Orbital learning: a novel, actively orchestrated decentralised learning for healthcare.

A novel collaborative and continual learning across a network of decentralised healthcare units, avoiding identifiable data-sharing capacity, is proposed. Currently available methodologies, such as federated learning and swarm learning, have demonstrated decentralised learning. However, the majority of them face shortcomings that affect their performance and accuracy. These shortcomings include a non-uniform rate of data accumulation, non-uniform patient demographics, biased human labelling, and erroneous or malicious training data. A novel method to reduce such shortcomings is proposed in the present work through selective grouping and displacing of actors in a network of many entities for intra-group sharing of learning with inter-group accessibility. The proposed system, known as Orbital Learning, incorporates various features from split learning and ensemble learning for a robust and secure performance of supervised models. A digital embodiment of the information quality and flow within a decentralised network, this platform also acts as a digital twin of healthcare network. An example of ECG classification for arrhythmia with 6 clients is used to analyse its performance and is compared against federated learning. In this example, four separate experiments are conducted with varied configurations, such as varied age demographics and clients with data tampering. The results obtained show an average area under receiver operating characteristic curve (AUROC) of 0.819 (95% CI 0.784-0.853) for orbital learning whereas 0.714 (95% CI 0.692-0.736) for federated learning. This result shows an increase in overall performance and establishes that the proposed system can address the majority of the issues faced by existing decentralised learning methodologies. Further, a scalability demo conducted establishes the versatility and scalability of this platform in handling state-of-the-art large language models.

Quantifying the efficacy of voltage protocols in characterising ion channel kinetics: A novel information-theoretic approach.

Voltage-clamp experiments are commonly utilised to characterise cellular ion channel kinetics. In these experiments, cells are stimulated using a known time-varying voltage, referred to as the voltage protocol, and the resulting cellular response, typically in the form of current, is measured. Parameters of models that describe ion channel kinetics are then estimated by solving an inverse problem which aims to minimise the discrepancy between the predicted response of the model and the actual measured cell response. In this paper, a novel framework to evaluate the information content of voltage-clamp protocols in relation to ion channel model parameters is presented. Additional quantitative information metrics that allow for comparisons among various voltage protocols are proposed. These metrics offer a foundation for future optimal design frameworks to devise novel, information-rich protocols. The efficacy of the proposed framework is evidenced through the analysis of seven voltage protocols from the literature. By comparing known numerical results for inverse problems using these protocols with the information-theoretic metrics, the proposed approach is validated. The essential steps of the framework are: (i) generate random samples of the parameters from chosen prior distributions; (ii) run the model to generate model output (current) for all samples; (iii) construct reduced-dimensional representations of the time-varying current output using proper orthogonal decomposition (POD); (iv) estimate information-theoretic metrics such as mutual information, entropy equivalent variance, and conditional mutual information using non-parametric methods; (v) interpret the metrics; for example, a higher mutual information between a parameter and the current output suggests the protocol yields greater information about that parameter, resulting in improved identifiability; and (vi) integrate the information-theoretic metrics into a single quantitative criterion, encapsulating the protocol's efficacy in estimating model parameters.

An AI based digital-twin for prioritising pneumonia patient treatment.

A digital-twin based three-tiered system is proposed to prioritise patients for urgent intensive care and ventilator support. The deep learning methods are used to build patient-specific digital-twins to identify and prioritise critical cases amongst severe pneumonia patients. The three-tiered strategy is proposed to generate severity indices to: (1) identify urgent cases, (2) assign critical care and mechanical ventilation, and (3) discontinue mechanical ventilation and critical care at the optimal time. The severity indices calculated in the present study are the probability of death and the probability of requiring mechanical ventilation. These enable the generation of patient prioritisation lists and facilitates the smooth flow of patients in and out of Intensive Therapy Units (ITUs). The proposed digital-twin is built on pre-trained deep learning models using data from more than 1895 pneumonia patients. The severity indices calculated in the present study are assessed using the standard benchmark of Area Under Receiving Operating Characteristic Curve (AUROC). The results indicate that the ITU and mechanical ventilation can be prioritised correctly to an AUROC value as high as 0.89. This model may be employed in its current form to COVID-19 patients, but transfer learning with COVID-19 patient data will improve the predictions. The digital-twin model developed and tested is available via accompanying Supplemental material.

