A design specification for Critical Illness Digital Twins (CIDTs) to cure sepsis: responding to the National Academies of Sciences, Engineering and Medicine Report "Foundational Research Gaps and Future Directions for Digital Twins".

On December 15, 2023, The National Academies of Sciences, Engineering and Medicine (NASEM) released a report entitled: "Foundational Research Gaps and Future Directions for Digital Twins." The ostensible purpose of this report was to bring some structure to the burgeoning field of digital twins by providing a working definition and a series of research challenges that need to be addressed to allow this technology to fulfill its full potential. In the work presented herein we focus on five specific findings from the NASEM Report: 1) definition of a Digital Twin, 2) using "fit-for-purpose" guidance, 3) developing novel approaches to Verification, Validation and Uncertainty Quantification (VVUQ) of Digital Twins, 4) incorporating control as an explicit purpose for a Digital Twin and 5) using a Digital Twin to guide data collection and sensor development, and describe how these findings are addressed through the design specifications for a Critical Illness Digital Twin (CIDT) aimed at curing sepsis.

Activation of Cryptochrome 4 from Atlantic Herring.

Marine fish migrate long distances up to hundreds or even thousands of kilometers for various reasons that include seasonal dependencies, feeding, or reproduction. The ability to perceive the geomagnetic field, called magnetoreception, is one of the many mechanisms allowing some fish to navigate reliably in the aquatic realm. While it is believed that the photoreceptor protein cryptochrome 4 (Cry4) is the key component for the radical pair-based magnetoreception mechanism in night migratory songbirds, the Cry4 mechanism in fish is still largely unexplored. The present study aims to investigate properties of the fish Cry4 protein in order to understand the potential involvement in a radical pair-based magnetoreception. Specifically, a computationally reconstructed atomistic model of Cry4 from the Atlantic herring (Clupea harengus) was studied employing classical molecular dynamics (MD) and quantum mechanics/molecular mechanics (QM/MM) methods to investigate internal electron transfers and the radical pair formation. The QM/MM simulations reveal that electron transfers occur similarly to those found experimentally and computationally in Cry4 from European robin (Erithacus rubecula). It is therefore plausible that the investigated Atlantic herring Cry4 has the physical and chemical properties to form radical pairs that in turn could provide fish with a radical pair-based magnetic field compass sensor.

A Combined Magnetoelectric Sensor Array and MRI-Based Human Head Model for Biomagnetic FEM Simulation and Sensor Crosstalk Analysis.

Magnetoelectric (ME) magnetic field sensors are novel sensing devices of great interest in the field of biomagnetic measurements. We investigate the influence of magnetic crosstalk and the linearity of the response of ME sensors in different array and excitation configurations. To achieve this aim, we introduce a combined multiscale 3D finite-element method (FEM) model consisting of an array of 15 ME sensors and an MRI-based human head model with three approximated compartments of biological tissues for skin, skull, and white matter. A linearized material model at the small-signal working point is assumed. We apply homogeneous magnetic fields and perform inhomogeneous magnetic field excitation for the ME sensors by placing an electric point dipole source inside the head. Our findings indicate significant magnetic crosstalk between adjacent sensors leading down to a 15.6% lower magnetic response at a close distance of 5 mm and an increasing sensor response with diminishing crosstalk effects at increasing distances up to 5 cm. The outermost sensors in the array exhibit significantly less crosstalk than the sensors located in the center of the array, and the vertically adjacent sensors exhibit a stronger crosstalk effect than the horizontally adjacent ones. Furthermore, we calculate the ratio between the electric and magnetic sensor responses as the sensitivity value and find near-constant sensitivities for each sensor, confirming a linear relationship despite magnetic crosstalk and the potential to simulate excitation sources and sensor responses independently.

COMMBINI: an experimentally-informed COmputational Model of Macrophage dynamics in the Bone INjury Immunoresponse.

