TF-Unet:An automatic cardiac MRI image segmentation method.

Personalized heart models are widely used to study the mechanisms of cardiac arrhythmias and have been used to guide clinical ablation of different types of arrhythmias in recent years. MRI images are now mostly used for model building. In cardiac modeling studies, the degree of segmentation of the heart image determines the success of subsequent 3D reconstructions. Therefore, a fully automated segmentation is needed. In this paper, we combine U-Net and Transformer as an alternative approach to perform powerful and fully automated segmentation of medical images. On the one hand, we use convolutional neural networks for feature extraction and spatial encoding of inputs to fully exploit the advantages of convolution in detail grasping; on the other hand, we use Transformer to add remote dependencies to high-level features and model features at different scales to fully exploit the advantages of Transformer. The results show that, the average dice coefficients for ACDC and Synapse datasets are 91.72 and 85.46%, respectively, and compared with Swin-Unet, the segmentation accuracy are improved by 1.72% for ACDC dataset and 6.33% for Synapse dataset.

Current progress of computational modeling for guiding clinical atrial fibrillation ablation.

Atrial fibrillation (AF) is one of the most common arrhythmias, associated with high morbidity, mortality, and healthcare costs, and it places a significant burden on both individuals and society. Anti-arrhythmic drugs are the most commonly used strategy for treating AF. However, drug therapy faces challenges because of its limited efficacy and potential side effects. Catheter ablation is widely used as an alternative treatment for AF. Nevertheless, because the mechanism of AF is not fully understood, the recurrence rate after ablation remains high. In addition, the outcomes of ablation can vary significantly between medical institutions and patients, especially for persistent AF. Therefore, the issue of which ablation strategy is optimal is still far from settled. Computational modeling has the advantages of repeatable operation, low cost, freedom from risk, and complete control, and is a useful tool for not only predicting the results of different ablation strategies on the same model but also finding optimal personalized ablation targets for clinical reference and even guidance. This review summarizes three-dimensional computational modeling simulations of catheter ablation for AF, from the early-stage attempts such as Maze III or circumferential pulmonary vein isolation to the latest advances based on personalized substrate-guided ablation. Finally, we summarize current developments and challenges and provide our perspectives and suggestions for future directions.

Fully Automatic Scar Segmentation for Late Gadolinium Enhancement MRI Images in Left Ventricle with Myocardial Infarction.

Numerous methods have been published to segment the infarct tissue in the left ventricle, most of them either need manual work, post-processing, or suffer from poor reproducibility. We proposed an automatic segmentation method for segmenting the infarct tissue in left ventricle with myocardial infarction. Cardiac images of a total of 60 diseased hearts (55 human hearts and 5 porcine hearts) were used in this study. The epicardial and endocardial boundaries of the ventricles in every 2D slice of the cardiac magnetic resonance with late gadolinium enhancement images were manually segmented. The subsequent pipeline of infarct tissue segmentation is fully automatic. The segmentation results with the automatic algorithm proposed in this paper were compared to the consensus ground truth. The median of Dice overlap between our automatic method and the consensus ground truth is 0.79. We also compared the automatic method with the consensus ground truth using different image sources from different centers with different scan parameters and different scan machines. The results showed that the Dice overlap with the public dataset was 0.83, and the overall Dice overlap was 0.79. The results show that our method is robust with respect to different MRI image sources, which were scanned by different centers with different image collection parameters. The segmentation accuracy we obtained is comparable to or better than that of the conventional semi-automatic methods. Our segmentation method may be useful for processing large amount of dataset in clinic.

Preliminary Study: Learning the Impact of Simulation Time on Reentry Location and Morphology Induced by Personalized Cardiac Modeling.

