Deriving phenotype-representative left ventricular flow patterns by reduced-order modeling and classification.

BACKGROUND: Extracting phenotype-representative flow patterns and their associated numerical metrics is a bottleneck in the clinical translation of advanced cardiac flow imaging modalities. We hypothesized that reduced-order models (ROMs) are a suitable strategy for deriving simple and interpretable clinical metrics of intraventricular flow suitable for further assessments. Combined with machine learning (ML) flow-based ROMs could provide new insight to help diagnose and risk-stratify patients. METHODS: We analyzed 2D color-Doppler echocardiograms of 81 non-ischemic dilated cardiomyopathy (DCM) patients, 51 hypertrophic cardiomyopathy (HCM) patients, and 77 normal volunteers (Control). We applied proper orthogonal decomposition (POD) to build patient-specific and cohort-specific ROMs of LV flow. Each ROM aggregates a low number of components representing a spatially dependent velocity map modulated along the cardiac cycle by a time-dependent coefficient. We tested three classifiers using deliberately simple ML analyses of these ROMs with varying supervision levels. In supervised models, hyperparameter grid search was used to derive the ROMs that maximize classification power. The classifiers were blinded to LV chamber geometry and function. We ran vector flow mapping on the color-Doppler sequences to help visualize flow patterns and interpret the ML results. RESULTS: POD-based ROMs stably represented each cohort through 10-fold cross-validation. The principal POD mode captured >80 % of the flow kinetic energy (KE) in all cohorts and represented the LV filling/emptying jets. Mode 2 represented the diastolic vortex and its KE contribution ranged from <1 % (HCM) to 13 % (DCM). Semi-unsupervised classification using patient-specific ROMs revealed that the KE ratio of these two principal modes, the vortex-to-jet (V2J) energy ratio, is a simple, interpretable metric that discriminates DCM, HCM, and Control patients. Receiver operating characteristic curves using V2J as classifier had areas under the curve of 0.81, 0.91, and 0.95 for distinguishing HCM vs. Control, DCM vs. Control, and DCM vs. HCM, respectively. CONCLUSIONS: Modal decomposition of cardiac flow can be used to create ROMs of normal and pathological flow patterns, uncovering simple interpretable flow metrics with power to discriminate disease states, and particularly suitable for further processing using ML.

Cardiac Stasis Imaging, Stroke and Silent Brain Infarcts in Patients with Non-Ischemic Dilated Cardiomyopathy.

Cardioembolic stroke is one of the most devastating complications of non-ischemic dilated cardiomyopathy (NIDCM). However, in clinical trials of primary prevention, the benefits of anticoagulation are hampered by the risk of bleeding. Indices of cardiac blood stasis may account for the risk of stroke and be useful to individualize primary prevention treatments. We performed a cross-sectional study in patients with NIDCM and no history of atrial fibrillation (AF) from two sources: 1) a prospective enrollment of unselected patients with left ventricular (LV) ejection fraction <45% and 2) a retrospective identification of patients with a history of previous cardioembolic neurological event. The primary endpoint integrated a history of ischemic stroke or the presence intraventricular thrombus, or a silent brain infarction (SBI) by imaging. From echocardiography, we calculated blood flow inside the LV, its residence time (RT) maps and its derived stasis indices. Of the 89 recruited patients, 18 showed a positive endpoint: 9 had a history stroke or TIA and 9 were diagnosed with SBIs in the brain imaging. Averaged RT, performed good to identify the primary endpoint (AUC (95% CI)= 0.75 (0.61-0.89), p= 0.001). When accounting only for identifying a history of stroke or TIA, AUC for was 0.92 (0.85-1.00) with and odds ratio= 7.2 (2.3 - 22.3) per cycle, p< 0.001. These results suggest that, in patients with NIDCM in sinus rhythm, stasis imaging derived from echocardiography may account for the burden of stroke.

Deriving Explainable Metrics of Left Ventricular Flow by Reduced-Order Modeling and Classification.

