Function, metabolic, and flow heterogeneity of the heart: the view is getting better

Detection of myocardial ischemia due to clinically asymptomatic coronary artery stenosis at rest using supervised artificial intelligence-enabled vectorcardiography - A five-fold cross validation of accuracy

BACKGROUND

Coronary artery disease (CAD) is a leading cause of death and disability. Conventional non-invasive diagnostic modalities for the detection of stable CAD at rest are subject to significant limitations: low sensitivity, and personal expertise. We aimed to develop a reliable and time-cost efficient screening tool for the detection of coronary ischemia using machine learning.

METHODS

We developed a supervised artificial intelligence algorithm combined with a five lead vectorcardiography (VCG) approach (i.e. Cardisiography, CSG) for the diagnosis of CAD. Using vectorcardiography, the excitation process of the heart can be described as a three-dimensional signal. A diagnosis can be received, by first, calculating specific physical parameters from the signal, and subsequently, analyzing them with a machine learning algorithm containing neuronal networks. In this multi-center analysis, the primary evaluated outcome was the accuracy of the CSG Diagnosis System, validated by a five-fold nested cross-validation in comparison to angiographic findings as the gold standard. Individuals with 1, 2, or 3- vessel disease were defined as being affected.

RESULTS

Of the 595 patients, 62·0% (n = 369) had 1, 2 or 3- vessel disease identified by coronary angiography. CSG identified a CAD at rest with a sensitivity of 90·2 ± 4·2% for female patients (male: 97·2 ± 3·1%), specificity of 74·4 ± 9·8% (male: 76·1 ± 8·5%), and overall accuracy of 82·5 ± 6·4% (male: 90·7 ± 3·3%).

CONCLUSION

These findings demonstrate that supervised artificial intelligence-enabled vectorcardiography can overcome limitations of conventional non-invasive diagnostic modalities for the detection of coronary ischemia at rest and is capable as a highly valid screening tool.

Reperfusion Microvascular Ischemia After Prolonged Coronary Occlusion: Implications And Treatment With Local Supersaturated Oxygen Delivery

Following a prolonged coronary arterial occlusion, heterogeneously scattered, focal regions of low erythrocyte flow are commonly found throughout the reperfused myocardium. Experimental studies have also demonstrated the presence of widespread, focally patchy regions of microvascular ischemia during reperfusion (RMI). However, the potential contribution of RMI to tissue viability and function has received little attention in the absence of practical clinical methods for its detection. In this review, the anatomic/functional basis of RMI is summarized, along with the evidence for its presence in reperfused myocardium. Advances in microcirculation research related to obstructive responses of vascular endothelial cells and blood elements to the effects of hypoxia and low shear stress are discussed, and a potential cycle of intensification of RMI from such responses and progressive loss of functional capillary density is presented. In capillaries with impaired erythrocyte flow, compensatory increases in the delivery of oxygen, because of its low solubility in plasma, are effective only at high partial pressures. As discussed herein, attenuation of the cycle with oxygen at hyperbaric levels in plasma is, very likely, responsible for improved tissue level perfusion noted experimentally. Observed clinical benefits from intracoronary SuperSaturated oxygen (SSO₂) delivery, including infarct size reduction, can be attributed to attenuation of RMI with improvement in microvascular blood flow.

Optical Cryoimaging Reveals a Heterogeneous Distribution of Mitochondrial Redox State in ex vivo Guinea Pig Hearts and Its Alteration During Ischemia and Reperfusion

Oxidation of substrates to generate ATP in mitochondria is mediated by redox reactions of NADH and FADH2. Cardiac ischemia and reperfusion (IR) injury compromises mitochondrial oxidative phosphorylation. We hypothesize that IR alters the metabolic heterogeneity of mitochondrial redox state of the heart that is only evident in the 3-D optical cryoimaging of the perfused heart before, during, and after IR. The study involved four groups of hearts: time control (TC: heart perfusion without IR), global ischemia (Isch), global ischemia followed by reperfusion (IR) and TC with PCP (a mitochondrial uncoupler) perfusion. Mitochondrial NADH and FAD autofluorescence signals were recorded spectrofluorometrically online in guinea pig ex vivo-perfused hearts in the Langendorff mode. At the end of each specified protocol, hearts were rapidly removed and snap frozen in liquid N2 for later 3-D optical cryoimaging of the mitochondrial NADH, FAD, and NADH/FAD redox ratio (RR). The TC hearts revealed a heterogeneous spatial distribution of NADH, FAD, and RR. Ischemia and IR altered the spatial distribution and caused an overall increase and decrease in the RR by 55% and 64%, respectively. Uncoupling with PCP resulted in the lowest level of the RR (73% oxidation) compared with TC. The 3-D optical cryoimaging of the heart provides novel insights into the heterogeneous distribution of mitochondrial NADH, FAD, RR, and metabolism from the base to the apex during ischemia and IR. This 3-D information of the mitochondrial redox state in the normal and ischemic heart was not apparent in the dynamic spectrofluorometric data.

