were large between studies, and their thresholds were inconsistent. Most studies (51%) included 1 to 10 cases. The median number of observers or experts was 7 (range, 2-35). The intraclass correlation coefficient was reported in only 9% of cases. Investigating the feasibility of studying IOV in delineation, a minimum of 8 observers with 3 cases, or 11 observers with 2 cases, was required to demonstrate moderate reproducibility.

CONCLUSIONS

Implementation of future IOV studies would benefit from a more standardized methodology: clear definitions of the gold standard and metrics and a justification of the tradeoffs made in the choice of the number of observers and number of delineated cases should be provided.

GCW-UNet segmentation of cardiac magnetic resonance images for evaluation of left atrial enlargement

BACKGROUND AND OBJECTIVE

Left atrial enlargement (LAE) is an anatomical variation of the left atrium and the result of the long-term increase of left atrial pressure. Most of the increase in stress or volume is due to potential cardiovascular disease. Studies have shown that LAE can independently predict the development of clinically significant cardiovascular disease and heart failure. If the left atrial volume is accurately measured, it will be an essential indicator of human health and an essential means for doctors to find patients' potential diseases. We can analyze the dynamic changes in the left atrial structure and analyze left atrial dilation. However, manual segmentation was inefficient and error-prone before the 3D reconstruction of the left atrium. In order to solve this problem, a convolution neural network (CNN) method based on cardiac magnetic resonance image (MRI) is proposed to automatically segment the left atrial region.

METHODOLOGY

In this paper, we have proposed and developed a novel U-Net with Gaussian blur and channel weight neural network (GCW-UNet) to automatically segment the left atrial region in the MRI of a patient with LAE. After Gaussian blur, different resolutions of the MRI are obtained. High-resolution MRI clearly shows the detailed features of the left atrium, while low-resolution MRI clearly shows the overall outline of the left atrium, which can solve the problem of more minor MRI features. Adaptive channel weights can enhance the atrial segmentation capability of the network.

RESULTS

Compared with the state-of-the-art left atrial segmentation methods, our CNN-based technique results in the segmentation of the left atrium being closer to the manual segmentation by an experienced radiologist. On the test datasets, the mean Dice similarity coefficient reaches 93.57%.

CONCLUSION

Firstly, MRI has a small number of imaging artifacts, which results in low segmentation accuracy. Our method successfully solves the problem. Secondly, due to the high similarity between the background (the area outside the left atrium) and the foreground (the left atrium) in MRI, traditional neural networks misclassify the background as the foreground. Our GCW-Unit can address the imbalanced number of pixels between the foreground and background. Finally, after segmenting the left atrium in the MRI by GCW-Unit, we reconstructed the left atrium to model a three-dimensional heart of a patient suffering from LAE. Based on the different time frames of one heartbeat, we could present the dynamics of the left atrial structure during a cardiac cycle. This can better assist in the evaluation of LAE in heart patients.

Recent Advances in Fibrosis and Scar Segmentation From Cardiac MRI: A State-of-the-Art Review and Future Perspectives

Segmentation of cardiac fibrosis and scars is essential for clinical diagnosis and can provide invaluable guidance for the treatment of cardiac diseases. Late Gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) has been successful in guiding the clinical diagnosis and treatment reliably. For LGE CMR, many methods have demonstrated success in accurately segmenting scarring regions. Co-registration with other non-contrast-agent (non-CA) modalities [e.g., balanced steady-state free precession (bSSFP) cine magnetic resonance imaging (MRI)] can further enhance the efficacy of automated segmentation of cardiac anatomies. Many conventional methods have been proposed to provide automated or semi-automated segmentation of scars. With the development of deep learning in recent years, we can also see more advanced methods that are more efficient in providing more accurate segmentations. This paper conducts a state-of-the-art review of conventional and current state-of-the-art approaches utilizing different modalities for accurate cardiac fibrosis and scar segmentation.

The role of regional myocardial topography post-myocardial infarction on infarct extension

Infarct extension involves necrosis of healthy myocardium in the border zone (BZ), progressively enlarging the infarct zone (IZ) and recruiting the remote zone (RZ) into the BZ, eventually leading to heart failure. The mechanisms underlying infarct extension remain unclear, but myocyte stretching has been suggested as the most likely cause. Using human patient-specific left-ventricular (LV) numerical simulations established from cardiac magnetic resonance imaging (MRI) of myocardial infarction (MI) patients, the correlation between infarct extension and regional mechanics abnormality was investigated by analysing the fibre stress-strain loops (FSSLs). FSSL abnormality was characterised using the directional regional external work (DREW) index, which measures FSSL area and loop direction. Sensitivity studies were also performed to investigate the effect of infarct stiffness on regional myocardial mechanics and potential for infarct extension. We found that infarct extension was correlated to severely abnormal FSSL in the form of counter-clockwise loop at the RZ close to the infarct, as indicated by negative DREW values. In regions demonstrating negative DREW values, we observed substantial fibre stretching in the isovolumic relaxation (IVR) phase accompanied by a reduced rate of systolic shortening. Such stretching in IVR phase in part of the RZ was due to its inability to withstand the high LV pressure that was still present and possibly caused by regional myocardial stiffness inhomogeneity. Further analysis revealed that the occurrence of severely abnormal FSSL due to IVR fibre stretching near the RZ-BZ boundary was due to a large amount of surrounding infarcted tissue, or an excessively stiff IZ.