Automating fractional flow reserve (FFR) calculation from CT scans: A rapid workflow using unsupervised learning and computational fluid dynamics.

Fractional flow reserve (FFR) provides the functional relevance of coronary atheroma. The FFR-guided strategy has been shown to reduce unnecessary stenting, improve overall health outcome, and to be cost-saving. The non-invasive, coronary computerised tomography (CT) angiography-derived FFR (cFFR) is an emerging method in reducing invasive catheter based measurements. This computational fluid dynamics-based method is laborious as it requires expertise in multidisciplinary analysis of combining image analysis and computational mechanics. In this work, we present a rapid method, powered by unsupervised learning, to automatically calculate cFFR from CT scans without manual intervention.

Biomechanics of cells and subcellular components: A comprehensive review of computational models and applications.

Cells are a fundamental structural, functional and biological unit for all living organisms. Up till now, considerable efforts have been made to study the responses of single cells and subcellular components to an external load, and understand the biophysics underlying cell rheology, mechanotransduction and cell functions using experimental and in silico approaches. In the last decade, computational simulation has become increasingly attractive due to its critical role in interpreting experimental data, analysing complex cellular/subcellular structures, facilitating diagnostic designs and therapeutic techniques, and developing biomimetic materials. Despite the significant progress, developing comprehensive and accurate models of living cells remains a grand challenge in the 21st century. To understand current state of the art, this review summarises and classifies the vast array of computational biomechanical models for cells. The article covers the cellular components at multi-spatial levels, that is, protein polymers, subcellular components, whole cells and the systems with scale beyond a cell. In addition to the comprehensive review of the topic, this article also provides new insights into the future prospects of developing integrated, active and high-fidelity cell models that are multiscale, multi-physics and multi-disciplinary in nature. This review will be beneficial for the researchers in modelling the biomechanics of subcellular components, cells and multiple cell systems and understanding the cell functions and biological processes from the perspective of cell mechanics.

A proof of concept study for machine learning application to stenosis detection.

This proof of concept (PoC) assesses the ability of machine learning (ML) classifiers to predict the presence of a stenosis in a three vessel arterial system consisting of the abdominal aorta bifurcating into the two common iliacs. A virtual patient database (VPD) is created using one-dimensional pulse wave propagation model of haemodynamics. Four different machine learning (ML) methods are used to train and test a series of classifiers-both binary and multiclass-to distinguish between healthy and unhealthy virtual patients (VPs) using different combinations of pressure and flow-rate measurements. It is found that the ML classifiers achieve specificities larger than 80% and sensitivities ranging from 50 to 75%. The most balanced classifier also achieves an area under the receiver operative characteristic curve of 0.75, outperforming approximately 20 methods used in clinical practice, and thus placing the method as moderately accurate. Other important observations from this study are that (i) few measurements can provide similar classification accuracies compared to the case when more/all the measurements are used; (ii) some measurements are more informative than others for classification; and (iii) a modification of standard methods can result in detection of not only the presence of stenosis, but also the stenosed vessel. Graphical Abstract An overview of methodology fo the creation of virtual patients and their classification.

A physiologically realistic virtual patient database for the study of arterial haemodynamics.