3D IMAGING OF THE MITOCHONDRIAL REDOX STATE OF RAT HEARTS UNDER NORMAL AND FASTING CONDITIONS

The heart requires continuous ATP availability that is generated in the mitochondria. Although studies using the cell culture and perfused organ models have been carried out to investigate the biochemistry in the mitochondria in response to a change in substrate supply, mitochondrial bioenergetics of heart under normal feed or fasting conditions has not been studied at the tissue level with a sub-millimeter spatial resolution either in vivo or ex vivo. Oxidation of many food-derived metabolites to generate ATP in the mitochondria is realized through the NADH/NAD+ couple acting as a central electron carrier. We employed the Chance redox scanner - the low-temperature fluorescence scanner to image the three-dimensional (3D) spatial distribution of the mitochondrial redox states in heart tissues of rats under normal feeding or an overnight starvation for 14.5 h. Multiple consecutive sections of each heart were imaged to map three redox indices, i.e., NADH, oxidized flavoproteins (Fp, including flavin adenine dinucleotide (FAD)) and the redox ratio NADH/Fp. The imaging results revealed the micro-heterogeneity and the spatial distribution of these redox indices. The quantitative analysis showed that in the fasted hearts the standard deviation of both NADH and Fp, i.e., SD_NADH and SD_Fp, significantly decreased with a p value of 0.032 and 0.045, respectively, indicating that the hearts become relatively more homogeneous after fasting. The fasted hearts contained 28.6% less NADH (p = 0.038). No significant change in Fp was found (p = 0.4). The NADH/Fp ratio decreased with a marginal p value (0.076). The decreased NADH in the fasted hearts is consistent with the cardiac cells' reliance of fatty acids consumption for energy metabolism when glucose becomes scarce. The experimental observation of NADH decrease induced by dietary restriction in the heart at tissue level has not been reported to our best knowledge. The Chance redox scanner demonstrated the feasibility of 3D imaging of the mitochondrial redox state in the heart and provides a useful tool to study heart metabolism and function under normal, dietary-change and pathological conditions at tissue level.

NIR spectroscopic imaging to map hemoglobin + myoglobin oxygenation, their concentration and optical pathlength across a beating pig heart during surgery

The purpose of this paper is to demonstrate that near-infrared (NIR) spectroscopic imaging can provide spatial distribution (maps) of the absolute concentration of hemoglobin + myoglobin, oxygen saturation parameter and optical pathlength, reporting on the biochemico-physiological status of a beating heart in vivo. The method is based on processing the NIR spectroscopic images employing a first-derivative approach. Blood-pressure-controlled gating compensated the effect of heart motion on the imaging. All the maps are available simultaneously and noninvasively at a spatial resolution in the submillimeter range and can be obtained in a couple of minutes. The equipment has no mechanical contact with the tissue, thereby leaving the heart unaffected during the measurement.

QUANTITATIVE DOUBLE-TRACER DIGITALRADIOGRAPHY BASED ON DESMETHYLIMIPRAMINE DEPOSITION: APPLICATION TO STUDIES ON MYOCARDIAL PERFUSION HETEROGENEITY

Desmethylimipramine (DMI), an α2-adrenergic antagonist, is a nearly ideal deposition tracer for evaluating the myocardial flow distribution with the least artifactual effects on microcirculation. Myocardial retentions of tritium- and iodine-125-labeled DMI (HDMI, IDMI) were confirmed to be satisfactory; the retentions of IDMI and HDMI at 1 min were 95 and 91% respectively in isolated Tyrode-perfused rabbit hearts (n=6) at a perfusion rate of 8.1 ml/min/g, and 98 and 96% respectively, in blood-perfused rat hearts (n=4) at a perfusion rate of 3.1 ml/min/g. Using these tracers combined with subtraction digitalradiography, it allowed the assessing of changes of myocardial flow distribution with 400×400 μ m 2 resolution. In blood-perfused rat hearts (n=4), the validity of this method was verified by the strong cross-correlation between regional densities of two tracers injected simultaneously (r=0.94) and the regression line having a slope close to one. Furthermore, in Tyrode-perfused rabbit hearts, the flow distributions were evaluated before and after decreasing perfusion rate moderately by 34% (n=7) and severely by 70% (n=7). Severe flow reduction increased the coefficient of variation of tracer density (CV) significantly from 19 to 25%, but CV did not change with moderate flow reduction (20 vs. 19%). Regional densities of two tracers were cross-correlated still substantially under severe low-flow perfusion (r=0.84). Accordingly, flow differences between originally high- and low-flow regions were enlarged under severe flow reduction. In conclusion, double-tracer digitalradiography based on the DMI deposition will be a potent method for the analysis of flow heterogeneity at microvascular levels.

