Three-Dimensional Self-Navigated T2 Mapping for the Detection of Acute Cellular Rejection After Orthotopic Heart Transplantation

Prediction of heart transplant rejection from routine pathology slides with self-supervised deep learning

Aims

One of the most important complications of heart transplantation is organ rejection, which is diagnosed on endomyocardial biopsies by pathologists. Computer-based systems could assist in the diagnostic process and potentially improve reproducibility. Here, we evaluated the feasibility of using deep learning in predicting the degree of cellular rejection from pathology slides as defined by the International Society for Heart and Lung Transplantation (ISHLT) grading system.

Methods and results

We collected 1079 histopathology slides from 325 patients from three transplant centres in Germany. We trained an attention-based deep neural network to predict rejection in the primary cohort and evaluated its performance using cross-validation and by deploying it to three cohorts. For binary prediction (rejection yes/no), the mean area under the receiver operating curve (AUROC) was 0.849 in the cross-validated experiment and 0.734, 0.729, and 0.716 in external validation cohorts. For a prediction of the ISHLT grade (0R, 1R, 2/3R), AUROCs were 0.835, 0.633, and 0.905 in the cross-validated experiment and 0.764, 0.597, and 0.913; 0.631, 0.633, and 0.682; and 0.722, 0.601, and 0.805 in the validation cohorts, respectively. The predictions of the artificial intelligence model were interpretable by human experts and highlighted plausible morphological patterns.

Conclusion

We conclude that artificial intelligence can detect patterns of cellular transplant rejection in routine pathology, even when trained on small cohorts.

The Road Toward Reproducibility of Parametric Mapping of the Heart: A Technical Review

Parametric mapping of the heart has become an essential part of many cardiovascular magnetic resonance imaging exams, and is used for tissue characterization and diagnosis in a broad range of cardiovascular diseases. These pulse sequences are used to quantify the myocardial T₁, T₂, T2*, and T_1ρ relaxation times, which are unique surrogate indices of fibrosis, edema and iron deposition that can be used to monitor a disease over time or to compare patients to one another. Parametric mapping is now well-accepted in the clinical setting, but its wider dissemination is hindered by limited inter-center reproducibility and relatively long acquisition times. Recently, several new parametric mapping techniques have appeared that address both of these problems, but substantial hurdles remain for widespread clinical adoption. This review serves both as a primer for newcomers to the field of parametric mapping and as a technical update for those already well at home in it. It aims to establish what is currently needed to improve the reproducibility of parametric mapping of the heart. To this end, we first give an overview of the metrics by which a mapping technique can be assessed, such as bias and variability, as well as the basic physics behind the relaxation times themselves and what their relevance is in the prospect of myocardial tissue characterization. This is followed by a summary of routine mapping techniques and their variations. The problems in reproducibility and the sources of bias and variability of these techniques are reviewed. Subsequently, novel fast, whole-heart, and multi-parametric techniques and their merits are treated in the light of their reproducibility. This includes state of the art segmentation techniques applied to parametric maps, and how artificial intelligence is being harnessed to solve this long-standing conundrum. We finish up by sketching an outlook on the road toward inter-center reproducibility, and what to expect in the future.

Respiratory Motion-Registered Isotropic Whole-Heart T2 Mapping in Patients With Acute Non-ischemic Myocardial Injury

Background: T₂ mapping is a magnetic resonance imaging technique that can be used to detect myocardial edema and inflammation. However, the focal nature of myocardial inflammation may render conventional 2D approaches suboptimal and make whole-heart isotropic 3D mapping desirable. While self-navigated 3D radial T₂ mapping has been demonstrated to work well at a magnetic field strength of 3T, it results in too noisy maps at 1.5T. We therefore implemented a novel respiratory motion-resolved compressed-sensing reconstruction in order to improve the 3D T₂ mapping precision and accuracy at 1.5T, and tested this in a heterogeneous patient cohort. Materials and Methods: Nine healthy volunteers and 25 consecutive patients with suspected acute non-ischemic myocardial injury (sarcoidosis, n = 19; systemic sclerosis, n = 2; acute graft rejection, n = 2, and myocarditis, n = 2) were included. The free-breathing T₂ maps were acquired as three ECG-triggered T₂-prepared 3D radial volumes. A respiratory motion-resolved reconstruction was followed by image registration of the respiratory states and pixel-wise T₂ mapping. The resulting 3D maps were compared to routine 2D T₂ maps. The T₂ values of segments with and without late gadolinium enhancement (LGE) were compared in patients. Results: In the healthy volunteers, the myocardial T₂ values obtained with the 2D and 3D techniques were similar (45.8 ± 1.8 vs. 46.8 ± 2.9 ms, respectively; P = 0.33). Conversely, in patients, T₂ values did differ between 2D (46.7 ± 3.6 ms) and 3D techniques (50.1 ± 4.2 ms, P = 0.004). Moreover, with the 2D technique, T₂ values of the LGE-positive segments were similar to those of the LGE-negative segments (T_2LGE-= 46.2 ± 3.7 vs. T_2LGE+ = 47.6 ± 4.1 ms; P = 0.49), whereas the 3D technique did show a significant difference (T_2LGE- = 49.3 ± 6.7 vs. T_2LGE+ = 52.6 ± 8.7 ms, P = 0.006). Conclusion: Respiratory motion-registered 3D radial imaging at 1.5T led to accurate isotropic 3D whole-heart T₂ maps, both in the healthy volunteers and in a small patient cohort with suspected non-ischemic myocardial injury. Significantly higher T₂ values were found in patients as compared to controls in 3D but not in 2D, suggestive of the technique's potential to increase the sensitivity of CMR at earlier stages of disease. Further study will be needed to demonstrate its accuracy.