Bone fracture healing is a well-orchestrated but complex process that involves numerous regulations at different scales. This complexity becomes particularly evident during the inflammatory stage, as immune cells invade the healing region and trigger a cascade of signals to promote a favorable regenerative environment. Thus, the emergence of criticalities during this stage might hinder the rest of the process. Therefore, the investigation of the many interactions that regulate the inflammation has a primary importance on the exploration of the overall healing progression. In this context, an in silico model named COMMBINI (COmputational Model of Macrophage dynamics in the Bone INjury Immunoresponse) has been developed to investigate the mechano-biological interactions during the early inflammatory stage at the tissue, cellular and molecular levels. An agent-based model is employed to simulate the behavior of immune cells, inflammatory cytokines and fracture debris as well as their reciprocal multiscale biological interactions during the development of the early inflammation (up to 5 days post-injury). The strength of the computational approach is the capacity of the in silico model to simulate the overall healing process by taking into account the numerous hidden events that contribute to its success. To calibrate the model, we present an in silico immunofluorescence method that enables a direct comparison at the cellular level between the model output and experimental immunofluorescent images. The combination of sensitivity analysis and a Genetic Algorithm allows dynamic cooperation between these techniques, enabling faster identification of the most accurate parameter values, reducing the disparity between computer simulation and histological data. The sensitivity analysis showed a higher sensibility of the computer model to the macrophage recruitment ratio during the early inflammation and to proliferation in the late stage. Furthermore, the Genetic Algorithm highlighted an underestimation of macrophage proliferation by in vitro experiments. Further experiments were conducted using another externally fixated murine model, providing an independent validation dataset. The validated COMMBINI platform serves as a novel tool to deepen the understanding of the intricacies of the early bone regeneration phases. COMMBINI aims to contribute to designing novel treatment strategies in both the biological and mechanical domains.

Data-driven multiscale model of macaque auditory thalamocortical circuits reproduces in vivo dynamics.

We developed a detailed model of macaque auditory thalamocortical circuits, including primary auditory cortex (A1), medial geniculate body (MGB), and thalamic reticular nucleus, utilizing the NEURON simulator and NetPyNE tool. The A1 model simulates a cortical column with over 12,000 neurons and 25 million synapses, incorporating data on cell-type-specific neuron densities, morphology, and connectivity across six cortical layers. It is reciprocally connected to the MGB thalamus, which includes interneurons and core and matrix-layer-specific projections to A1. The model simulates multiscale measures, including physiological firing rates, local field potentials (LFPs), current source densities (CSDs), and electroencephalography (EEG) signals. Laminar CSD patterns, during spontaneous activity and in response to broadband noise stimulus trains, mirror experimental findings. Physiological oscillations emerge spontaneously across frequency bands comparable to those recorded in vivo. We elucidate population-specific contributions to observed oscillation events and relate them to firing and presynaptic input patterns. The model offers a quantitative theoretical framework to integrate and interpret experimental data and predict its underlying cellular and circuit mechanisms.

Origin of Heterogeneous Stripping of Lithium in Liquid Electrolytes.

Lithium metal batteries suffer from low cycle life. During discharge, parts of the lithium are not stripped reversibly and remain isolated from the current collector. This isolated lithium is trapped in the insulating remaining solid-electrolyte interphase (SEI) shell and contributes to the capacity loss. However, a fundamental understanding of why isolated lithium forms and how it can be mitigated is lacking. In this article, we perform a combined theoretical and experimental study to understand isolated lithium formation during stripping. We derive a thermodynamic consistent model of lithium dissolution and find that the interaction between lithium and SEI leads to locally preferred stripping and isolated lithium formation. Based on a cryogenic transmission electron microscopy (cryo TEM) setup, we reveal that these local effects are particularly pronounced at kinks of lithium whiskers. We find that lithium stripping can be heterogeneous both on a nanoscale and on a larger scale. Cryo TEM observations confirm our theoretical prediction that isolated lithium occurs less at higher stripping current densities. The origin of isolated lithium lies in local effects, such as heterogeneous SEI, stress fields, or the geometric shape of the deposits. We conclude that in order to mitigate isolated lithium, a uniform lithium morphology during plating and a homogeneous SEI are indispensable.

Theoretical quantification of the polyvalent binding of nanoparticles coated with peptide-major histocompatibility complex to T cell receptor-nanoclusters.