Origins of heterogeneity in tissue perfusion and metabolism

AIMS

In the heart and other tissues, perfusion and metabolic activity are heterogeneous and spatially correlated. The goal of this work is to investigate the causes of this behaviour. Theoretical simulations are used to examine the effects on flow distribution and oxygen levels in terminal vascular beds of inherent irregularity in network structure, considering structural adaptation of vessel diameters to haemodynamic and metabolic stimuli, and adaptation of oxygen demand to local oxygen availability.

METHODS AND RESULTS

A mathematical model based on experimentally observed microvascular network structures (rat mesentery and m. sartorius) is used to simulate blood flow, oxygen transport, and adaptation of vessel diameters and tissue oxygen demand. Inherent geometric heterogeneities of vascular networks cause heterogeneity of blood flow and oxygen levels that cannot be eliminated by increasing metabolic sensitivity of diameter adaptation. Adaptation of oxygen demand to differences in oxygen availability causes increased oxygen extraction, implying improved functional capacity, and establishes a correlation between local oxygen demand and flow rate, as observed experimentally. Such a correlation is not predicted if the heterogeneity of oxygen demand is assumed to be an intrinsic tissue property.

CONCLUSION

A central mechanism generating heterogeneous perfusion is the inevitable structural heterogeneity of terminal vascular beds, which cannot be fully compensated by structural adaptation of vessel diameters. Heterogeneity of metabolism may result from adaptation of tissue function to the heterogeneous oxygen availability. These findings are of interest for the understanding of tissue function, including the heart, and of results obtained by corresponding imaging approaches.

Microheterogeneity of regional myocardial blood flows in low-perfused rat hearts evaluated by double-tracer digital radiography

Using (3)H- and (125)I-labeled desmethylimipramine (DMI) for regional flow tracers, we established a two-time measurement method for the spatial pattern of myocardial perfusion in cross-circulated rat hearts. Myocardial extractions and retentions of these tracers were confirmed to be satisfactory; however, the latter were less than 90% after 3 min at a perfusion rate of 2.9 ml/min/g, limiting the present application to a short-time perfusion measurement. Distributions of myocardial depositions were separated by subtraction digital radiography with 400-microm pixel resolution. Its feasibility was examined by regression analysis between local deposition densities of (3)H- and (125)I-DMI injected simultaneously. The slope, y-intercept, and correlation coefficient (r) of the regression line were 0.98+/-0.04, 0.02+/-0.04, and 0.95+/-0.03, respectively, indicating the validity of the present image subtraction technique. The spatial pattern of myocardial perfusion in response to flow reduction was evaluated by the injections of (3)H- and (125)I-DMI, respectively, before and after a nearly 70% flow reduction. A significant correlation between normalized density distributions of these tracers was found in both subepicardium (r=0.77+/-0.12) and subendocardium (r=0.73+/-0.20), indicating the stable pattern of myocardial perfusion. However, the coefficient of variation of tracer densities showed a decrease of subendocardial flow heterogeneity from 35+/-15% to 31+/-16%. Thus, flow differences between originally high- and low-flow regions in subendocardium were reduced on a relative basis during low perfusion.

Regional differences of myocardial infarct development and ischemic preconditioning