Personalized cardiac modeling is widely used for studying the mechanisms of cardiac arrythmias. Due to the high demanding of computational resource of modeling, the arrhythmias induced in the models are usually simulated for just a few seconds. In clinic, it is common that arrhythmias last for more than several minutes and the morphologies of reentries are not always stable, so it is not clear that whether the simulation of arrythmias for just a few seconds is long enough to match the arrhythmias detected in patients. This study aimed to observe how long simulation of the induced arrhythmias in the personalized cardiac models is sufficient to match the arrhythmias detected in patients. A total of 5 contrast enhanced MRI datasets of patient hearts with myocardial infarction were used in this study. Then, a classification method based on Gaussian mixture model was used to detect the infarct tissue. For each reentry, 3 s and 10 s were simulated. The characteristics of each reentry simulated for different duration were studied. Reentries were induced in all 5 ventricular models and sustained reentries were induced at 39 stimulation sites in the model. By analyzing the simulation results, we found that 41% of the sustained reentries in the 3 s simulation group terminated in the longer simulation groups (10 s). The second finding in our simulation was that only 23.1% of the sustained reentries in the 3 s simulation did not change location and morphology in the extended 10 s simulation. The third finding was that 35.9% reentries were stable in the 3 s simulation and should be extended for the simulation time. The fourth finding was that the simulation results in 10 s simulation matched better with the clinical measurements than the 3 s simulation. It was shown that 10 s simulation was sufficient to make simulation results stable. The findings of this study not only improve the simulation accuracy, but also reduce the unnecessary simulation time to achieve the optimal use of computer resources to improve the simulation efficiency and shorten the simulation time to meet the time node requirements of clinical operation on patients.

Computationally guided personalized targeted ablation of persistent atrial fibrillation.

Atrial fibrillation (AF)-the most common arrhythmia-significantly increases the risk of stroke and heart failure. Although catheter ablation can restore normal heart rhythms, patients with persistent AF who develop atrial fibrosis often undergo multiple failed ablations, and thus increased procedural risks. Here, we present personalized computational modelling for the reliable predetermination of ablation targets, which are then used to guide the ablation procedure in patients with persistent AF and atrial fibrosis. First, we show that a computational model of the atria of patients identifies fibrotic tissue that, if ablated, will not sustain AF. Then, we report the results of integrating the target ablation sites in a clinical mapping system and testing its feasibility in ten patients with persistent AF. The computational prediction of ablation targets avoids lengthy electrical mapping and could improve the accuracy and efficacy of targeted AF ablation in patients while eliminating the need for repeat procedures.

Characterizing Conduction Channels in Postinfarction Patients Using a Personalized Virtual Heart.

Patients with myocardial infarction have an abundance of conduction channels (CC); however, only a small subset of these CCs sustain ventricular tachycardia (VT). Identifying these critical CCs (CCCs) in the clinic so that they can be targeted by ablation remains a significant challenge. The objective of this study is to use a personalized virtual-heart approach to conduct a three-dimensional (3D) assessment of CCCs sustaining VTs of different morphologies in these patients, to investigate their 3D structural features, and to determine the optimal ablation strategy for each VT. To achieve these goals, ventricular models were constructed from contrast enhanced magnetic resonance imagings of six postinfarction patients. Rapid pacing induced VTs in each model. CCCs that sustained different VT morphologies were identified. CCCs' 3D structure and type and the resulting rotational electrical activity were examined. Ablation was performed at the optimal part of each CCC, aiming to terminate each VT with a minimal lesion size. Predicted ablation locations were compared to clinical. Analyzing the simulation results, we found that the observed VTs in each patient model were sustained by a limited number (2.7 ± 1.2) of CCCs. Further, we identified three types of CCCs sustaining VTs: I-type and T-type channels, with all channel branches bounded by scar, and functional reentry channels, which were fully or partially bounded by conduction block surfaces. The different types of CCCs accounted for 43.8, 18.8, and 37.4% of all CCCs, respectively. The mean narrowest width of CCCs or a branch of CCC was 9.7 ± 3.6 mm. Ablation of the narrowest part of each CCC was sufficient to terminate VT. Our results demonstrate that a personalized virtual-heart approach can determine the possible VT morphologies in each patient and identify the CCCs that sustain reentry. The approach can aid clinicians in identifying accurately the optimal VT ablation targets in postinfarction patients.

Sensitivity of Ablation Targets Prediction to Electrophysiological Parameter Variability in Image-Based Computational Models of Ventricular Tachycardia in Post-infarction Patients.