This study creates a physiologically realistic virtual patient database (VPD), representing the human arterial system, for the primary purpose of studying the effects of arterial disease on haemodynamics. A low dimensional representation of an anatomically detailed arterial network is outlined, and a physiologically realistic posterior distribution for its parameters constructed through the use of a Bayesian approach. This approach combines both physiological/geometrical constraints and the available measurements reported in the literature. A key contribution of this work is to present a framework for including all such available information for the creation of virtual patients (VPs). The Markov Chain Monte Carlo (MCMC) method is used to sample random VPs from this posterior distribution, and the pressure and flow-rate profiles associated with each VP computed through a physics based model of pulse wave propagation. This combination of the arterial network parameters (representing a virtual patient) and the haemodynamics waveforms of pressure and flow-rates at various locations (representing functional response and potential measurements that can be acquired in the virtual patient) makes up the VPD. While 75,000 VPs are sampled from the posterior distribution, 10,000 are discarded as the initial burn-in period of the MCMC sampler. A further 12,857 VPs are subsequently removed due to the presence of negative average flow-rate, reducing the VPD to 52,143. Due to undesirable behaviour observed in some VPs-asymmetric under- and over-damped pressure and flow-rate profiles in left and right sides of the arterial system-a filter is proposed to remove VPs showing such behaviour. Post application of the filter, the VPD has 28,868 subjects. It is shown that the methodology is appropriate by comparing the VPD statistics to those reported in literature across real populations. Generally, a good agreement between the two is found while respecting physiological/geometrical constraints.

On the poro-elastic models for microvascular blood flow resistance: An in vitro validation.

Nowadays, adequate and accurate representation of the microvascular flow resistance constitutes one of the major challenges in computational haemodynamic studies. In this work, a theoretical, porous media framework, ultimately designed for representing downstream resistance, is presented and compared against an in vitro experimental results. The resistor consists of a poro-elastic tube, with either a constant or variable porosity profile in space. The underlying physics, characterizing the fluid flow through the porous media, is analysed by considering flow variables at different network locations. Backward reflections, originated in the reservoir of the in vitro model, are accounted for through a reflection coefficient imposed as an outflow network condition. The simulation results are in good agreement with the measurements for both the homogenous and heterogeneous porosity conditions. In addition, the comparison allows identification of the range of values representing experimental reservoir reflection coefficients. The pressure drops across the heterogeneous porous media increases with respect to the simpler configuration, whilst flow remains almost unchanged. The effect of some fluid network features, such as tube Young's modulus and fluid viscosity, on the theoretical results is also elucidated, providing a reference for the invitro and insilico simulation of different microvascular conditions.

Towards enabling a cardiovascular digital twin for human systemic circulation using inverse analysis.

An exponential rise in patient data provides an excellent opportunity to improve the existing health care infrastructure. In the present work, a method to enable cardiovascular digital twin is proposed using inverse analysis. Conventionally, accurate analytical solutions for inverse analysis in linear problems have been proposed and used. However, these methods fail or are not efficient for nonlinear systems, such as blood flow in the cardiovascular system (systemic circulation) that involves high degree of nonlinearity. To address this, a methodology for inverse analysis using recurrent neural network for the cardiovascular system is proposed in this work, using a virtual patient database. Blood pressure waveforms in various vessels of the body are inversely calculated with the help of long short-term memory (LSTM) cells by inputting pressure waveforms from three non-invasively accessible blood vessels (carotid, femoral and brachial arteries). The inverse analysis system built this way is applied to the detection of abdominal aortic aneurysm (AAA) and its severity using neural networks.

Diagnostic performance of virtual fractional flow reserve derived from routine coronary angiography using segmentation free reduced order (1-dimensional) flow modelling.