The spatial and temporal development of myocardial infarction depends on the area at risk (AAR), the severity and duration of blood flow reduction (energy supply) as well as on heart rate and regional wall function (energy demand). Both supply and demand can vary within the AAR of a given heart, potentially resulting in differences in infarct development. We therefore retrospectively analyzed infarct size (IS, %AAR, TTC) in 24 anesthetized pigs in vivo following 90 min hypoperfusion and 120 min reperfusion of the LAD coronary artery, which supplies parts of the LV septum (LVS) and anterior free wall (LVAFW). The total LAD perfusion territory averaged 49.8 +/- 14.2 (SD) g (49.2 +/- 8.4% of LV); 61.4 +/- 8.1% of the AAR was LVAFW. IS within the LVS was 25.3 +/- 15.1%, while IS within the LVAFW was 16.6 +/-10.1% (p<0.05). While ischemic blood flow (radiolabeled microspheres) did not differ between LVS (0.05 +/- 0.02 ml/min/g) and LVAFW (0.05 +/- 0.03 ml/min/g), perivascular connective tissue (56 +/- 9 vs. 38+/-7 microm(2), p < 0.05) and the capillary-to-myocyte distance (1.65 +/- 0.23 vs. 1.18 +/- 0.23 mm, p < 0.05) were larger in LVS than in LVAFW. Interestingly, IS in LVS (9.3 +/- 9.6%, n = 24) and LVAFW (9.2 +/- 9.1%) were reduced to the same absolute extent by ischemic preconditioning with one cycle of 10 min ischemia and 15 min reperfusion, suggesting that a similar regional difference exists also in the protection afforded by ischemic preconditioning. The mechanism(s) for that remain(s) to be established.

CONCLUSION

In pigs, regional differences in infarct development and protection from it exist in the LAD perfusion territory, which are independent of ischemic blood flow but apparently related to pre-existing structural differences.

High-resolution imaging reveals a limit in spatial resolution of blood flow measurements by microspheres

Density of 15-μm microspheres after left atrial application is the standard measure of regional perfusion. In the heart, substantial differences in microsphere density are seen at spatial resolutions <5 ml, implying perfusion heterogeneity. Microsphere deposition imaging permits a superior evaluation of the distribution pattern. Therefore, fluorescent microspheres (FMS) were applied, FMS deposition in the canine heart was imaged by epifluorescence microscopy in vitro, and the patterns were observed compared with MR images of iron oxide microspheres (IMS) obtained in vivo and in vitro. FMS deposition in myocardial slices revealed the following: 1) a nonrandom distribution, with sequentially applied FMS of different color stacked within the same vessel, 2) general FMS clustering, and 3) rather large areas devoid of FMS ( n = 3). This pattern was also seen in reconstructed three-dimensional images (<1 nl resolution) of FMS distribution ( n = 4). Surprisingly, the deposition pattern of sequentially applied FMS remained virtually identical over 3 days. Augmenting flow by intracoronary adenosine (>2 μM) enhanced local microsphere density, but did not alter the deposition pattern ( n = 3). The nonrandom, temporally stable pattern was quantitatively confirmed by a three-dimensional intermicrosphere distance analysis of sequentially applied FMS. T2-weighted short-axis MR images (2-μl resolution) of IMS revealed similar patterns in vivo and in vitro ( n = 6), as seen with FMS. The observed temporally stable microsphere patterns are not consistent with the notion that microsphere deposition is solely governed by blood flow. We propose that at high spatial resolution (<2 μl) structural aspects of the vascular network dominate microsphere distribution, resulting in the organized patterns observed.

Relationship between arterial diameter and perfused tissue volume in myocardial microcirculation: a micro-CT-based analysis

The volume of myocardial tissue that is perfused by an epicardial coronary artery has been shown to be predictably related to the diameter of the epicardial arterial lumen. However, to what extent the intramyocardial microvasculature follows the epicardial rules remains unclear. To explore the relationship between the diameter of coronary arterioles and their subsequent perfused myocardial volumes, we quantified the volume of nonperfused myocardium resulting from an embolized arteriole of a certain diameter. We injected a single dose of microspheres selected from one of nine possible microsphere combinations (10, 30, and 100 μm diameter, each at three possible doses) into the left anterior descending coronary and/or left circumflex arteries of seven anesthetized pigs. At postmortem, the coronary arteries were infused with a radiopaque silicon polymer. Embolized myocardium (1 cm3) was scanned with a microcomputerized tomography scanner and resulted in three-dimensional images that consisted of 20 μm/side cubic voxels and a subvolume of the specimen with 4 μm/side cubic voxels. Image analysis provided the number and volumes of myocardial perfusion defects for each size and dose of microspheres. The smallest individual myocardial perfusion defects, which correspond to the volume of myocardium perfused by a single embolized arteriole, were found to be 0.0004 ± 0.0002, 0.02 ± 0.004, and 0.62 ± 0.099 mm3 for the 10-, 30-, and 100-μm microspheres, respectively. The number of myocardial perfusion defects in the embolized myocardium was inversely related to the dose of the injected microspheres. This reflects a clustering behavior that is consistent with a random distribution process of the individual embolized perfusion defects.