Background: Extracting explainable flow metrics is a bottleneck to the clinical translation of advanced cardiac flow imaging modalities. We hypothesized that reduced-order models (ROMs) of intraventricular flow are a suitable strategy for deriving simple and interpretable clinical metrics suitable for further assessments. Combined with machine learning (ML) flow-based ROMs could provide new insight to help diagnose and risk-stratify patients. Methods: We analyzed 2D color-Doppler echocardiograms of 81 non-ischemic dilated cardiomyopathy (DCM) patients, 51 hypertrophic cardiomyopathy (HCM) patients, and 77 normal volunteers (Control). We applied proper orthogonal decomposition (POD) to build patient-specific and cohort-specific ROMs of LV flow. Each ROM aggregates a low number of components representing a spatially dependent velocity map modulated along the cardiac cycle by a time-dependent coefficient. We tested three classifiers using deliberately simple ML analyses of these ROMs with varying supervision levels. In supervised models, hyperparameter gridsearch was used to derive the ROMs that maximize classification power. The classifiers were blinded to LV chamber geometry and function. We ran vector flow mapping on the color-Doppler sequences to help visualize flow patterns and interpret the ML results. Results: POD-based ROMs stably represented each cohort through 10-fold cross-validation. The principal POD mode captured >80% of the flow kinetic energy (KE) in all cohorts and represented the LV filling/emptying jets. Mode 2 represented the diastolic vortex and its KE contribution ranged from <1% (HCM) to 13% (DCM). Semi-unsupervised classification using patient-specific ROMs revealed that the KE ratio of these two principal modes, the vortex-to-jet (V2J) energy ratio, is a simple, interpretable metric that discriminates DCM, HCM, and Control patients. Receiver operating characteristic curves using V2J as classifier had areas under the curve of 0.81, 0.91, and 0.95 for distinguishing HCM vs. Control, DCM vs. Control, and DCM vs. HCM, respectively. Conclusions: Modal decomposition of cardiac flow can be used to create ROMs of normal and pathological flow patterns, uncovering simple interpretable flow metrics with power to discriminate disease states, and particularly suitable for further processing using ML.

A new species of Struthanthus Mart. (Loranthaceae) from Oaxaca, Mexico.

Struthanthusibe-dzisp. nov. is a new species described and illustrated from the cloud and pine-oak forests of the Sierra Madre del Sur in Oaxaca, Mexico. This species shares similarities of leaf shape and inflorescence type with S.deppeanus, S.quercicola, and S.ramiro-cruzii. However, S.ibe-dzi can be recognized by its glaucous branches, leaves and inflorescences; compressed nodes; convoluted distal half of styles in pistillate flowers; and staminate flowers with asymmetrical thecae and an extended connective forming an apiculate horn in both anther series. A distribution map and an identification key are provided to separate S.ibe-dzi from morphologically similar congeners present in the region.

Clinical assessment of intraventricular blood transport in patients undergoing cardiac resynchronization therapy.

In the healthy heart, left ventricular (LV) filling generates different flow patterns which have been proposed to optimize blood transport by coupling diastole and systole. This work presents a novel image-based method to assess how different flow patterns influence LV blood transport in patients undergoing cardiac resynchronization therapy (CRT). Our approach is based on solving the advection equation for a passive scalar field from time-resolved blood velocity fields. Imposing time-varying inflow boundary conditions for the scalar field provides a straightforward method to distinctly track the transport of blood entering the LV in the different filling waves of a given cardiac cycle, as well as the transport barriers which couple filling and ejection. We applied this method to analyze flow transport in a group of patients with implanted CRT devices and a group of healthy volunteers. Velocity fields were obtained using echocardiographic color Doppler velocimetry, which provides two-dimensional time-resolved flow maps in the apical long axis three-chamber view of the LV. In the patients under CRT, the device programming was varied to analyze flow transport under different values of the atrioventricular conduction delay, and to model tachycardia (100 bpm). Using this method, we show how CRT influences the transit of blood inside the left ventricle, contributes to conserving kinetic energy, and favors the generation of hemodynamic forces that accelerate blood in the direction of the LV outflow tract. These novel aspects of ventricular function are clinically accessible by quantitative analysis of color-Doppler echocardiograms.

A clinical method for mapping and quantifying blood stasis in the left ventricle.

In patients at risk of intraventrcular thrombosis, the benefits of chronic anticoagulation therapy need to be balanced with the pro-hemorrhagic effects of therapy. Blood stasis in the cardiac chambers is a recognized risk factor for intracardiac thrombosis and potential cardiogenic embolic events. In this work, we present a novel flow image-based method to assess the location and extent of intraventricular stasis regions inside the left ventricle (LV) by digital processing flow-velocity images obtained either by phase-contrast magnetic resonance (PCMR) or 2D color-Doppler velocimetry (echo-CDV). This approach is based on quantifying the distribution of the blood Residence Time (TR) from time-resolved blood velocity fields in the LV. We tested the new method in illustrative examples of normal hearts, patients with dilated cardiomyopathy and one patient before and after the implantation of a left ventricular assist device (LVAD). The method allowed us to assess in-vivo the location and extent of the stasis regions in the LV. Original metrics were developed to integrate flow properties into simple scalars suitable for a robust and personalized assessment of the risk of thrombosis. From a clinical perspective, this work introduces the new paradigm that quantitative flow dynamics can provide the basis to obtain subclinical markers of intraventricular thrombosis risk. The early prediction of LV blood stasis may result in decrease strokes by appropriate use of anticoagulant therapy for the purpose of primary and secondary prevention. It may also have a significant impact on LVAD device design and operation set-up.