INTRODUCTION: Fractional flow reserve (FFR) improves assessment of the physiological significance of coronary lesions compared with conventional angiography. However, it is an invasive investigation. We tested the performance of a virtual FFR (1D-vFFR) using routine angiographic images and a rapidly performed reduced order computational model. METHODS: Quantitative coronary angiography (QCA) was performed in 102 with coronary lesions assessed by invasive FFR. A 1D-vFFR for each lesion was created using reduced order (one-dimensional) computational flow modelling derived from conventional angiographic images and patient specific estimates of coronary flow. The diagnostic accuracy of 1D-vFFR and QCA derived stenosis was compared against the gold standard of invasive FFR using area under the receiver operator characteristic curve (AUC). RESULTS: QCA revealed the mean coronary stenosis diameter was 44% ± 12% and lesion length 13 ± 7 mm. Following angiography calculation of the 1DvFFR took less than one minute. Coronary stenosis (QCA) had a significant but weak correlation with FFR (r = -0.2, p = 0.04) and poor diagnostic performance to identify lesions with FFR <0.80 (AUC 0.39, p = 0.09), (sensitivity - 58% and specificity - 26% at a QCA stenosis of 50%). In contrast, 1D-vFFR had a better correlation with FFR (r = 0.32, p = 0.01) and significantly better diagnostic performance (AUC 0.67, p = 0.007), (sensitivity - 92% and specificity - 29% at a 1D-vFFR of 0.7). CONCLUSIONS: 1D-vFFR improves the determination of functionally significant coronary lesions compared with conventional angiography without requiring a pressure-wire or hyperaemia induction. It is fast enough to influence immediate clinical decision-making but requires further clinical evaluation.

Artificial intelligence approaches to predict coronary stenosis severity using non-invasive fractional flow reserve.

Fractional flow reserve is the current reference standard in the assessment of the functional impact of a stenosis in coronary heart disease. In this study, three models of artificial intelligence of varying degrees of complexity were compared to fractional flow reserve measurements. The three models are the multivariate polynomial regression, which is a statistical method used primarily for correlation; the feed-forward neural network; and the long short-term memory, which is a type of recurrent neural network that is suited to modelling sequences. The models were initially trained using a virtual patient database that was generated from a validated one-dimensional physics-based model. The feed-forward neural network performed the best for all test cases considered, which were a single vessel case from a virtual patient database, a multi-vessel network from a virtual patient database, and 25 clinically invasive fractional flow reserve measurements from real patients. The feed-forward neural network model achieved around 99% diagnostic accuracy in both tests involving virtual patients, and a respectable 72% diagnostic accuracy when compared to the invasive fractional flow reserve measurements. The multivariate polynomial regression model performed well in the single vessel case, but struggled on network cases as the variation of input features was much larger. The long short-term memory performed well for the single vessel cases, but tended to have a bias towards a positive fractional flow reserve prediction for the virtual multi-vessel case, and for the patient cases. Overall, the feed-forward neural network shows promise in successfully predicting fractional flow reserve in real patients, and could be a viable option if trained using a large enough data set of real patients.

Computational instantaneous wave-free ratio (IFR) for patient-specific coronary artery stenoses using 1D network models.

In this work, we estimate the diagnostic threshold of the instantaneous wave-free ratio (iFR) through the use of a one-dimensional haemodynamic framework. To this end, we first compared the computed fractional flow reserve (cFFR) predicted from a 1D computational framework with invasive clinical measurements. The framework shows excellent promise and utilises minimal patient data from a cohort of 52 patients with a total of 66 stenoses. The diagnostic accuracy of the cFFR model was 75.76%, with a sensitivity of 71.43%, a specificity of 77.78%, a positive predictive value of 60%, and a negative predictive value of 85.37%. The validated model was then used to estimate the diagnostic threshold of iFR. The model determined a quadratic relationship between cFFR and the ciFR. The iFR diagnostic threshold was determined to be 0.8910 from a receiver operating characteristic curve that is in the range of 0.89 to 0.9 that is normally reported in clinical studies.

Non-invasive coronary CT angiography-derived fractional flow reserve: A benchmark study comparing the diagnostic performance of four different computational methodologies.

