Artificial Intelligence-Assisted Auscultation of Heart Murmurs: Validation by Virtual Clinical Trial.

Artificial intelligence (AI) has potential to improve the accuracy of screening for valvular and congenital heart disease by auscultation. However, despite recent advances in signal processing and classification algorithms focused on heart sounds, clinical acceptance of this technology has been limited, in part due to lack of objective performance data. We hypothesized that a heart murmur detection algorithm could be quantitatively and objectively evaluated by virtual clinical trial. All cases from the Johns Hopkins Cardiac Auscultatory Recording Database (CARD) with either a pathologic murmur, an innocent murmur or no murmur were selected. The test algorithm, developed independently of CARD, analyzed each recording using an automated batch processing protocol. 3180 heart sound recordings from 603 outpatient visits were selected from CARD. Algorithm estimation of heart rate was similar to gold standard. Sensitivity and specificity for detection of pathologic cases were 93% (CI 90-95%) and 81% (CI 75-85%), respectively, with accuracy 88% (CI 85-91%). Performance varied according to algorithm certainty measure, age of patient, heart rate, murmur intensity, location of recording on the chest and pathologic diagnosis. This is the first reported comprehensive and objective evaluation of an AI-based murmur detection algorithm to our knowledge. The test algorithm performed well in this virtual clinical trial. This strategy can be used to efficiently compare performance of other algorithms against the same dataset and improve understanding of the potential clinical usefulness of AI-assisted auscultation.

Computerized Automatic Diagnosis of Innocent and Pathologic Murmurs in Pediatrics: A Pilot Study.

OBJECTIVE: Computer-aided auscultation in the differentiation of pathologic (AHA class I) from no or innocent murmurs (AHA class III) would be of great value to the general practitioner. This would allow objective screening for structural heart disease, standardized documentation of auscultation findings, and may avoid unnecessary referrals to pediatric cardiologists. Our goal was to assess the quality of a novel computerized algorithm that automatically classifies murmurs in phonocardiograms (PCGs) acquired in a pediatric population. DESIGN: This is a pilot study testing the ability of a novel computerized algorithm to accurately diagnose PCGs compared with interpreted echocardiograms as a gold standard. SETTING: This study was performed in pediatric cardiology clinics at a tertiary care hospital. PATIENTS: All incoming patients were recruited, including patients with no murmurs, innocent murmurs, and pathologic murmurs (106 patients). INTERVENTION: Using an electronic stethoscope, PCGs were acquired by the pediatric cardiologist from each patient. The PCGs were analyzed by the algorithm and diagnoses were compared with findings by echocardiograms interpreted by pediatric cardiologists which were used as the gold standard. OUTCOME MEASURES: Sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were calculated. RESULTS: When compared with echocardiography as a gold standard in diagnosing murmurs, the computerized algorithm tested on N=34 PCGs, yielded a sensitivity of 87% and specificity of 100%, a positive predictive value of 100%, negative predictive value of 90% and an accuracy of 94%. CONCLUSION: With echocardiogram as a gold standard, this computerized algorithm can detect pathologic murmurs with high sensitivity, specificity and accuracy, comparable to if not better than published results of pediatric cardiologists and neonatologists. This study confirms the high quality and "real-world" robustness of a novel computational algorithm in the assessment of pediatric murmurs.

A microstructurally based continuum model of cartilage viscoelasticity and permeability incorporating measured statistical fiber orientations.

The remarkable mechanical properties of cartilage derive from an interplay of isotropically distributed, densely packed and negatively charged proteoglycans; a highly anisotropic and inhomogeneously oriented fiber network of collagens; and an interstitial electrolytic fluid. We propose a new 3D finite strain constitutive model capable of simultaneously addressing both solid (reinforcement) and fluid (permeability) dependence of the tissue's mechanical response on the patient-specific collagen fiber network. To represent fiber reinforcement, we integrate the strain energies of single collagen fibers-weighted by an orientation distribution function (ODF) defined over a unit sphere-over the distributed fiber orientations in 3D. We define the anisotropic intrinsic permeability of the tissue with a structure tensor based again on the integration of the local ODF over all spatial fiber orientations. By design, our modeling formulation accepts structural data on patient-specific collagen fiber networks as determined via diffusion tensor MRI. We implement our new model in 3D large strain finite elements and study the distributions of interstitial fluid pressure, fluid pressure load support and shear stress within a cartilage sample under indentation. Results show that the fiber network dramatically increases interstitial fluid pressure and focuses it near the surface. Inhomogeneity in the tissue's composition also increases fluid pressure and reduces shear stress in the solid. Finally, a biphasic neo-Hookean material model, as is available in commercial finite element codes, does not capture important features of the intra-tissue response, e.g., distributions of interstitial fluid pressure and principal shear stress.

Constitutive modelling of arteries considering fibre recruitment and three-dimensional fibre distribution.

