Identification of supraventricular tachycardia mechanisms with surface electrocardiograms using a convolutional neural network.

Background: It remains difficult to definitively distinguish supraventricular tachycardia (SVT) mechanisms using a 12-lead electrocardiogram (ECG) alone. Machine learning may identify visually imperceptible changes on 12-lead ECGs and may improve ability to determine SVT mechanisms. Objective: We sought to develop a convolutional neural network (CNN) that identifies the SVT mechanism according to the gold standard of SVT ablation and to compare CNN performance against experienced electrophysiologists among patients with atrioventricular nodal re-entrant tachycardia (AVNRT), atrioventricular reciprocating tachycardia (AVRT), and atrial tachycardia (AT). Methods: All patients with 12-lead surface ECG during sinus rhythm and SVT and had successful SVT ablation from 2013 to 2020 were included. A CNN was trained using data from 1505 surface ECGs that were split into 1287 training and 218 test ECG datasets. We compared the CNN performance against independent adjudication by 2 experienced cardiac electrophysiologists on the test dataset. Results: Our dataset comprised 1505 ECGs (368 AVNRT, 304 AVRT, 95 AT, and 738 sinus rhythm) from 725 patients. The CNN areas under the receiver-operating characteristic curve for AVNRT, AVRT, and AT were 0.909, 0.867, and 0.817, respectively. When fixing the specificity of the CNN to the electrophysiologist adjudicators' specificity, the CNN identified all SVT classes with higher sensitivity: (1) AVNRT (91.7% vs 65.9%), (2) AVRT (78.4% vs 63.6%), and (3) AT (61.5% vs 50.0%). Conclusion: A CNN can be trained to differentiate SVT mechanisms from surface 12-lead ECGs with high overall performance, achieving similar performance to experienced electrophysiologists at fixed specificities.

Dynamic loading of immature epiphyseal cartilage pumps nutrients out of vascular canals.

The potential influence of mechanical loading on transvascular transport in vascularized soft tissues has not been explored extensively. This experimental investigation introduced and explored the hypothesis that dynamic mechanical loading can pump solutes out of blood vessels and into the surrounding tissue, leading to faster uptake and higher solute concentrations than could otherwise be achieved under unloaded conditions. Immature epiphyseal cartilage was used as a model tissue system, with fluorescein (332 Da), dextran (3, 10, and 70 kDa) and transferrin (80 kDa) as model solutes. Cartilage disks were either dynamically loaded (± 10% compression over a 10% static offset strain, at 0.2 Hz) or maintained unloaded in solution for up to 20 h. Results demonstrated statistically significant solute uptake in dynamically loaded (DL) explants relative to passive diffusion (PD) controls for all solutes except unbound fluorescein, as evidenced by the DL:PD concentration ratios after 20 h (1.0 ± 0.2, 2.4 ± 1.1, 6.1 ± 3.3, 9.0 ± 4.0, and 5.5 ± 1.6 for fluorescein, 3, 10, and 70 kDa dextran, and transferrin). Significant uptake enhancements were also observed within the first 30s of loading. Termination of dynamic loading produced dissipation of enhanced solute uptake back to PD control values. Confocal images confirmed that solute uptake occurred from cartilage canals into their surrounding extracellular matrix. The incidence of this loading-induced transvascular solute pumping mechanism may significantly alter our understanding of the interaction of mechanical loading and tissue metabolism.

Validation of theoretical framework explaining active solute uptake in dynamically loaded porous media.

Solute transport in biological tissues is a fundamental process necessary for cell metabolism. In connective soft tissues, such as articular cartilage, cells are embedded within a dense extracellular matrix that hinders the transport of solutes. However, according to a recent theoretical study (Mauck et al., 2003, J. Biomech. Eng. 125, 602-614), the convective motion of a dynamically loaded porous solid matrix can also impart momentum to solutes, pumping them into the tissue and giving rise to concentrations which exceed those achived under passive diffusion alone. In this study, the theoretical predictions of this model are verified against experimental measurements. The mechanical and transport properties of an agarose-dextran model system were characterized from independent measurements and substituted into the theory to predict solute uptake or desorption under dynamic mechanical loading for various agarose concentrations and dextran molecular weights, as well as different boundary and initial conditions. In every tested case, agreement was observed between experiments and theoretical predictions as assessed by coefficients of determination ranging from R(2)=0.61 to 0.95. These results provide strong support for the hypothesis that dynamic loading of a deformable porous tissue can produce active transport of solutes via a pumping mechanisms mediated by momentum exchange between the solute and solid matrix.