Nanoparticles (NPs) coated with peptide-major histocompatibility complexes (pMHCs) can be used as a therapy to treat autoimmune diseases. They do so by inducing the differentiation and expansion of disease-suppressing T regulatory type 1 (Tr1) cells by binding to their T cell receptors (TCRs) expressed as TCR-nanoclusters (TCRnc). Their efficacy can be controlled by adjusting NP size and number of pMHCs coated on them (referred to as valence). The binding of these NPs to TCRnc on T cells is thus polyvalent and occurs at three levels: the TCR-pMHC, NP-TCRnc and T cell levels. In this study, we explore how this polyvalent interaction is manifested and examine if it can facilitate T cell activation downstream. This is done by developing a multiscale biophysical model that takes into account the three levels of interactions and the geometrical complexity of the binding. Using the model, we quantify several key parameters associated with this interaction analytically and numerically, including the insertion probability that specifies the number of remaining pMHC binding sites in the contact area between T cells and NPs, the dwell time of interaction between NPs and TCRnc, carrying capacity of TCRnc, the distribution of covered and bound TCRs, and cooperativity in the binding of pMHCs within the contact area. The model was fit to previously published dose-response curves of interferon-γ obtained experimentally by stimulating a population of T cells with increasing concentrations of NPs at various valences and NP sizes. Exploring the parameter space of the model revealed that for an appropriate choice of the contact area angle, the model can produce moderate jumps between dose-response curves at low valences. This suggests that the geometry and kinetics of NP binding to TCRnc can act in synergy to facilitate T cell activation.

A multiscale hybrid model for exploring the effect of Resolvin D1 on macrophage polarization during acute inflammation.

Dysregulated inflammation underlies various diseases. Specialized pro-resolving mediators (SPMs) like Resolvin D1 (RvD1) have been shown to resolve inflammation and halt disease progression. Macrophages, key immune cells that drive inflammation, respond to the presence of RvD1 by polarizing to an anti-inflammatory type (M2). However, RvD1's mechanisms, roles, and utility are not fully understood. This paper introduces a gene-regulatory network (GRN) model that contains pathways for RvD1 and other SPMs and proinflammatory molecules like lipopolysaccharides. We couple this GRN model to a partial differential equation-agent-based hybrid model using a multiscale framework to simulate an acute inflammatory response with and without the presence of RvD1. We calibrate and validate the model using experimental data from two animal models. The model reproduces the dynamics of key immune components and the effects of RvD1 during acute inflammation. Our results suggest RvD1 can drive macrophage polarization through the G protein-coupled receptor 32 (GRP32) pathway. The presence of RvD1 leads to an earlier and increased M2 polarization, reduced neutrophil recruitment, and faster apoptotic neutrophil clearance. These results support a body of literature that suggests that RvD1 is a promising candidate for promoting the resolution of acute inflammation. We conclude that once calibrated and validated on human data, the model can identify critical sources of uncertainty, which could be further elucidated in biological experiments and assessed for clinical use.

RootSlice-A novel functional-structural model for root anatomical phenotypes.

Root anatomy is an important determinant of root metabolic costs, soil exploration, and soil resource capture. Root anatomy varies substantially within and among plant species. RootSlice is a multicellular functional-structural model of root anatomy developed to facilitate the analysis and understanding of root anatomical phenotypes. RootSlice can capture phenotypically accurate root anatomy in three dimensions of different root classes and developmental zones, of both monocotyledonous and dicotyledonous species. Several case studies are presented illustrating the capabilities of the model. For maize nodal roots, the model illustrated the role of vacuole expansion in cell elongation; and confirmed the individual and synergistic role of increasing root cortical aerenchyma and reducing the number of cortical cell files in reducing root metabolic costs. Integration of RootSlice for different root zones as the temporal properties of the nodal roots in the whole-plant and soil model OpenSimRoot/maize enabled the multiscale evaluation of root anatomical phenotypes, highlighting the role of aerenchyma formation in enhancing the utility of cortical cell files for improving plant performance over varying soil nitrogen supply. Such integrative in silico approaches present avenues for exploring the fitness landscape of root anatomical phenotypes.

A Computational Study of Blood Flow Dynamics in the Pulmonary Arteries.

In this work we study the blood dynamics in the pulmonary arteries by means of a 3D-0D geometric multiscale approach, where a detailed 3D model for the pulmonary arteries is coupled with a lumped parameters (0D) model of the cardiovascular system. We propose to investigate three strategies for the numerical solution of the 3D-0D coupled problem: the Splitting-Explicit and Implicit algorithms, where information are exchanged between 3D and 0D models at each time step at the interfaces, and the One-Way algorithm, where the 0D is solved first off-line. In our numerical experiments performed in a realistic patient-specific 3D domain with a physiologically calibrated 0D model, we discuss first the issue on instabilities that may arise when not suitable connections are considered between 3D and 0D models; second we compare the performance and accuracy of the three proposed numerical strategies. Finally, we report a comparison between a healthy and a hypertensive case, providing a preliminary result highlighting how our method could be used in future for clinical purposes.