Structurally motivated material models may provide increased insights into the underlying mechanics and physics of arteries under physiological loading conditions. We propose a multiscale model for arterial tissue capturing three different scales (i) a single collagen fibre; (ii) bundle of collagen fibres; and (iii) collagen network within the tissue. The waviness of collagen fibres is introduced by a probability density function for the recruitment stretch at which the fibre starts to bear load. The three-dimensional distribution of the collagen fibres is described by an orientation distribution function using the bivariate von Mises distribution, and fitted to experimental data. The strain energy for the tissue is decomposed additively into a part related to the matrix material and a part for the collagen fibres. Volume fractions account for the matrix/fibre constituents. The proposed model only uses two parameters namely a shear modulus of the matrix material and a (stiffness) parameter related to a single collagen fibre. A fit of the multiscale model to representative experimental data obtained from the individual layers of a human thoracic aorta shows that the proposed model is able to adequately capture the nonlinear and anisotropic behaviour of the aortic layers.

An affine continuum mechanical model for cross-linked F-actin networks with compliant linker proteins.

Cross-linked actin networks are important building blocks of the cytoskeleton. In order to gain deeper insight into the interpretation of experimental data on actin networks, adequate models are required. In this paper we introduce an affine constitutive network model for cross-linked F-actin networks based on nonlinear continuum mechanics, and specialize it in order to reproduce the experimental behavior of in vitro reconstituted model networks. The model is based on the elastic properties of single filaments embedded in an isotropic matrix such that the overall properties of the composite are described by a free-energy function. In particular, we are able to obtain the experimentally determined shear and normal stress responses of cross-linked actin networks typically observed in rheometer tests. In the present study an extensive analysis is performed by applying the proposed model network to a simple shear deformation. The single filament model is then extended by incorporating the compliance of cross-linker proteins and further extended by including viscoelasticity. All that is needed for the finite element implementation is the constitutive model for the filaments, the linkers and the matrix, and the associated elasticity tensor in either the Lagrangian or Eulerian formulation. The model facilitates parameter studies of experimental setups such as micropipette aspiration experiments and we present such studies to illustrate the efficacy of this modeling approach.

Advances in the mechanical modeling of filamentous actin and its cross-linked networks on multiple scales.

The protein actin is a part of the cytoskeleton and, therefore, responsible for the mechanical properties of the cells. Starting with the single molecule up to the final structure, actin creates a hierarchical structure of several levels exhibiting a remarkable behavior. The hierarchy spans several length scales and limitations in computational power; therefore, there is a call for different mechanical modeling approaches for the different scales. On the molecular level, we may consider each atom in molecular dynamics simulations. Actin forms filaments by combining the molecules into a double helix. In a model, we replace molecular subdomains using coarse-graining methods, allowing the investigation of larger systems of several atoms. These models on the nanoscale inform continuum mechanical models of large filaments, which are based on worm-like chain models for polymers. Assemblies of actin filaments are connected with cross-linker proteins. Models with discrete filaments, so-called Mikado models, allow us to investigate the dependence of the properties of networks on the parameters of the constituents. Microstructurally motivated continuum models of the networks provide insights into larger systems containing cross-linked actin networks. Modeling of such systems helps to gain insight into the processes on such small scales. On the other hand, they call for verification and hence trigger the improvement of established experiments and the development of new methods.

A new approach to model cross-linked actin networks: multi-scale continuum formulation and computational analysis.

The mechanical properties of a cell are defined mainly by the cytoskeleton. One contributor within this three-dimensional structure is the actin cortex which is located underneath the lipid bilayer. It forms a nearly isotropic and densely cross-linked protein network. We present a continuum mechanical formulation for describing the mechanical properties of in vitro model systems based on their micro-structure, i.e. the behavior of a single filament and its spatial arrangement. The network is considered elastic, viscous effects being neglected. Filamentous actin is a biopolymer with a highly nonlinear force-stretch relationship. This can be well described by a worm-like chain model that includes extensibility of the filament, which we call the ß-model. A comparison with experimental data shows good agreement with values for the physically interpretable parameters. To make these properties applicable to three dimensions we used a non-affine micro-sphere network, which accounts for filaments, equally distributed in space. The assembled model results in a strain-energy density which is a function of the deformation gradient, and it is validated with experimental data from rheological experiments of in vitro reconstituted actin networks. The Cauchy stress and elasticity tensors are obtained within the continuum mechanics framework and implemented into a finite element program to solve boundary-value problems.

Viscoelasticity of cross-linked actin networks: experimental tests, mechanical modeling and finite-element analysis.

Filamentous actin is one of the main constituents of the eukaryotic cytoskeleton. The actin cortex, a densely cross-linked network, resides underneath the lipid bilayer. In the present work we propose a continuum mechanical formulation for describing the viscoelastic properties of in vitro actin networks, which serve as model systems for the cortex, by including the microstructure, i.e. the behavior of a single filament and its spatial arrangement. The modeling of the viscoelastic response in terms of physically interpretable parameters is conducted using a multiscale approach consisting of two steps: modeling of the single filament response of F-actin by a worm-like chain model including the extensibility of the filament, and assembling the three-dimensional biopolymer network by using the microsphere model which accounts for filaments equally distributed in space. The viscoelastic effects of the network are taken into account using a generalized Maxwell model. The Cauchy stress and elasticity tensors are obtained within a continuum mechanics framework and implemented into a finite-element program. The model is validated on the network level using large strain experiments on reconstituted actin gels. Comparisons of the proposed model with rheological experiments recover reasonable values for the material parameters. Finite-element simulations of the indentation of a sphere on a network slab and the aspiration of a droplet in a micropipette allow for further insights of the viscoelastic behavior of actin networks.