Non-invasive coronary computed tomography (CT) angiography-derived fractional flow reserve (cFFR) is an emergent approach to determine the functional relevance of obstructive coronary lesions. Its feasibility and diagnostic performance has been reported in several studies. It is unclear if differences in sensitivity and specificity between these studies are due to study design, population, or "computational methodology." We evaluate the diagnostic performance of four different computational workflows for the prediction of cFFR using a limited data set of 10 patients, three based on reduced-order modelling and one based on a 3D rigid-wall model. The results for three of these methodologies yield similar accuracy of 6.5% to 10.5% mean absolute difference between computed and measured FFR. The main aspects of modelling which affected cFFR estimation were choice of inlet and outlet boundary conditions and estimation of flow distribution in the coronary network. One of the reduced-order models showed the lowest overall deviation from the clinical FFR measurements, indicating that reduced-order models are capable of a similar level of accuracy to a 3D model. In addition, this reduced-order model did not include a lumped pressure-drop model for a stenosis, which implies that the additional effort of isolating a stenosis and inserting a pressure-drop element in the spatial mesh may not be required for FFR estimation. The present benchmark study is the first of this kind, in which we attempt to homogenize the data required to compute FFR using mathematical models. The clinical data utilised in the cFFR workflows are made publicly available online.

A novel numerical modelling approach for keratoplasty eye procedure.

Objective of the work is to investigate stress and deformation that conrneal tissue and donor graft undergo during endothelial keratoplasty. In order to attach the donor graft to the cornea, different air bubble pressure profiles acting on the graft are considered. This study is carried out by employing a three-dimensional nonlinear finite element methodology, combined with a contact algorithm. The ocular tissues are treated as isotropic, hyper-elastic and nearly-incompressible materials. The contact algorithm, based on the penalty-based node-to-surface approach, is used to model the donor graft-corneal interface region. First, the proposed computational methodology is tested against benchmark data for bending of the plates over a cylinder. Then, the influence of geometrical and material parameters of the graft on the corneal contact-structural response is investigated. The results are presented in terms of Von Mises stress intensity, displacement and mean contact force. Results clearly indicate that the air bubble pressure plays a key role in the corneal stress and strain, as well as graft stiffness and thickness.

Electromechanical vibration of microtubules and its application in biosensors.

An electric field (EF) has the potential to excite the vibration of polarized microtubules (MTs) and thus enable their use as a biosensor for the biophysical properties of MTs or cells. To facilitate the development, this paper aims to capture the EF-induced vibration modes and the associated frequency for MTs. The analyses were carried out based on a molecular structural mechanics model accounting for the structural details of MTs. Transverse vibration, radial breathing vibration and axial vibration were achieved for MTs subject to a transverse or an axial EF. The frequency shift and stiffness alteration of MTs were also examined due to the possible changes of the tubulin interactions in physiological or pathological processes. The strong correlation achieved between the tubulin interaction and MT vibration excited by EF provides a new avenue to a non-contacting technique for the structural or property changes in MTs, where frequency shift is used as a biomarker. This technique can be used for individual MTs and is possible for those in cells when the cytosol damping on MT vibrations is largely reduced by the unique features of MT-cytosol interface.

A novel, FFT-based one-dimensional blood flow solution method for arterial network.

In the present work, we propose an FFT-based method for solving blood flow equations in an arterial network with variable properties and geometrical changes. An essential advantage of this approach is in correctly accounting for the vessel skin friction through the use of Womersley solution. To incorporate nonlinear effects, a novel approximation method is proposed to enable calculation of nonlinear corrections. Unlike similar methods available in the literature, the set of algebraic equations required for every harmonic is constructed automatically. The result is a generalized, robust and fast method to accurately capture the increasing pulse wave velocity downstream as well as steepening of the pulse front. The proposed method is shown to be appropriate for incorporating correct convection and diffusion coefficients. We show that the proposed method is fast and accurate and it can be an effective tool for 1D modelling of blood flow in human arterial networks.

A novel porous media-based approach to outflow boundary resistances of 1D arterial blood flow models.