Virtual cells in a virtual microenvironment recapitulate early development-like patterns in human pluripotent stem cell colonies.

The mechanism by which morphogenetic signals engage the regulatory networks responsible for early embryonic tissue patterning is incompletely understood. Here, we developed a minimal gene regulatory network (GRN) model of human pluripotent stem cell (hPSC) lineage commitment and embedded it into "cellular" agents that respond to a dynamic morphogenetic signaling microenvironment. Simulations demonstrated that GRN wiring had significant non-intuitive effects on tissue pattern order, composition, and dynamics. Experimental perturbation of GRN connectivities supported model predictions and demonstrated the role of OCT4 as a master regulator of peri-gastrulation fates. Our so-called GARMEN strategy provides a multiscale computational platform to understand how single-cell-based regulatory interactions scale to tissue domains. This foundation provides new opportunities to simulate the impact of network motifs on normal and aberrant tissue development.

An Investigation of Left Ventricular Valve Disorders and the Mechano-Electric Feedback Using a Synergistic Lumped Parameter Cardiovascular Numerical Model.

Cardiac diseases and failure make up one of largest contributions to global mortality and significantly detriment the quality of life for millions of others. Disorders in the valves of the left ventricle are a prominent example of heart disease, with prolapse, regurgitation, and stenoses-the three main valve disorders. It is widely known that mitral valve prolapse increases the susceptibility to cardiac arrhythmia. Here, we investigate stenoses and regurgitation of the mitral and aortic valves in the left ventricle using a synergistic low-order numerical model. The model synergy derives from the incorporation of the mechanical, chemical, and electrical elements. As an alternative framework to the time-varying elastance (TVE) method, it allows feedback mechanisms at work in the heart to be considered. The TVE model imposes the ventricular pressure-volume relationship using a periodic function rather than calculating it consistently. Using our synergistic approach, the effects of valve disorders on the mechano-electric-feedback (MEF) are investigated. The MEF is the influence of cellular mechanics on the electrical activity, and significantly contributes to the generation of arrhythmia. We further investigate stenoses and regurgitation of the mitral and aortic valves and their relationship with the MEF and generation of arrhythmia. Mitral valve stenosis is found to increase the sensitivity to arrhythmia-stimulating systolic stretch, and reduces the sensitivity to diastolic stretch. Aortic valve stenosis does not change the sensitivity to arrhythmia-stimulating stretch, and regurgitation reduces it. A key result is found when valve regurgitation is accompanied by diastolic stretch. In the presence of MEF disorder, ectopic beats become far more frequent when accompanied by valve regurgitation. Therefore, arrhythmia resulting from a disorder in the MEF will be more severe when valve regurgitation is present.

Assessing the therapeutic response of tumors to hypoxia-targeted prodrugs with an in silico approach.

Tumor hypoxia is commonly recognized as a condition stimulating the progress of the aggressive phenotype of tumor cells. Hypoxic tumor cells inhibit the delivery of cytotoxic drugs, causing hypoxic areas to receive insufficient amounts of anticancer agents, which results in adverse treatment responses. Being such an obstruction to conventional therapies for cancer, hypoxia might be considered a target to facilitate the efficacy of treatments in the resistive environment of tumor sites. In this regard, benefiting from prodrugs that selectively target hypoxic regions remains an effective approach. Additionally, combining hypoxia-activated prodrugs (HAPs) with conventional chemotherapeutic drugs has been used as a promising strategy to eradicate hypoxic cells. However, determining the appropriate sequencing and scheduling of the combination therapy is also of great importance in obtaining favorable results in anticancer therapy. Here, benefiting from a modeling approach, we study the efficacy of HAPs in combination with chemotherapeutic drugs on tumor growth and the treatment response. Different treatment schedules have been investigated to see the importance of determining the optimal schedule in combination therapy. The effectiveness of HAPs in varying hypoxic conditions has also been explored in the study. The model provides qualitative conclusions about the treatment response, as the maximal benefit is obtained from combination therapy with greater cell death for highly hypoxic tumors. It has also been observed that the antitumor effects of HAPs show a hypoxia-dependent profile.

A Multiscale Model of COVID-19 Dynamics.

