MICCAI Computational Biomechanics for Medicine XVI workshop programme 2021 October 01
V 09:00 UTC to 10:30 UTC Sessions
> Chair: Martyn Nash
V 09:00 UTC Welcome
* Poul Nielsen
V 09:10 UTC Podium 1
*
Efficient and accurate computation of quantitative flow ratio (QFR) for physiological assessment of coronary artery stenosis from a single angiographic view
V George C. Bourantas, Grigorios Tsigkas, Konstantinos Katsanos, Fivos V. Bekiris, Benjamin F. Zwick, Adam Wittek, Karol Miller, Periklis Davlouros
V In this study we develop a fast (less than 30 sec) method to compute the quantitative flow ratio (QFR) from angiographic images. We evaluate its diagnostic accuracy in patients with intermediate coronary stenosis through comparison with invasively measured fractional flow reserve (FFR). Our method uses a single angiographic view with minimal vessel foreshortening and overlap to compute geometrical data (vessel length, stenosis length, diameter of normal unobstructed vessel, diameter of stenosed vessel). Finally, we evaluate the diagnostic performance of our 2D-based QFR in the physiological assessment of intermediate coronary lesions.
> speaker: George Bourantas
* institution: University of Western Australia
* country: Australia
* city: Perth
* local time: 17:10 AWST
* 09:30 UTC Podium 2
*
Assessing fibre reorientation in soft tissues with simultaneous mueller matrix imaging and mechanical testing
V Alexander Dixon, Andrew Taberner, Martyn Nash, Poul M. F. Nielsen
V The implantation of bioprosthetic heart valves constructed from heterograft soft tissue membranes, such as bovine pericardium, porcine pericardium, and porcine heart valves, is a common method of treatment for valvular heart disease. A greater understanding of the load-dependent behaviour is needed to improve durability and functional performance of these tissues in such applications. However, existing techniques that provide the required experimental data are destructive in nature. This limits measurements to pre- and post- mechanical testing states and prevents further use of the tissue in medical devices. Thus, there is a need to develop an instrument implementing wide-field non-destructive imaging techniques that can assess collagen fibre architecture in soft tissue membranes undergoing mechanical testing. In this work, a novel optomechanical system was developed that integrates a Mueller matrix imaging polarimeter with a biaxial mechanical testing instrument, suitable for characterising the mechanical material properties of soft tissue membranes. A preliminary mechanical test demonstrated that this instrument provides a non-destructive method for detecting fibre realignment along loading directions in these tissues. The additional measurements now possible with this system may offer new insights into structure-function relationships of soft tissue membranes.
> speaker: Alexander Dixon
* institution: University of Auckland
* country: New Zealand
* city: Auckland
* local time: 22:30 NZDT
V 09:50 UTC Podium 3
*
Predicting plaque progression in patient-specific carotid bifurcation
V Tijana Djukic, Smiljana Djorovic, Branko Arsic, Branko Gakovic, Igor Koncar, N. Filipovic
V Atherosclerosis is an inflammatory disease that is characterized by the accumulation of lipids and formation of plaque within the arterial wall. It would be clinically useful to analyze this inflammation process and its progress in more detail for each particular patient. In this study, fully coupled biomechanics software for plaque progression was used to simulate the growth of plaque within patient-specific carotid bifurcation. The numerical model couples computational fluid mechanics, transport of relevant molecules, inflammatory process and plaque growth. The simulation is performed using the geometry of the specific patient that is reconstructed by applying deep learning techniques to the images obtained from clinical ultrasound examination. The numerical model used can help to predict the evolution of atherosclerotic plaque which is very significant for appropriate diagnostics and treatment planning and represents one step further in applying biomechanics modeling within the concept of computer-integrated medicine.
> speaker: Smiljana Tomašević
* institution: University of Kragujevac
* country: Serbia
* city: Kragujevac
* local time: 11:50 CEST
V 10:10 UTC Podium 4
*
Measuring three-dimensional surface deformations of skin using a stereoscopic system and intrinsic features
V Amir HajiRassouliha, Debbie Zhao, Emily J. Lam Po Tang, Dong Hoon Choi, Andrew J. Taberner, Martyn P. Nash, Poul M. F. Nielsen
V Measurement of three-dimensional (3D) surface deformations is an important step in characterising mechanical properties and developing computational models of skin. Compared to two-dimensional (2D) measurements, the ability to acquire 3D information involves extra levels of complexity, particularly during calibration and reconstruction. Furthermore, the addition of speckle patterns to the skin is typically required to enhance surface contrast. We have developed a method that uses a calibrated four-camera stereoscope to accurately measure 3D surface deformations of skin without requiring the addition of extrinsic surface features. To validate the method, a flat, rigid disc was reconstructed in 3D, at different orientations, and subsequently compared against its known dimensions. The root mean squared error of our method in measuring geometric features on the disc was 44 μm ± 25 μm over a 100 mm × 100 mm field of view. Our method could accurately measure disc displacements, with a maximum error of 7.8 μm (relative error 0.0031) at an applied translation of 2500 μm. Deformations of a uniaxially stretched rubber membrane were also measured, showing close agreement with the expected values, assuming a homogenous and linear stress/strain response. Finally, we demonstrated 3D deformation measurement of unpatterned post-mortem pig skin subject to uniaxial stretch using four cameras oriented as the edges of an octahedron with 90º angles between their optical axes. The ability to measure 3D full-field deformations of unpatterned skin, at wide camera angles, enables our method to be used in various skin experiments, including tissue indentation and dissection.
