A Simple Method for Automatic 3D Reconstruction of Coronary Arteries From X-Ray Angiography.

Automatic three-dimensional (3-D) reconstruction of the coronary arteries (CA) from medical imaging modalities is still a challenging task. In this study, we present a deep learning-based method of automatic identification of the two ends of the vessel from X-ray coronary angiography (XCA). We also present a method of using template models of CA in matching the two-dimensional segmented vessels from two different angles of XCA. For the deep learning network, we used a U-net consisting of an encoder (Resnet) and a decoder. The two ends of the vessel were manually labeled to generate training images. The network was trained with 2,342, 1,907, and 1,523 labeled images for the left anterior descending (LAD), left circumflex (LCX), and right coronary artery (RCA), respectively. For template models of CA, ten reconstructed 3-D models were averaged for each artery. The accuracy of correspondence using template models was compared with that of manual matching. The deep learning network pointed the proximal region (20% of the total length) in 97.7, 97.5, and 96.4% of 315, 201, and 167 test images for LAD, LCX, and RCA, respectively. The success rates in pointing the distal region were 94.9, 89.8, and 94.6%, respectively. The average distances between the projected points from the reconstructed 3-D model to the detector and the points on the segmented vessels were not statistically different between the template and manual matchings. The computed FFR was not significantly different between the two matchings either. Deep learning methodology is feasible in identifying the two ends of the vessel in XCA, and the accuracy of using template models is comparable to that of manual correspondence in matching the segmented vessels from two angles.

Electrophysiological significance of the interatrial conduction including cavo-tricuspid isthmus during atrial fibrillation.

KEY POINTS: The interatrial conduction, including Bachmann's bundle, the posterior septal conduction, the anterior septal conduction, and the cavo-tricuspid isthmus, contributes to the maintenance mechanisms of atrial fibrillation in a 3D biatrial model. The interatrial conduction ablation including a cavo-tricuspid isthmus ablation significantly affects the wave dynamics of atrial fibrillation (AF) and facilitates the AF termination or atrial tachycardia conversion of the AF after the circumferential pulmonary vein isolation. Additional cavo-tricuspid isthmus ablation after the circumferential pulmonary vein isolation improves long-term rhythm outcome after clinical AF catheter ablation. ABSTRACT: Although it is known that atrial fibrillation (AF) is mainly a left atrial (LA) disease, the role of the right atrium (RA) and interatrial conduction (IAC), including the cavo-tricuspid isthmus (CTI), has not been clearly defined. We tested AF wave dynamics with or without IAC in computational modelling and the rhythm outcome of AF catheter ablation (AFCA) including CTI ablation in clinical cohort data. We evaluated the dominant frequency (DF) in 3D biatrial AF simulations integrated with 3D-computed tomograms obtained from 10 patients. The IAC was implemented at Bachmann's bundle, posterior septum and the CTI. After virtual circumferential PV isolation (CPVI), we disconnected IACs one by one, and observed the wave dynamics. We compared the long-term rhythm outcome after CPVI alone and additional CTI ablation in 846 patients with AFCA. LA-DF was higher than RA-DF in AF (P < 0.001). After CPVI, the DF decreased significantly by additional IAC ablation (P = 0.003), especially in the LA (P = 0.016). The amount of DF reduction (P = 0.020) and rates of AF termination (P < 0.001) or AT conversion (P = 0.021) were significantly higher after IAC ablations including CTI than those without. In clinical AFCA, the AF recurrence rate was significantly lower in patients with additional CTI ablation than CPVI alone during 25 ± 20 months' follow-up (hazard ratio 0.60 [0.46-0.79], P < 0.001, Log rank P < 0.001). IAC contributes to the maintenance mechanism of AF, and IAC including CTI ablation affects AF wave dynamics, facilitating AF termination in 3D biatrial modelling. Additional CTI ablation after CPVI improves the long-term rhythm outcome in clinical AFCA, potentially in a paroxysmal type with accompanying atrial flutter, or atrial dimension close to normal.

