Computational Biomechanics for Medicine

Childbirth-related trauma is a recurrent and widespread topic due to the disorders it can trigger, such as urinary and/or anal incontinence, and pelvic organ prolapse, affecting women at various levels. Pelvic floor dysfunction often results from weakening or direct damage to the pelvic floor muscles (PFM) or connective tissue, and vaginal delivery is considered the primary risk factor. Elucidating the normal labor mechanisms and the impact of vaginal delivery in PFM can lead to the development of preventive and therapeutic strategies to minimize the most common injuries. By providing some understanding of the function of the pelvic floor during childbirth, the existing biomechanical models attempt to respond to this problem. These models have been used to estimate the mechanical changes on PFM during delivery, to analyze fetal descent, the effect of the fetal head molding, and delivery techniques that potentially contribute to facilitating labor and reducing the risk of muscle injury.

Biomechanical models of childbirth should be sufficiently well-informed and functional for personalized planning of birth and obstetric interventions. Some challenges to be addressed with a focus on customization will be discussed including the in vivo acquisition of individual-specific pelvic floor mechanical properties.

Numerical simulation is useful to help understand the behavior of pelvic system, and eventually to assist the diagnostic and surgery. Patient-specific simulation is expected to optimize the treatment of patients. Despite the requirement of mechanical properties and loading, patient-specific simulation requires first 3D geometry adapted to patient. Manual 3D reconstruction of the patient-specific anatomy is time-consuming and introduces uncertainties. In this paper, we propose an efficient computer-assisted approach to modeling 3D geometries well suited to MRI data. A well-controlled physical model is also proposed, and manufactured, to estimate uncertainties of the presented method.

This work presents the results of a numerical analysis of the pelvic ring model subjected to multidirectional impact impulse loading. The action mechanism behind widespread pelvic fractures resulting from combined lateral, vertical, and longitudinal forces has not been sufficiently explained in the literature yet. The elaborated finite element model of the human pelvis based on CT scans contains a bi-layered structure of bone, varying stiffness of pelvic ligaments and hyperelastic behavior of cartilage. The numerical analysis was performed using a force value of 10 kN, equivalent to a velocity of 12 m/s acting in the range ±45° in each direction. The performed analysis indicates the von Mises stress is concentrated in the femur under lateral–vertical impact load, in the frontal part of the pelvic ring under vertical–lateral, and in the wings of the ilium under lateral–longitudinal impact load. The most extensive injuries to the pelvic ring were observed under vertical–longitudinal impact load, causing interruption of the pelvic continuity.

Head impact can cause traumatic brain injury (TBI) through axonal overstretch or subsequent inflammation and understanding the biomechanics of the impact event is useful for TBI prevention research. Tagged magnetic resonance imaging (MRI) acquired during a mild-acceleration impact has enabled measurement and visualization of brain deformation in vivo. However, measurements using MRI are subject to error, and having independent validation while imaging in vivo is very difficult. Thus, characterizing the accuracy of these measurements needs to be done in a separate experiment using a phantom where a gold standard is available. This study describes a method for error quantification using a calibration phantom compatible with MRI and high-speed video (the gold standard). During linear acceleration, the maximum shear strain (MSS) in the phantom ranged from 0 to 12%, which is similar to in vivo brain deformation at a similar acceleration. The mean displacement error against video was 0.3 ± 0.3 mm, and the MSS error was 1.4 ± 0.3%. To match resolutions, video data was filtered temporally using an averaging filter. Compared to the unfiltered results, resolution matching improved the agreement between MRI and video results by 15%. In conclusion, tagged MRI analysis compares well to video data provided that resolutions are matched—a finding that is also applicable when using MRI to validate simulations.

Needle insertion (placement) into human body organs is a frequently performed procedure in clinical practice. Its success largely depends on the accuracy with which the needle tip reaches the anatomical target. As the tissue deforms due to interactions with the needle, the target tends to change its position. One possible way to decrease the risk of missing the target can be to account for tissue deformations when planning the needle insertion. This can be achieved by employing computational biomechanics models to predict the tissue deformations.

In this study, for computing the tissue deformations due to needle insertion, we employed a meshless formulation of computational mechanics that uses a spatial discretisation in a form of a cloud of points. We used the previously verified Meshless Total Lagrangian Explicit Dynamics (MTLED) algorithm that facilitates accurate and robust prediction of soft continua/soft tissues mechanical responses under large deformations. For modelling of interactions between the needle and soft tissues, we propose a kinematic approach that directly links deformation of the tissue adjacent to the needle with the needle motion. This approach does not require any assumptions about the exact mechanisms of such interactions. Its parameters can be determined directly from observation of the tissue sample/body organ deformations during needle insertion.

We evaluated the performance of our kinematic approach for modelling the interactions between the needle and the tissue through application in modelling of needle insertion into a cylindrical sample (diameter of 30 mm and height of 17 mm) of Sylgard 527 (by Dow Corning) silicone gel and comparing the results obtained from the model with the experimentally measured force acting on the needle. The needle insertion depth was up to 10 mm. For modelling of the constitutive responses of Sylgard 527 gel, we used the neo-Hookean hyperelastic material model with the shear modulus experimentally determined from compression of Sylgard 527 gel samples.

The general behaviour of the needle force–insertion depth relationship was correctly predicted by our framework that combines a meshless method of computational mechanics for computing the deformations and kinematic approach for modelling the interactions between the needle and soft tissues. The predicted force magnitude differed by only 25% from the experimentally observed value. These differences require further analysis. One possible explanation can be that the neo-Hookean material model we used here may not be sufficient to correctly represent Sylgard 527 gel constitutive behaviour.