> speaker: Amir HajiRassouliha
* institution: University of Auckland
* country: New Zealand
* city: Auckland
* local time: 23:10 NZDT
* 10:30 UTC to 11:20 UTC Break (50 minute)
V 11:20 UTC to 12:40 UTC Sessions
> Chair: Adam Wittek
V 11:20 UTC Podium 5
*
A direct geometry processing cartilage generation method for datasets with poor cartilage visibility
V Faezeh Moshfeghifar, Max Kragballle Nielsen, José D. Tascón-Vidarte, Sune Darkner, Kenny Erleben
V We present a method to generate subject-specific cartilage for the hip joint. Given bone geometry, our approach is agnostic to image modality, creates conforming interfaces, and is well suited for finite element analysis. We demonstrate our method on ten hip joints showing anatomical shape consistency and well-behaved stress patterns. Our method is fast and may assist in large-scale biomechanical population studies of the hip joint when manual segmentation or training data is not feasible.
> speaker: Faezeh Moshfeghifar
* institution: University of Copenhagen
* country: Denmark
* city: Copenhagen
* local time: 13:20 CEST
V 11:40 UTC Podium 6
*
Development of an open source, low-cost imaging system for continuous lung monitoring in ICU
V Samuel Richardson, Andrew Creegan, Alex Dixon, Llewellyn Sim Johns, Haribalan Kumar, Kelly Burrowes, Poul M.F. Nielsen, J Geoffrey Chase, Merryn H. Tawhai
V A lost-cost open-source electrical impedance tomography (EIT) device was equipped with a novel lidar based workflow to extract torso and electrode position which was then used in the EIT image reconstruction. EIT data was gathered from 9 healthy volunteers (5 male, 4 female) whilst undergoing a controlled breathing protocol. Four different reconstruction configurations were undertaken: a subject specific lidar based mesh vs a generic oval mesh, and subject specific lidar based electrode placements vs generic equal spaced electrode placements. Our results showed that torso shape error and electrode position errors can be drastically reduced with the lidar-based method allowing for the future utilization of patient-specific information. Good correlation was observed between volume delta and the EIT difference image.
> speaker: Samuel Richardson
* institution: University of Auckland
* country: New Zealand
* city: Auckland
* local time: 00:40 NZDT (2021-10-02)
V 12:00 UTC Podium 7
*
Towards accurate measurement of abdominal aortic aneurysm wall thickness from CT and MRI
V Andy T. Huynh, K. Miller
V Abdominal Aortic Aneurysm (AAA) is the focal dilation or widening of the infrarenal artery. It is a vascular disease commonly found in older adults with prevalence increasing steadily with age. The disease is often discovered by unrelated medical examinations and screenings due to its asymptomatic nature. People unaware of their condition may only find out after the catastrophic event of a ruptured AAA, where most patients will not survive if left untreated. The current clinical rupture risk indicator for AAA repair is a AAA diameter exceeding 5.5 cm. There are many limitations with the clinical rupture risk indicator due to its derivation coming from population statistics and not patient-specific circumstances. Computation of AAA wall stress using three-dimensional (3D) reconstructions of patient CT scans have often been used by researchers as a potential patient-specific rupture risk indicator. A property that has a great influence on the stress distribution and magnitude is the aortic wall thickness. Unfortunately, there are no validated, non-invasive methods for measuring aortic wall thickness of patients with AAA. Researchers have utilised either CT or MRI as input into their custom wall detection algorithms, however, there has not yet been a study which utilises both. Therefore, this study aims to develop a non-invasive, and patient-specific method of detecting aortic wall thickness utilising both CT and MRI scans.