Ventricular tachycardia (VT), which could lead to sudden cardiac death, occurs frequently in patients with myocardial infarction. Computational modeling has emerged as a powerful platform for the non-invasive investigation of lethal heart rhythm disorders in post-infarction patients and for guiding patient VT ablation. However, it remains unclear how VT dynamics and predicted ablation targets are influenced by inter-patient variability in action potential duration (APD) and conduction velocity (CV). The goal of this study was to systematically assess the effect of changes in the electrophysiological parameters on the induced VTs and predicted ablation targets in personalized models of post-infarction hearts. Simulations were conducted in 5 patient-specific left ventricular models reconstructed from late gadolinium-enhanced magnetic resonance imaging scans. We comprehensively characterized all possible pre-ablation and post-ablation VTs in simulations conducted with either an "average human VT"-based electrophysiological representation (i.e., EPavg) or with ±10% APD or CV (i.e., EPvar); additional simulations were also executed in some models for an extended range of these parameters. The results showed that: (1) a subset of reentries (76.2-100%, depending on EP parameter set) conducted with ±10% APD/CV was observed in approximately the same locations as reentries observed in EPavg cases; (2) emergent VTs could be induced sometimes after ablation in EPavg models, and these emergent VTs often corresponded to the pre-ablation reentries in simulations with EPvar parameter sets. These findings demonstrate that the VT ablation target uncertainty in patient-specific ventricular models with an average representation of VT-remodeled electrophysiology is relatively low and the ablation targets stable, as the localization of the induced VTs was primarily driven by the remodeled structural substrate. Thus, personalized ventricular modeling with an average representation of infarct-remodeled electrophysiology may uncover most targets for VT ablation.

Antibiotics in a typical karst river system in China: Spatiotemporal variation and environmental risks.

Karst aquifers are highly susceptible to contamination because compounds in water from the land surface are able to enter aquifers directly through sinkholes and travel rapidly through conduits. To investigate the occurrence and profiles of antibiotics in the typical karst river system in Kaiyang, southwest China, 34 aqueous samples were collected periodically to delineate seasonal trends in antibiotic levels. Thirty-five antibiotics, including nine sulfonamides, four tetracyclines, five macrolides, 16 quinolones and chloramphenicol, were analysed via solid phase extraction combined with ultra-performance liquid chromatography-tandem mass spectrometry. A total of 25 antibiotics were detected with the highest detection frequency reaching 94.1%, indicating the ubiquity of antibiotics in the study area. The total concentration of antibiotics ranged from 0.37 to 508.6 ng/L, with the dominating proportion including macrolides and quinolones based on the distribution profiles and seasonal variation. Due to the natural attenuation, the total concentration of antibiotics gradually decreased with the flow direction in the southern part of the river. The total concentrations of antibiotics in the mainstream were significantly higher in the dry season than in the rainy seasons. However, the distribution profiles were susceptible to anthropogenic activities, such as the leakage of septic tank wastewater. The dendrogram and heatmap revealed that three clusters of sample sites represented tributaries and the upstream areas, the downstream areas, and the potential pollutant source, and three clusters of antibiotics represented different concentration patterns. The high ecological risks of tetracycline, erythromycin and ciprofloxacin for algae and ofloxacin for plants were determined. These findings contributed to the establishment of a database for future monitoring and control of antibiotics in karst areas.

Optimal contrast-enhanced MRI image thresholding for accurate prediction of ventricular tachycardia using ex-vivo high resolution models.

Patient specific models created from contrast-enhanced (i.e. late-gadolinium, LGE) MRI images can be used for prediction of reentry location and clinical ablation planning. However, there is still a need for direct and systematic comparison between characteristics of ventricular tachycardia (VT) morphologies predicted in computational models and those acquired in clinical or experimental protocols. In this study, we aimed to: 1) assess the differences in VT morphologies predicted by modeling and recorded in experiments in terms of patterns and location of reentries, earliest and latest activation sites, and cycle lengths; and 2) define the optimal range of infarct tissue threshold values which provide best match between simulation and experimental results. To achieve these goals, we utilized LGE-MRI images from 4 swine hearts with inducible monomorphic VT. The images were segmented to identify non-infarcted myocardium, semi viable gray zone (GZ), and core scar based on pixel intensity. Several models were reconstructed from each LGE-MRI scan, with voxels of intensity between that of non-infarcted myocardium and 20-50% of the maximum intensity (in 10% increments) in the infarct region classified as GZ. VT induction was simulated in each model. Our simulation results showed that using GZ intensity thresholds of 20% or 30% resulted in the best match of simulated propagation patterns and reentry locations with those from the experiment. Overall, we matched 70% (7/10) morphologies for all the hearts. Our simulation shows that MRI-based computational models of hearts with myocardial infarction can accurately reproduce the majority of experimentally recorded post-infarction VTs.

Personalized virtual-heart technology for guiding the ablation of infarct-related ventricular tachycardia.