COVID-19, caused by the infection of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has been a global pandemic and created unprecedented public health challenges throughout the world. Despite significant progresses in understanding the disease pathogenesis and progression, the epidemiological triad of pathogen, host, and environment remains unclear. In this paper, we develop a multiscale model to study the coupled within-host and between-host dynamics of COVID-19. The model includes multiple transmission routes (both human-to-human and environment-to-human) and connects multiple scales (both the population and individual levels). A detailed analysis on the local and global dynamics of the fast system, slow system and full system shows that rich dynamics, including both forward and backward bifurcations, emerge with the coupling of viral infection and epidemiological models. Model fitting to both virological and epidemiological data facilitates the evaluation of the influence of a few infection characteristics and antiviral treatment on the spread of the disease. Our work underlines the potential role that the environment can play in the transmission of COVID-19. Antiviral treatment of infected individuals can delay but cannot prevent the emergence of disease outbreaks. These results highlight the implementation of comprehensive intervention measures such as social distancing and wearing masks that aim to stop airborne transmission, combined with surface disinfection and hand hygiene that can prevent environmental transmission. The model also provides a multiscale modeling framework to study other infectious diseases when the environment can serve as a reservoir of pathogens.

A Multiscale Mathematical Model of Plasmodium Vivax Transmission.

Malaria is caused by Plasmodium parasites which are transmitted to humans by the bite of an infected Anopheles mosquito. Plasmodium vivax is distinct from other malaria species in its ability to remain dormant in the liver (as hypnozoites) and activate later to cause further infections (referred to as relapses). Mathematical models to describe the transmission dynamics of P. vivax have been developed, but most of them fail to capture realistic dynamics of hypnozoites. Models that do capture the complexity tend to involve many governing equations, making them difficult to extend to incorporate other important factors for P. vivax, such as treatment status, age and pregnancy. In this paper, we have developed a multiscale model (a system of integro-differential equations) that involves a minimal set of equations at the population scale, with an embedded within-host model that can capture the dynamics of the hypnozoite reservoir. In this way, we can gain key insights into dynamics of P. vivax transmission with a minimum number of equations at the population scale, making this framework readily scalable to incorporate more complexity. We performed a sensitivity analysis of our multiscale model over key parameters and found that prevalence of P. vivax blood-stage infection increases with both bite rate and number of mosquitoes but decreases with hypnozoite death rate. Since our mathematical model captures the complex dynamics of P. vivax and the hypnozoite reservoir, it has the potential to become a key tool to inform elimination strategies for P. vivax.

A theoretical analysis of the scale separation in a model to predict solid tumour growth.

Solid tumour growth depends on a host of factors which affect the cell life cycle and extracellular matrix vascularization that leads to a favourable environment. The whole solid tumour can either grow or wither in response to the action of the immune system and therapeutics. A personalised mathematical model of such behaviour must consider both the intra- and inter-cellular dynamics and the mechanics of the solid tumour and its microenvironment. However, such wide range of spatial and temporal scales can hardly be modelled in a single model, and require the so-called multiscale models, defined as orchestrations of single-scale component models, connected by relation models that transform the data for one scale to another. While multiscale models are becoming common, there is a well-established engineering approach to the definition of the scale separation, e.g., how the spatiotemporal continuum is split in the various component models. In most studies scale separation is defined as natural, linked to anatomical concepts such as organ, tissue, or cell; but these do not provide reliable definition of scales: for examples skeletal organs can be as large as 500 mm (femur), or as small as 3 mm (stapes). Here we apply a recently proposed scale-separation approach based on the actual experimental and computational limitations to a patient-specific model of the growth of neuroblastoma. The resulting multiscale model can be properly informed with the available experimental data and solved in a reasonable timeframe with the available computational resources.

Effect of Interface Transition Zone and Coarse Aggregate on Microscopic Diffusion Behavior of Chloride Ion.

Concrete is a multiphase composite material composed of coarse aggregate, cement mortar, and interface transition zone (ITZ). It is of great significance to study the effect of ITZ and coarse aggregate on chloride microscopic diffusion behavior for predicting the service life of reinforced concrete (RC) structures. By introducing the random distribution function, a random coarse aggregate model considering the randomness of the thickness of the ITZ was established. Furthermore, a two-dimensional (2D) chloride ion diffusion mesoscopic model was developed by specifying different diffusion properties for different phase materials of concrete. Moreover, the effects of coarse aggregate rate, ITZ thickness, and ITZ diffusion property on chloride ion diffusion behavior were investigated in this paper. The research showed that the aggregate has hindrance and agglomeration action on chloride ion diffusion. Although the volume content of the ITZ was very small, less than 0.2% of the total volume of concrete, the effect of the ITZ on the chloride diffusion in concrete cannot be ignored. More importantly, the mechanism of promoting chloride diffusion in the ITZ was revealed through the chloride diffusion trajectory. The research revealed the transmission mechanism of chloride ions in the meso-structure of concrete and provides theoretical support for the design of RC structures in coastal areas.