Pattern differences between distributions of microregional myocardial flows in crystalloid- and blood-perfused rat hearts

Regional myocardial flow distributions in Langendorff rat hearts under Tyrode and blood perfusion were assessed by tracer digital radiography (100-μm resolution). Flow distributions during baseline and maximal hyperemia following a 60-s flow cessation were evaluated by the coefficient of variation of regional flows (CV; related to global flow heterogeneity) and the correlation between adjacent regional flows (CA; inversely related to local flow randomness). These values were obtained for the original images (642 pixels) and for coarse-grained images (322, 162, and 82 blocks of nearby pixels). At a given point in time during baseline, both CV and CA were higher in blood ( n = 7) than in Tyrode perfusion ( n = 7) over all pixel aggregates ( P < 0.05, two-way ANOVA). During the maximal hyperemia, CV and CA were still significantly higher in blood ( n = 7) than in Tyrode perfusion ( n = 7); however, these values decreased substantially in blood perfusion and the CV and CA differences became smaller than those at baseline accordingly. During basal blood perfusion, the 60-s average flow distribution ( n = 7) showed a smaller CV and CA than those at a given point in time ( P < 0.05, two-way ANOVA). Coronary flow reserve was significantly higher in blood than in Tyrode perfusion. In conclusion, the flow heterogeneity and the local flow similarity are both higher in blood than in Tyrode perfusion, probably due to the different degree of coronary tone preservation and the presence or absence of blood corpuscles. Under blood perfusion, temporal flow fluctuations over 60-s order are largely involved in shaping microregional flow distributions.

Fractal properties of perfusion heterogeneity in optimized arterial trees: a model study

Regional blood flows in the heart muscle are remarkably heterogeneous. It is very likely that the most important factor for this heterogeneity is the metabolic need of the tissue rather than flow dispersion by the branching network of the coronary vasculature. To model the contribution of tissue needs to the observed flow heterogeneities we use arterial trees generated on the computer by constrained constructive optimization. This method allows to prescribe terminal flows as independent boundary conditions, rather than obtaining these flows by the dispersive effects of the tree structure. We study two specific cases: equal terminal flows (model 1) and terminal flows set proportional to the volumes of Voronoi polyhedra used as a model for blood supply regions of terminal segments (model 2). Model 1 predicts, depending on the number Nterm of end-points, fractal dimensions D of perfusion heterogeneities in the range 1.20 to 1.40 and positively correlated nearest-neighbor regional flows, in good agreement with experimental data of the normal heart. Although model 2 yields reasonable terminal flows well approximated by a lognormal distribution, it fails to predict D and nearest-neighbor correlation coefficients r1 of regional flows under normal physiologic conditions: model 2 gives D = 1.69 +/- 0.02 and r1 = -0.18 +/- 0.03 (n = 5), independent of Nterm and consistent with experimental data observed under coronary stenosis and under the reduction of coronary perfusion pressure. In conclusion, flow heterogeneity can be modeled by terminal positions compatible with an existing tree structure without resorting to the flow-dispersive effects of a specific branching tree model to assign terminal flows.

Myocardial proteome analysis reveals reduced NOS inhibition and enhanced glycolytic capacity in areas of low local blood flow

In the heart, in situ local myocardial blood flow (MBF) varies greater than 10-fold between individual areas and displays a spatially heterogeneous pattern. To analyze its molecular basis, we analyzed protein expression of low and high flow samples (300 mg, <50% or >150% of mean MBF, each n=30) of six beagle dogs by 2-D polyacrylamide gel electrophoresis (380 +/- 78 spots/gel). In low flow samples, dimethylarginine dimethylaminohydrolase (DDAH1) was increased greatly (+377%, compared with high flow samples). This increase resulted in a 75% reduction of asymmetric dimethylarginine (ADMA), the potent endogenous inhibitor of NO synthase, whereas eNOS showed no difference. Low flow samples exhibited enhanced expression of GAPDH (+89%) and phosphoglycerate kinase (+100%), whereas hydroxyacyl-CoA dehydrogenase, electron transfer flavoprotein, myoglobin, and desmin were decreased. Assessing local MBF on different days within 2 weeks revealed a high degree of MBF stability (r2 > 0.79). Thus, stable differences in local MBF are associated with significant differences in local gene and protein expression. In low flow areas, the increased DDAH1 reduces ADMA concentration and NOS inhibition, which strongly suggests enhanced NO formation. Low flow areas are also characterized by a higher glycolytic and a lower fatty acid oxidation capacity. Both the shift in substrate utilization and the rise in NO may contribute to the known lower oxygen consumption in these areas.