In this paper we introduce a novel method for prescribing terminal boundary conditions in one-dimensional arterial flow networks. This is carried out by coupling the terminal arterial vessel with a poro-elastic tube, representing the flow resistance offered by microcirculation. The performance of the proposed porous media-based model has been investigated through several different numerical examples. First, we investigate model parameters that have a profound influence on the flow and pressure distributions of the system. The simulation results have been compared against the waveforms generated by three elements (RCR) Windkessel model. The proposed model is also integrated into a realistic arterial tree, and the results obtained have been compared against experimental data at different locations of the network. The accuracy and simplicity of the proposed model demonstrates that it can be an excellent alternative for the existing models.

A semi-active human digital twin model for detecting severity of carotid stenoses from head vibration-A coupled computational mechanics and computer vision method.

In this work, we propose a methodology to detect the severity of carotid stenosis from a video of a human face with the help of a coupled blood flow and head vibration model. This semi-active digital twin model is an attempt to link noninvasive video of a patient face to the percentage of carotid occlusion. The pulsatile nature of blood flow through the carotid arteries induces a subtle head vibration. This vibration is a potential indicator of carotid stenosis severity, and it is exploited in the present study. A head vibration model has been proposed in the present work that is linked to the forces generated by blood flow with or without occlusion. The model is used to generate a large number of virtual head vibration data for different degrees of occlusion. In order to determine the in vivo head vibration, a computer vision algorithm is adopted to use human face videos. The in vivo vibrations are compared against the virtual vibration data generated from the coupled computational blood flow/vibration model. A comparison of the in vivo vibration is made against the virtual data to find the best fit between in vivo and virtual data. The preliminary results on healthy subjects and a patient clearly indicate that the model is accurate and it possesses the potential for detecting approximate severity of carotid artery stenoses.

Influence of ageing on human body blood flow and heat transfer: A detailed computational modelling study.

Ageing plays a fundamental role in arterial blood transport and heat transfer within a human body. The aim of this work is to provide a comprehensive methodology, based on biomechanical considerations, for modelling arterial flow and energy exchange mechanisms in the body accounting for age-induced changes. The study outlines a framework for age-related modifications within several interlinked subsystems, which include arterial stiffening, heart contractility variations, tissue volume and property changes, and thermoregulatory system deterioration. Some of the proposed age-dependent governing equations are directly extrapolated from experimental data sets. The computational framework is demonstrated through numerical experiments, which show the impact of such age-related changes on arterial blood pressure, local temperature distribution, and global body thermal response. The proposed numerical experiments show that the age-related changes in arterial convection do not significantly affect the tissue temperature distribution. Results also highlight age-related effects on the sweating mechanism, which lead to a significant reduction in heat dissipation and a subsequent rise in skin and core temperatures.

What does not kill a tumour may make it stronger: In silico insights into chemotherapeutic drug resistance.

Tumour recurrence post chemotherapy is an established clinical problem and many cancer types are often observed to be increasingly drug resistant subsequent to chemotherapy treatments. Drug resistance in cancer is a multipart phenomenon which can be derived from several origins and in many cases it has been observed that cancer cells have the ability to possess, acquire and communicate drug resistant traits. Here, an in silico framework is developed in order to study drug resistance and drug response in cancer cell populations exhibiting various drug resistant features. The framework is based on an on-lattice hybrid multiscale mathematical model and is equipped to simulate multiple mechanisms on different scales that contribute towards chemotherapeutic drug resistance in cancer. This study demonstrates how drug resistant tumour features may depend on the interplay amongst intracellular, extracelluar and intercellular factors. On a cellular level, drug resistant cell phenotypes are here derived from inheritance or mutations that are spontaneous, drug-induced or communicated via exosomes. Furthermore intratumoural heterogeneity and spatio-temporal drug dynamics heavily influences drug delivery and the development of drug resistant cancer cell subpopulations. Chemotherapy treatment strategies are here optimised for various in silico tumour scenarios and treatment objectives. We demonstrate that optimal chemotherapy treatment strategies drastically depend on which drug resistant mechanisms are activated, and that furthermore suboptimal chemotherapy administration may promote drug resistance.