Improved Quantification of Myocardium Scar in Late Gadolinium Enhancement Images: Deep Learning Based Image Fusion Approach

BACKGROUND

Quantification of myocardium scarring in late gadolinium enhanced (LGE) cardiac magnetic resonance imaging can be challenging due to low scar-to-background contrast and low image quality. To resolve ambiguous LGE regions, experienced readers often use conventional cine sequences to accurately identify the myocardium borders.

PURPOSE

To develop a deep learning model for combining LGE and cine images to improve the robustness and accuracy of LGE scar quantification.

STUDY TYPE

Retrospective.

POPULATION

A total of 191 hypertrophic cardiomyopathy patients: 1) 162 patients from two sites randomly split into training (50%; 81 patients), validation (25%, 40 patients), and testing (25%; 41 patients); and 2) an external testing dataset (29 patients) from a third site.

FIELD STRENGTH/SEQUENCE

1.5T, inversion-recovery segmented gradient-echo LGE and balanced steady-state free-precession cine sequences ASSESSMENT: Two convolutional neural networks (CNN) were trained for myocardium and scar segmentation, one with and one without LGE-Cine fusion. For CNN with fusion, the input was two aligned LGE and cine images at matched cardiac phase and anatomical location. For CNN without fusion, only LGE images were used as input. Manual segmentation of the datasets was used as reference standard.

STATISTICAL TESTS

Manual and CNN-based quantifications of LGE scar burden and of myocardial volume were assessed using Pearson linear correlation coefficients (r) and Bland-Altman analysis.

RESULTS

Both CNN models showed strong agreement with manual quantification of LGE scar burden and myocardium volume. CNN with LGE-Cine fusion was more robust than CNN without LGE-Cine fusion, allowing for successful segmentation of significantly more slices (603 [95%] vs. 562 (89%) of 635 slices; P < 0.001). Also, CNN with LGE-Cine fusion showed better agreement with manual quantification of LGE scar burden than CNN without LGE-Cine fusion (%ScarLGE-cine = 0.82 × %Scarmanual , r = 0.84 vs. %ScarLGE  = 0.47 × %Scarmanual , r = 0.81) and myocardium volume (VolumeLGE-cine = 1.03 × Volumemanual , r = 0.96 vs. VolumeLGE  = 0.91 × Volumemanual , r = 0.91).

DATA CONCLUSION

CNN based LGE-Cine fusion can improve the robustness and accuracy of automated scar quantification.

LEVEL OF EVIDENCE

3 TECHNICAL EFFICACY: 1.

Phenotyping of hypertensive heart disease and hypertrophic cardiomyopathy using personalized 3D modelling and cardiac cine MRI

Differential diagnosis of hypertensive heart disease (HHD) and hypertrophic cardiomyopathy (HCM) is clinically challenging but important for treatment management. This study aims to phenotype HHD and HCM in 3D + time domain by using a multiparametric motion-corrected personalized modeling algorithm and cardiac magnetic resonance (CMR). 44 CMR data, including 12 healthy, 16 HHD and 16 HCM cases, were examined. Multiple CMR phenotype data consisting of geometric and dynamic variables were extracted globally and regionally from the models over a full cardiac cycle for comparison against healthy models and clinical reports. Statistical classifications were used to identify the distinctive characteristics and disease subtypes with overlapping functional data, providing insights into the challenges for differential diagnosis of both types of disease. While HCM is characterized by localized extreme hypertrophy of the LV, wall thickening/contraction/strain was found to be normal and in sync, though it was occasionally exaggerated at normotrophic/less severely hypertrophic regions during systole to preserve the overall ejection fraction (EF) and systolic functionality. Additionally, we observed that hypertrophy in HHD could also be localized, although at less extreme conditions (i.e. more concentric). While fibrosis occurs mostly in those HCM cases with aortic obstruction, only minority of HHD patients were found affected by fibrosis. We demonstrate that subgroups of HHD (i.e. preserved and reduced EF: HHDpEF & HHDrEF) have different 3D + time CMR characteristics. While HHDpEF has cardiac functions in normal range, dilation and heart failure are indicated in HHDrEF as reflected by low LV wall thickening/contraction/strain and synchrony, as well as much reduced EF.