Ventricular tachycardia (VT), which can lead to sudden cardiac death, occurs frequently in patients with myocardial infarction. Catheter-based radiofrequency ablation of cardiac tissue has achieved only modest efficacy, owing to the inaccurate identification of ablation targets by current electrical mapping techniques, which can lead to extensive lesions and to a prolonged, poorly tolerated procedure. Here we show that personalized virtual-heart technology based on cardiac imaging and computational modelling can identify optimal infarct-related VT ablation targets in retrospective animal (5 swine) and human studies (21 patients) and in a prospective feasibility study (5 patients). We first assessed in retrospective studies (one of which included a proportion of clinical images with artifacts) the capability of the technology to determine the minimum-size ablation targets for eradicating all VTs. In the prospective study, VT sites predicted by the technology were targeted directly, without relying on prior electrical mapping. The approach could improve infarct-related VT ablation guidance, where accurate identification of patient-specific optimal targets could be achieved on a personalized virtual heart prior to the clinical procedure.

Sensitivity of reentrant driver localization to electrophysiological parameter variability in image-based computational models of persistent atrial fibrillation sustained by a fibrotic substrate.

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, causing morbidity and mortality in millions worldwide. The atria of patients with persistent AF (PsAF) are characterized by the presence of extensive and distributed atrial fibrosis, which facilitates the formation of persistent reentrant drivers (RDs, i.e., spiral waves), which promote fibrillatory activity. Targeted catheter ablation of RD-harboring tissues has shown promise as a clinical treatment for PsAF, but the outcomes remain sub-par. Personalized computational modeling has been proposed as a means of non-invasively predicting optimal ablation targets in individual PsAF patients, but it remains unclear how RD localization dynamics are influenced by inter-patient variability in the spatial distribution of atrial fibrosis, action potential duration (APD), and conduction velocity (CV). Here, we conduct simulations in computational models of fibrotic atria derived from the clinical imaging of PsAF patients to characterize the sensitivity of RD locations to these three factors. We show that RDs consistently anchor to boundaries between fibrotic and non-fibrotic tissues, as delineated by late gadolinium-enhanced magnetic resonance imaging, but those changes in APD/CV can enhance or attenuate the likelihood that an RD will anchor to a specific site. These findings show that the level of uncertainty present in patient-specific atrial models reconstructed without any invasive measurements (i.e., incorporating each individual's unique distribution of fibrotic tissue from medical imaging alongside an average representation of AF-remodeled electrophysiology) is sufficiently high that a personalized ablation strategy based on targeting simulation-predicted RD trajectories alone may not produce the desired result.

A feasibility study of arrhythmia risk prediction in patients with myocardial infarction and preserved ejection fraction.

AIM: To predict arrhythmia susceptibility in myocardial infarction (MI) patients with left ventricular ejection fraction (LVEF) >35% using a personalized virtual heart simulation approach. METHODS AND RESULTS: A total of four contrast enhanced magnetic resonance imaging (MRI) datasets of patient hearts with MI and average LVEF of 44.0 ± 2.6% were used in this study. Because of the preserved LVEF, the patients were not indicated for implantable cardioverter defibrillator (ICD) insertion. One patient had spontaneous ventricular tachycardia (VT) prior to the MRI scan; the others had no arrhythmic events. Simulations of arrhythmia susceptibility were blind to clinical outcome. Models were constructed from patient MRI images segmented to identify myocardium, grey zone, and scar based on pixel intensity. Grey zone was modelled as having altered electrophysiology. Programmed electrical stimulation (PES) was performed to assess VT inducibility from 19 bi-ventricular sites in each heart model. Simulations successfully predicted arrhythmia risk in all four patients. For the patient with arrhythmic event, in-silico PES resulted in VT induction. Simulations correctly predicted that VT was non-inducible for the three patients with no recorded VT events. CONCLUSIONS: Results demonstrate that the personalized virtual heart simulation approach may provide a novel risk stratification modality to non-invasively and effectively identify patients with LVEF >35% who could benefit from ICD implantation.

Accuracy of prediction of infarct-related arrhythmic circuits from image-based models reconstructed from low and high resolution MRI.