High-Spatial-Resolution 3D Whole-Heart MRI T2 Mapping for Assessment of Myocarditis

Background Clinical guidelines recommend the use of established T2 mapping sequences to detect and quantify myocarditis and edema, but T2 mapping is performed in two dimensions with limited coverage and repetitive breath holds. Purpose To assess the reproducibility of an accelerated free-breathing three-dimensional (3D) whole-heart T2 MRI mapping sequence in phantoms and participants without a history of cardiac disease and to investigate its clinical performance in participants with suspected myocarditis. Materials and Methods Eight participants (three women, mean age, 31 years ± 4 [standard deviation]; cohort 1) without a history of cardiac disease and 25 participants (nine women, mean age, 45 years ± 17; cohort 2) with clinically suspected myocarditis underwent accelerated free-breathing 3D whole-heart T2 mapping with 100% respiratory scanning efficiency at 1.5 T. The participants were enrolled from November 2018 to August 2020. Three repeated scans were performed on 2 separate days in cohort 1. Segmental variations in T2 relaxation times of the left ventricular myocardium were assessed, and intrasession and intersession reproducibility were measured. In cohort 2, segmental myocardial T2 values, detection of focal inflammation, and map quality were compared with those obtained from clinical breath-hold two-dimensional (2D) T2 mapping. Statistical differences were assessed using the nonparametric Mann-Whitney and Kruskal-Wallis tests, whereas the paired Wilcoxon signed-rank test was used to assess subjective scores. Results Whole-heart T2 maps were acquired in a mean time of 6 minutes 53 seconds ± 1 minute 5 seconds at 1.5 mm3 resolution. Breath-hold 2D and free-breathing 3D T2 mapping had similar intrasession (mean T2 change of 3.2% and 2.3% for 2D and 3D, respectively) and intersession (4.8% and 4.9%, respectively) reproducibility. The two T2 mapping sequences showed similar map quality (P = .23, cohort 2). Abnormal myocardial segments were identified with confidence (score 3) in 14 of 25 participants (56%) with 3D T2 mapping and only in 10 of 25 participants (40%) with 2D T2 mapping. Conclusion High-spatial-resolution three-dimensional (3D) whole-heart T2 mapping shows high intrasession and intersession reproducibility and helps provide T2 myocardial characterization in agreement with clinical two-dimensional reference, while enabling 3D assessment of focal disease with higher confidence. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Friedrich in this issue.

3D whole-heart isotropic-resolution motion-compensated joint T1 /T2 mapping and water/fat imaging

PURPOSE

To develop a free-breathing isotropic-resolution whole-heart joint T₁ and T₂ mapping sequence with Dixon-encoding that provides coregistered 3D T₁ and T₂ maps and complementary 3D anatomical water and fat images in a single ~9 min scan.

METHODS

Four interleaved dual-echo Dixon gradient echo volumes are acquired with a variable density Cartesian trajectory and different preparation pulses: 1) inversion recovery-preparation, 2) and 3) no preparations, and 4) T₂ preparation. Image navigators are acquired to correct each echo for 2D translational respiratory motion; the 8 echoes are jointly reconstructed with a low-rank patch-based reconstruction. A water/fat separation algorithm is used to obtain water and fat images for each acquired volume. T₁ and T₂ maps are generated by matching the signal evolution of the water images to a simulated dictionary. Complementary bright-blood and fat volumes for anatomical visualization are obtained from the T₂ -prepared dataset. The proposed sequence was tested in phantom experiments and 10 healthy subjects and compared to standard 2D MOLLI T₁ mapping, 2D balance steady-state free precession T₂ mapping, and 3D T₂ -prepared Dixon coronary MR angiography.

RESULTS

High linear correlation was found between T₁ and T₂ quantification with the proposed approach and phantom spin echo measurements (y = 1.1 × -11.68, R² = 0.98; and y = 0.85 × +5.7, R² = 0.99). Mean myocardial values of T₁ /T₂ = 1116 ± 30.5 ms/45.1 ± 2.38 ms were measured in vivo. Biases of T₁ /T₂ = 101.8 ms/-0.77 ms were obtained compared to standard 2D techniques.

CONCLUSION

The proposed joint T₁ /T₂ sequence permitted the acquisition of motion-compensated isotropic-resolution 3D T₁ and T₂ maps and complementary coronary MR angiography and fat volumes, showing promising results in terms of T₁ and T₂ quantification and visualization of cardiac anatomy and pericardial fat.