Influence of the partitioning of osmolytes by the cytoplasm on the passive response of cells to osmotic loading.

Due to the dense organization of organelles, cytoskeletal elements, and protein complexes that make up the intracellular environment, it is likely that membrane-permeant solutes may be excluded from a fraction of the interstitial space of the cytoplasm via steric restrictions, electrostatic interactions, and other long-range intermolecular forces. This study investigates the hypothesis that the intracellular partitioning of membrane-permeant solutes manifests itself as a partial volume recovery in response to hyperosmotic loading, based on prior theoretical and biomimetic experimental studies. Osmotic loading experiments are performed on immature bovine chondrocytes using culture conditions where regulatory volume responses are shown to be insignificant. Osmotic loading with membrane-permeant glycerol (92 Da) and urea (60 Da) are observed to produce partial volume recoveries consistent with the proposed hypothesis, whereas loading with 1,2-propanediol (76 Da) produces complete volume recovery. Combining these experimental results with the previous theoretical framework produces a measure for the intracellular partition coefficient of each of these solutes. At 1000 mOsm, 1,2-propanediol is the only osmolyte to yield a partition coefficient not statistically different from unity, kappa(p)(i) = 1.00 +/- 0.02. For glycerol, the partition coefficient increases with osmolarity from kappa(p)(i) = 0.48 +/- 0.19 at 200 mOsm to kappa(p)(i) = 0.80 +/- 0.07 at 1000 mOsm; urea exhibits no such dependence, with an average value of kappa(p)(i) = 0.87 +/- 0.07 for all osmolarities from 200 to 1000 mOsm. The finding that intracellular partitioning of membrane-permeant solutes manifests itself as a partial volume recovery under osmotic loading offers a simple method for characterizing the partition coefficient. These measurements suggest that significant partitioning may occur even for small membrane-permeant osmolytes. Furthermore, a positive correlation is observed, suggesting that a solute's cytoplasmic partition coefficient increases with increasing hydrophobicity.

Characterization of the Concentration-Dependence of Solute Diffusivity and Partitioning in a Model Dextran-Agarose Transport System.

This study reports experimental measurements of solute diffusivity and partition coefficient for various solute concentrations and gel porosities, and proposes novel constitutive relations to describe these observed values. The longer-term aim is to explore the theoretical ramifications of accommodating variations in diffusivity and partition coefficient with solute concentration and tissue porosity, and investigate whether they might suggest novel mechanisms not previously recognized in the field of solute transport in deformable porous media. The study implements a model transport system of agarose hydrogels to investigate the effect of solute concentration and hydrogel porosity on the transport of dextran polysaccharides. The proposed phenomenological constitutive relations are shown to provide better fits of experimental results than prior models proposed in the literature based on the microstructure of the gel. While these constitutive models were developed for the transport of dextran in agarose hydrogels, it is expected that they may also be applied to the transport of similar molecular weight solutes in other porous media. This quantification can assist in the application of biophysical models that describe biological transport in deformable tissues, as well as the cell cytoplasm.

Dynamic loading of deformable porous media can induce active solute transport.

Active solute transport mediated by molecular motors across porous membranes is a well-recognized mechanism for transport across the cell membrane. In contrast, active transport mediated by mechanical loading of porous media is a non-intuitive mechanism that has only been predicted recently from theory, but not yet observed experimentally. This study uses agarose hydrogel and dextran molecules as a model experimental system to explore this mechanism. Results show that dynamic loading can enhance the uptake of dextran by a factor greater than 15 over passive diffusion, for certain combinations of gel concentration and dextran molecular weight. Upon cessation of loading, the concentration reverts back to that achieved under passive diffusion. Thus, active solute transport in porous media can indeed be mediated by cyclical mechanical loading.