Osteonecrosis occurs due to the loss of blood supply to the bone, leading to spontaneous death of the trabecular bone. Delayed treatment of the involved patients results in collapse of the femoral head, which leads to a need for total hip arthroplasty surgery. Core decompression, as the most popular technique for treatment of the osteonecrosis, includes removal of the lesion area by drilling a straight tunnel to the lesion, debriding the dead bone and replacing it with bone substitutes. However, there are two drawbacks for this treatment method. First, due to the rigidity of the instruments currently used during core decompression, lesions cannot be completely removed and/or excessive healthy bone may also be removed with the lesion. Second, the use of bone substitutes, despite its biocompatibility and osteoconductivity, may not provide sufficient mechanical strength and support for the bone. To address these shortcomings, a novel robot-assisted curved core decompression (CCD) technique is introduced to provide surgeons with direct access to the lesions causing minimal damage to the healthy bone. In this study, with the aid of finite element (FE) simulations, we investigate biomechanical performance of core decompression using the curved drilling technique in the presence of normal gait loading. In this regard, we compare the result of the CCD using bone substitutes and flexible implants with other conventional core decompression techniques. The study finding shows that the maximum principal stress occurring at the superior domain of the neck is smaller in the CCD techniques (i.e., 52.847 MPa) compared to the other core decompression methods; furthermore, the peak value of normal stress at the interface for the CCD model is substantially smaller than traditional and advanced core decompression techniques (89% and 76%, respectively). FE results demonstrate the superior performance of CCD compared to the other approaches without any compromise to patient’s safety and markedly reducing the risk of femoral fracture in the postoperative phase for normal gait loading.

In this paper we present a hybrid 0D–1D model for the cerebral circulation and blood flow in large cerebral arteries. The 0D model contains the electrical analog circuit running from the aorta to the circle of Willis (CoW), and the venous network from the superior sagittal sinus (SSS) to the superior vena cava (SVC). To simulate the cerebral autoregulation, the vascular bed between the arterial and venous networks is implemented using an inductor/resistor couple. An artificial pulsatile pressure waveform includes the normal (∼100 mmHg), hypotensive (∼50 mmHg) and hypertensive (∼150 mmHg) phases. A 1D model is used to numerically solve the 1D Navier–Stokes equations coupled with an empirical arterial wall equation. The 1D model is then applied to the internal carotid, middle and anterior cerebral arteries (ICA, MCA and ACA) in the CoW, with the simulation results from the 0D model as boundary conditions. With this hybrid 0D–1D approach, we show that: (a) the cerebral flow may regain a normal flow rate value (∼600 mL/min) within several cardiac cycles; (b) an incomplete CoW can substantially affect the flow distribution in CoW; and (c) the flow rates in the MCA, ACA and PCA alter in response to the cerebral regulation. In conclusion a hybrid 0D–1D model for the cerebral blood flow is proposed, which can potentially be used for the cerebral flow modelling of different age groups or under different vascular diseases.

Non-invasive methods for estimating carotid artery (CA) pressure waveforms have been recently developed as a tool to detect heart abnormalities. Among non-invasive techniques, camera-based methods have the advantage of being non-contact, which enables measurements without applying external pressures to the artery. Camera-based methods measure skin deformation waveforms caused by arterial blood flow, which are assumed to have a similar shape to the pressure waveforms. Video recordings of the subject’s neck are analysed to quantify skin deformations caused by the carotid pressure pulse. However, in practice, unrelated motion, such as the relative movements of the camera and the subject, or movements due to breathing, can confound the skin measurements. One of the primary effects of this error is seen in the form of signal drift, which can make it difficult to analyse the shape of the skin displacement waveform. In this paper, we have investigated and compared two methods for removing the signal drift. One is to correct for the motion in the captured videos of the neck, and the second method is to use a high-pass wavelet filter. The results showed that, although both methods could reduce the signal drift, they had dissimilar effects on the shape of the CA displacement waveforms. The high-pass wavelet filter seemed to preserve the original measured shape of the CA displacement waveforms better than the motion-correction method. However, in this study, it was not possible to quantify the performance since the true shape of the CA displacement waveforms was not known.

In this study, we simulate blood flow in complex geometries obtained by digital subtraction angiography (DSA) images. We represent the flow domain by a set of irregularly distributed nodes or uniform Cartesian embedded grid, and we numerically solve the non-stationary Navier–Stokes (N-S) equations, in their velocity–vorticity formulation, by using a meshless point collocation method. The spatial derivatives are computed with the discretization corrected particle strength exchange (DC PSE) method, a recently developed meshless interpolation method. For the transient term a fourth order Runge–Kutta time integration scheme is used.

Motor control is a set of time-varying muscle excitations which generate desired motions for a biomechanical system. Muscle excitations cannot be directly measured from live subjects. An alternative approach is to estimate muscle activations using inverse motion-driven simulation. In this article, we propose a deep reinforcement learning method to estimate the muscle excitations in simulated biomechanical systems. Here, we introduce a custom-made reward function which incentivizes faster point-to-point tracking of target motion. Moreover, we deploy two new techniques, namely episode-based hard update and dual buffer experience replay, to avoid feedback training loops. The proposed method is tested in four simulated 2D and 3D environments with 6–24 axial muscles. The results show that the models were able to learn muscle excitations for given motions after nearly 100,000 simulated steps. Moreover, the root mean square error in point-to-point reaching of the target across experiments was less than 1% of the length of the domain of motion. Our reinforcement learning method is far from the conventional dynamic approaches as the muscle control is derived functionally by a set of distributed neurons. This can open paths for neural activity interpretation of this phenomenon.