In silico evaluation of the acute occlusion effect of coronary artery on cardiac electrophysiology and the body surface potential map.

Body surface potential map, an electric potential distribution on the body torso surface, enables us to infer the electrical activities of the heart. Therefore, observing electric potential projected to the torso surface can be highly useful for diagnosing heart diseases such as coronary occlusion. The BSPM for the heart of a patient show a higher level of sensitivity than 12-lead ECG. Relevant research has been mostly based on clinical statistics obtained from patients, and, therefore, a simulation for a variety of pathological phenomena of the heart is required. In this study, by using computer simulation, a body surface potential map was implemented according to various occlusion locations (distal, mid, proximal occlusion) in the left anterior descending coronary artery. Electrophysiological characteristics of the body surface during the ST segment period were observed and analyzed based on an ST isointegral map. We developed an integrated system that takes into account the cellular to organ levels, and performed simulation regarding the electrophysiological phenomena of the heart that occur during the first 5 minutes (stage 1) and 10 minutes (stage 2) after commencement of coronary occlusion. Subsequently, we calculated the bipolar angle and amplitude of the ST isointegral map, and observed the correlation between the relevant characteristics and the location of coronary occlusion. In the result, in the ventricle model during the stage 1, a wider area of ischemia led to counterclockwise rotation of the bipolar angle; and, during the stage 2, the amplitude increased when the ischemia area exceeded a certain size.

Influence of LVAD function on mechanical unloading and electromechanical delay: a simulation study.

This study hypothesized that a left ventricular assist device (LVAD) shortens the electromechanical delay (EMD) by mechanical unloading. The goal of this study is to examine, by computational modeling, the influence of LVAD on EMD for four heart failure (HF) cases ranging from mild HF to severe HF. We constructed an integrated model of an LVAD-implanted cardiovascular system, then we altered the Ca2+ transient magnitude, with scaling factors 1, 0.9, 0.8, and 0.7 representing HF1, HF2, HF3, and HF4, respectively, in order of increasing HF severity. The four HF conditions are classified into two groups. Group one is the four HF conditions without LVAD, and group two is the conditions treated with continuous LVAD pump. The single-cell mechanical responses showed that EMD was prolonged with the higher load. The findings indicated that in group one, the HF-induced Ca2 + transient remodeling prolonged the mechanical activation time (MAT) and decreased the contractile tension, which reduced the left ventricle (LV) pressure, and increased the end-diastolic strain. In group two, LVAD shortened MAT of the ventricles. Furthermore, LVAD reduced the contractile tension, and end-diastolic strain, but increased the aortic pressure. The computational study demonstrated that LVAD shortens EMD by mechanical unloading of the ventricle.

Effectiveness of atrial fibrillation rotor ablation is dependent on conduction velocity: An in-silico 3-dimensional modeling study.

BACKGROUND: We previously reported that stable rotors are observed in in-silico human atrial fibrillation (AF) models, and are well represented by a dominant frequency (DF). In the current study, we hypothesized that the outcome of DF ablation is affected by conduction velocity (CV) conditions and examined this hypothesis using in-silico 3D-AF modeling. METHODS: We integrated 3D CT images of left atrium obtained from 10 patients with persistent AF (80% male, 61.8±13.5 years old) into in-silico AF model. We compared AF maintenance durations (max 300s), spatiotemporal stabilities of DF, phase singularity (PS) number, life-span of PS, and AF termination or defragmentation rates after virtual DF ablation with 5 different CV conditions (0.2, 0.3, 0.4, 0.5, and 0.6m/s). RESULTS: 1. AF maintenance duration (p<0.001), spatiotemporal mean variance of DF (p<0.001), and the number of PS (p = 0.023) showed CV dependent bimodal patterns (highest at CV0.4m/s and lowest at CV0.6m/s) consistently. 2. After 10% highest DF ablation, AF defragmentation rates were the lowest at CV0.4m/s (37.8%), but highest at CV0.5 and 0.6m/s (all 100%, p<0.001). 3. In the episodes with AF termination or defragmentation followed by 10% highest DF ablation, baseline AF maintenance duration was shorter (p<0.001), spatiotemporal mean variance of DF was lower (p = 0.014), and the number of PS was lower (p = 0.004) than those with failed AF defragmentation after DF ablation. CONCLUSION: Virtual ablation of DF, which may indicate AF driver, was more likely to terminate or defragment AF with spatiotemporally stable DF, but not likely to do so in long-lasting and sustained AF conditions, depending on CV.