Accelerated free-breathing whole-heart 3D T2 mapping with high isotropic resolution

PURPOSE

To enable free-breathing whole-heart 3D T2 mapping with high isotropic resolution in a clinically feasible and predictable scan time. This 3D motion-corrected undersampled signal matched (MUST) T2 map is achieved by combining an undersampled motion-compensated T2 -prepared Cartesian acquisition with a high-order patch-based reconstruction.

METHODS

The 3D MUST-T2 mapping acquisition consists of an electrocardiogram-triggered, T2 -prepared, balanced SSFP sequence with nonselective saturation pulses. Three undersampled T2 -weighted volumes are acquired using a 3D Cartesian variable-density sampling with increasing T2 preparation times. A 2D image-based navigator is used to correct for respiratory motion of the heart and allow 100% scan efficiency. Multicontrast high-dimensionality undersampled patch-based reconstruction is used in concert with dictionary matching to generate 3D T2 maps. The proposed framework was evaluated in simulations, phantom experiments, and in vivo (10 healthy subjects, 2 patients) with 1.5-mm3 isotropic resolution. Three-dimensional MUST-T2 was compared against standard multi-echo spin-echo sequence (phantom) and conventional breath-held single-shot 2D SSFP T2 mapping (in vivo).

RESULTS

Three-dimensional MUST-T2 showed high accuracy in phantom experiments (R2 > 0.99). The precision of T2 values was similar for 3D MUST-T2 and 2D balanced SSFP T2 mapping in vivo (5 ± 1 ms versus 4 ± 2 ms, P = .52). Slightly longer T2 values were observed with 3D MUST-T2 in comparison to 2D balanced SSFP T2 mapping (50.7 ± 2 ms versus 48.2 ± 1 ms, P < .05). Preliminary results in patients demonstrated T2 values in agreement with literature values.

CONCLUSION

The proposed approach enables free-breathing whole-heart 3D T2 mapping with high isotropic resolution in about 8 minutes, achieving accurate and precise T2 quantification of myocardial tissue in a clinically feasible scan time.

3D Whole-heart free-breathing qBOOST-T2 mapping

PURPOSE

To develop an accelerated motion corrected 3D whole-heart imaging approach (qBOOST-T2) for simultaneous high-resolution bright- and black-blood cardiac MR imaging and quantitative myocardial T2 characterization.

METHODS

Three undersampled interleaved balanced steady-state free precession cardiac MR volumes were acquired with a variable density Cartesian trajectory and different magnetization preparations: (1) T2-prepared inversion recovery (T2prep-IR), (2) T2-preparation, and (3) no preparation. Image navigators were acquired prior the acquisition to correct for 2D translational respiratory motion. Each 3D volume was reconstructed with a low-rank patch-based reconstruction. The T2prep-IR volume provides bright-blood anatomy visualization, the black-blood volume is obtained by means of phase sensitive reconstruction between first and third datasets, and T2 maps are generated by matching the signal evolution to a simulated dictionary. The proposed sequence has been evaluated in simulations, phantom experiments, 11 healthy subjects and compared with 3D bright-blood cardiac MR and standard 2D breath-hold balanced steady-state free precession T2 mapping. The feasibility of the proposed approach was tested on 4 patients with suspected cardiovascular disease.

RESULTS

High linear correlation (y = 1.09 × -0.83, R² = 0.99) was found between the proposed qBOOST-T2 and T2 spin echo measurements in phantom experiment. Good image quality was observed in vivo with the proposed 4x undersampled qBOOST-T2. Mean T2 values of 53.1 ± 2.1 ms and 55.8 ± 2.7 ms were measured in vivo for 2D balanced steady-state free precession T2 mapping and qBOOST-T2, respectively, with linear correlation of y = 1.02x+1.46 (R² = 0.61) and T2 bias = 2.7 ms.

CONCLUSION

The proposed qBOOST-T2 sequence allows the acquisition of 3D high-resolution co-registered bright- and black-blood volumes and T2 maps in a single scan of ~11 min, showing promising results in terms of T2 quantification.

Quantifying Intermembrane Distances with Serial Image Dilations

A recently-described extracellular nanodomain, termed the perinexus, has been implicated in ephaptic coupling, which is an alternative mechanism for electrical conduction between cardiomyocytes. The current method for quantifying this space by manual segmentation is slow and has low spatial resolution.We developed an algorithm that uses serial image dilations of a binary outline to count the number of pixels between two opposing 2 dimensional edges.This algorithm requires fewer man hours and has a higher spatial resolution than the manual method while preserving the reproducibility of the manual process.In fact, experienced and novice investigators were able to recapitulate the results of a previous study with this new algorithm.The algorithm is limited by the human input needed to manually outline the perinexus and computational power mainly encumbered by a pre-existing pathfinding algorithm.However, the algorithm's high-throughput capabilities, high spatial resolution and reproducibility make it a versatile and robust measurement tool for use across a variety of applications requiring the measurement of the distance between any 2-dimensional (2D) edges.