Prediction of Ultrasonic Pulse Velocity for Cement, Mortar, and Concrete through a Multiscale Homogenization Approach.

Ultrasonic testing (UT) is an important method for concrete, and ultrasonic pulse velocity is commonly used to evaluate the quality of concrete materials in existing studies. The ultrasonic pulse velocity of concrete materials is affected by many factors; therefore, it is necessary to establish a quantitative prediction model for the ultrasonic pulse velocity of concrete materials. Based on the multiscale homogenization method, concrete material is divided into different scales of homogenized materials, namely cement paste, mortar, and concrete. Then, a multiscale ultrasonic pulse velocity model is established through a combination of elasticity formulation and the hydration model. At the three scales of cement paste, mortar, and concrete, the elastic parameters and ultrasonic pulse velocity were predicted with the water-to-cement ratio of 0.35, 0.5, and 0.65, respectively. The ultrasonic pulse velocity of concrete with different water-to-cement ratios and different ages were measured in the test and predicted by the model. The results show that the predicted value of ultrasonic pulse velocity is within the error range of ±1.5% of the measured ultrasonic pulse velocity, suggesting that the established prediction model of ultrasonic pulse velocity can reliably predict the velocity change in concrete materials.

Multiscale Model of Antiviral Timing, Potency, and Heterogeneity Effects on an Epithelial Tissue Patch Infected by SARS-CoV-2.

We extend our established agent-based multiscale computational model of infection of lung tissue by SARS-CoV-2 to include pharmacokinetic and pharmacodynamic models of remdesivir. We model remdesivir treatment for COVID-19; however, our methods are general to other viral infections and antiviral therapies. We investigate the effects of drug potency, drug dosing frequency, treatment initiation delay, antiviral half-life, and variability in cellular uptake and metabolism of remdesivir and its active metabolite on treatment outcomes in a simulated patch of infected epithelial tissue. Non-spatial deterministic population models which treat all cells of a given class as identical can clarify how treatment dosage and timing influence treatment efficacy. However, they do not reveal how cell-to-cell variability affects treatment outcomes. Our simulations suggest that for a given treatment regime, including cell-to-cell variation in drug uptake, permeability and metabolism increase the likelihood of uncontrolled infection as the cells with the lowest internal levels of antiviral act as super-spreaders within the tissue. The model predicts substantial variability in infection outcomes between similar tissue patches for different treatment options. In models with cellular metabolic variability, antiviral doses have to be increased significantly (>50% depending on simulation parameters) to achieve the same treatment results as with the homogeneous cellular metabolism.

Personalised 3D Assessment of Trochanteric Soft Tissues Improves HIP Fracture Classification Accuracy.

Passive soft tissues surrounding the trochanteric region attenuate fall impact forces and thereby control hip fracture risk. The degree of attenuation is related to Soft Tissue Thickness (STT). STT at the neutral hip impact orientation, estimated using a regression relation in body mass index (BMI), was previously shown to influence the current absolute risk of hip fracture (ARF0) and its fracture classification accuracy. The present study investigates whether fracture classification using ARF0 improves when STT is determined from the subject's Computed-Tomography (CT) scans (i.e. personalised) in an orientation-specific (i.e. 3D) manner. STT is calculated as the shortest distance along any impact orientation between a semi-automatically segmented femur surface and an automatically segmented soft tissue/air boundary. For any subject, STT along any of the 33 impact orientations analysed always exceeds the value estimated using BMI. Accuracy of fracture classification using ARF0 improves when using personalised 3D STT estimates (AUC = 0.87) instead of the BMI-based STT estimate (AUC = 0.85). The improvement is smaller (AUC = 0.86) when orientation-specificity of CT-based STT is suppressed and is nil when personalisation is suppressed instead. Thus, fracture classification using ARF0 improves when CT is used to personalise STT estimates and improves further when, in addition, the estimates are orientation specific.