Ganglionated plexi stimulation induces pulmonary vein triggers and promotes atrial arrhythmogenecity: In silico modeling study.

BACKGROUND: The role of the autonomic nervous system (ANS) on atrial fibrillation (AF) is difficult to demonstrate in the intact human left atrium (LA) due to technical limitations of the current electrophysiological mapping technique. We examined the effects of the ANS on the initiation and maintenance of AF by employing a realistic in silico human left atrium (LA) model integrated with a model of ganglionated plexi (GPs). METHODS: We incorporated the morphology of the GP and parasympathetic nerves in a three-dimensional (3D) realistic LA model. For the model of ionic currents, we used a human atrial model. GPs were stimulated by increasing the IK[ACh], and sympathetic nerve stimulation was conducted through a homogeneous increase in the ICa-L. ANS-induced wave-dynamics changes were evaluated in a model that integrated a patient's LA geometry, and we repeated simulation studies using LA geometries from 10 different patients. RESULTS: The two-dimensional model of pulmonary vein (PV) cells exhibited late phase 3 early afterdepolarization-like activity under 0.05µM acetylcholine (ACh) stimulation. In the 3D simulation model, PV tachycardia was induced, which degenerated to AF via GP (0.05µM ACh) and sympathetic (7.0×ICa-L) stimulations. Under sustained AF, local reentries were observed at the LA-PV junction. We also observed that GP stimulation reduced the complex fractionated atrial electrogram (CFAE)-cycle length (CL, p<0.01) and the life span of phase singularities (p<0.01). GP stimulation also increased the overlap area of the GP and CFAE areas (CFAE-CL≤120ms, p<0.01). When 3 patterns of virtual ablations were applied to the 3D AF models, circumferential PV isolation including the GP was the most effective in terminating AF. CONCLUSION: Cardiac ANS stimulations demonstrated triggered activity, automaticity, and local reentries at the LA-PV junction, as well as co-localized GP and CFAE areas in the 3D in silico GP model of the LA.

Physiome approach for the analysis of vascular flow reserve in the heart and brain.

This work reviews the key aspects of coronary and neurovascular flow reserves with an emphasis on physiomic modeling characteristics by the use of a variety of numerical approaches. First, we explain the definition of fractional flow reserve (FFR) in coronary artery and introduce its clinical significance. Then, computational researches for obtaining FFR are reviewed, and their clinical outcomes are compared. In the case of cerebrovascular reserve (CVR), in spite of substantial progress in the simulation of cerebral hemodynamics, only a few computational studies exist. Thus, we discuss the limitations of CVR simulation study and suggest the challenging issue to overcome these. Also, the future direction of physiomic researches for the flow reserves in coronary arteries and cerebral arteries is described. Also, we introduce a machine learning algorithm trained by the existing physiomic simulation data of flow reserve and suggest a prospective research direction related to this.

Numerical simulation of motion-induced dynamic noise in a ubiquitous ECG application.

Wearable ubiquitous biomedical applications, such as ECG monitors, can generate dynamic noise as a person moves. However, the source of this noise is not clear. We postulated that the dynamic ECG noise has two causes: the change in displacement of the heart during motion and the change in the electrical impedance of the skin-gel interface due to motion-induced deformation of the skin-gel interface. Using a three-dimensional electrophysiological heart model coupled with a torso model, dynamic noise was simulated, while the displacement of the heart was changed in the vertical and horizontal directions, independently and while the skin-gel interface was deformed during motion. To determine the deformation rate of the skin and sol-gel layers, motion-induced deformation of the two layers was simulated using a three-dimensional finite element method.