Identification of optimal ablation sites in hearts with infarct-related ventricular tachycardia (VT) remains difficult to achieve with the current catheter-based mapping techniques. Limitations arise from the ambiguities in determining the reentrant pathways location(s). The goal of this study was to develop experimentally validated, individualized computer models of infarcted swine hearts, reconstructed from high-resolution ex-vivo MRI and to examine the accuracy of the reentrant circuit location prediction when models of the same hearts are instead reconstructed from low clinical-resolution MRI scans. To achieve this goal, we utilized retrospective data obtained from four pigs ~10 weeks post infarction that underwent VT induction via programmed stimulation and epicardial activation mapping via a multielectrode epicardial sock. After the experiment, high-resolution ex-vivo MRI with late gadolinium enhancement was acquired. The Hi-res images were downsampled into two lower resolutions (Med-res and Low-res) in order to replicate image quality obtainable in the clinic. The images were segmented and models were reconstructed from the three image stacks for each pig heart. VT induction similar to what was performed in the experiment was simulated. Results of the reconstructions showed that the geometry of the ventricles including the infarct could be accurately obtained from Med-res and Low-res images. Simulation results demonstrated that induced VTs in the Med-res and Low-res models were located close to those in Hi-res models. Importantly, all models, regardless of image resolution, accurately predicted the VT morphology and circuit location induced in the experiment. These results demonstrate that MRI-based computer models of hearts with ischemic cardiomyopathy could provide a unique opportunity to predict and analyze VT resulting for from specific infarct architecture, and thus may assist in clinical decisions to identify and ablate the reentrant circuit(s).

A study of mechanical optimization strategy for cardiac resynchronization therapy based on an electromechanical model.

An optimal electrode position and interventricular (VV) delay in cardiac resynchronization therapy (CRT) improves its success. However, the precise quantification of cardiac dyssynchrony and magnitude of resynchronization achieved by biventricular (BiV) pacing therapy with mechanical optimization strategies based on computational models remain scant. The maximum circumferential uniformity ratio estimate (CURE) was used here as mechanical optimization index, which was automatically computed for 6 different electrode positions based on a three-dimensional electromechanical canine model of heart failure (HF) caused by complete left bundle branch block (CLBBB). VV delay timing was adjusted accordingly. The heart excitation propagation was simulated with a monodomain model. The quantification of mechanical intra- and interventricular asynchrony was then investigated with eight-node isoparametric element method. The results showed that (i) the optimal pacing location from maximal CURE of 0.8516 was found at the left ventricle (LV) lateral wall near the equator site with a VV delay of 60 ms, in accordance with current clinical studies, (ii) compared with electrical optimization strategy of E(RMS), the LV synchronous contraction and the hemodynamics improved more with mechanical optimization strategy. Therefore, measures of mechanical dyssynchrony improve the sensitivity and specificity of predicting responders more. The model was subject to validation in future clinical studies.

Simulation of biatrial conduction via different pathways during sinus rhythm with a detailed human atrial model.

In order to better understand biatrial conduction, investigate various conduction pathways, and compare the differences between isotropic and anisotropic conductions in human atria, we present a simulation study of biatrial conduction with known/assumed conduction pathways using a recently developed human atrial model. In addition to known pathways: (1) Bachmann's bundle (BB), (2) limbus of fossa ovalis (LFO), and (3) coronary sinus (CS), we also hypothesize that there exist two fast conduction bundles that connect the crista terminalis (CT), LFO, and CS. Our simulation demonstrates that use of these fast conduction bundles results in a conduction pattern consistent with experimental data. The comparison of isotropic and anisotropoic conductions in the BB case showed that the atrial working muscles had small effect on conduction time and conduction speed, although the conductivities assigned in anisotropic conduction were two to four times higher than the isotropic conduction. In conclusion, we suggest that the hypothesized intercaval bundles play a significant role in the biatrial conduction and that myofiber orientation has larger effects on the conduction system than the atrial working muscles. This study presents readers with new insights into human atrial conduction.

An image-based model of the whole human heart with detailed anatomical structure and fiber orientation.

Many heart anatomy models have been developed to study the electrophysiological properties of the human heart. However, none of them includes the geometry of the whole human heart. In this study, an anatomically detailed mathematical model of the human heart was firstly reconstructed from the computed tomography images. In the reconstructed model, the atria consisted of atrial muscles, sinoatrial node, crista terminalis, pectinate muscles, Bachmann's bundle, intercaval bundles, and limbus of the fossa ovalis. The atrioventricular junction included the atrioventricular node and atrioventricular ring, and the ventricles had ventricular muscles, His bundle, bundle branches, and Purkinje network. The epicardial and endocardial myofiber orientations of the ventricles and one layer of atrial myofiber orientation were then measured. They were calculated using linear interpolation technique and minimum distance algorithm, respectively. To the best of our knowledge, this is the first anatomically-detailed human heart model with corresponding experimentally measured fibers orientation. In addition, the whole heart excitation propagation was simulated using a monodomain model. The simulated normal activation sequence agreed well with the published experimental findings.