> speaker: Andy Huynh
* institution: University of Western Australia
* country: Australia
* city: Perth
* local time: 20:00 AWST
V 12:20 UTC Podium 8
*
Morphological variation in an endothelial cell population: a virtual-cell model
V Yi Chung Lim, Michael Cooling, Sue McGlashan, David S. Long
V The mechanotransmission of fluid-induced forces in endothelial cells is a focal determinant of atherosclerosis. Understanding mechanotransmission is a key step towards developing endothelial-cell based therapies for cardiac diseases. Mechanotransmission is dependent on cell morphology, yet the effect of population morphological variation on mechanotransmission has not yet been examined. We have developed a set of morphological descriptors to quantify three- dimensional morphological variation in a population of human microvascular endothelial cells. From these data we determined quantitatively how the morphology of any cell compares to the overall population. Descriptors were used to generate virtual cells representative of the population morphologies. These virtual cells can be used as the spatial domain in a finite-element analysis. To the best of our knowledge, our study is the first to examine morphological variation in an EC population using a virtual cell approach.
> speaker: David Long
* institution: Wichita State University
* country: United States of America
* city: Wichita
* local time: 07:20 CDT
* 12:40 UTC to 14:00 UTC Break (80 minute)
V 14:00 UTC to 15:40 UTC Sessions
> Chair: Karol Miller
V 14:00 UTC Keynote 1
*
Non-Fourier heat conduction in soft tissues: experiments, models, waves
V Martin Ostoja-Starzewski
V Electrosurgery of soft tissue organs involves application of high voltage at high frequency to the interface between the surgical probe and the tissue boundary. In order to accomplish correct interaction of the electrosurgical probe with the tissue (i.e., coagulation of blood at the probe/tissue interface) as opposed to charring, it is imperative to understand the heat conduction phenomena due to moving heat sources. Recent research [1] has established experimental evidence for the damped-hyperbolic character of transient heat conduction – described by a Maxwell-Cattaneo (not a Fourier) model – in porcine muscle tissue and blood. In fact, a fractional derivative of order α = 0.5 offers the most appropriate model for heat conduction in blood while an integer model is sufficient to describe heat conduction in muscle. Since the thermal signal speeds are on the order of a few millimeters per second, by way of analogies, one needs to consider subsonic (M < 1) and even supersonic (M > 1) second sound phenomena during electrosurgery. In 2d (resp., 3d) problems, this translates into a possible formation of Mach wedges (resp., Mach cones). The subsonic case has recently been examined in [2]. Two types of sensitivity studies of supersonic 2d problems have been performed so far: (i) non-rectilinear paths of a heat source in two-phase inhomogeneous media and (ii) rectilinear paths of heat source in random fractal media. The latter case is motivated by the widely reported fractal character of many biological tissues, while its solution relies on a methodology outlined in [3]. Both types of studies are richly illustrated by 2d computer simulations of transient heat fields.
[1] A. Madhukar, Y. Park, W. Kim, H.J., R. Berlin, L.P. Chamorro, J. Bentsman, and M. Ostoja-Starzewski, “Heat conduction in porcine muscle and blood: Experiments and time-fractional telegraph equation model,” J. Roy. Soc. Interface 16, 20190726, 2019.
[2] Y. Povstenko and M. Ostoja-Starzewski, “Doppler effect described by the solutions of the Cattaneo telegraph equation,” Acta Mech. 232, 725-740, 2021.
[3] X. Zhang and M. Ostoja-Starzewski, “Impact force and moment problems on random mass density fields with fractal and Hurst effects,” in special issue "Advanced materials modelling via fractional calculus: challenges and perspectives," Phil. Trans. Roy. Soc. A 378(2172), 20190591, 2020.
> speaker: Martin Ostoja-Starzewski
* institution: University of Illinois at Urbana-Champaign
* country: United States of America
* city: Champaign
* local time: 09:00 CDT
V 15:00 UTC Podium 9
*
Automatic framework for patient-specific biomechanical computations of organ deformation: an epilepsy (EEG) case study
V Saima Safdar, Benjamin Zwick, George Bourantas, Grand R. Joldes, Simon K. Warfield, Damon E. Hyde, Adam Wittek, Karol Miller
V Our motivation is to enable non-specialists to use sophisticated biomechanical models in the clinic. To further this goal, in this study, we constructed a framework within 3D Slicer for automatically generating and solving patient-specific biomechanical models of the brain. This framework allows determining automatically patient-specific geometry from MRI data, generating patient-specific computational grid, defining boundary conditions and external loads, assigning material properties to intracranial constituents and solving the resulting set of differential equations. We used Meshless Total Lagrangian Explicit Dynamics Method (MTLED) to solve these equations. We demonstrated the effectiveness and appropriateness of our framework on a case study of brain tissue deformations caused by placement of electrodes on the brain surface in intracranial electro-encephalography (iEEG).
> speaker: Saima Safdar
* institution: University of Western Australia
* country: Australia
* city: Perth
* local time: 23:00 AWST
V 15:20 UTC Podium 10
*
Generating scoliotic computed tomography volumes from finite element spine models
V Austin Tapp, Isaac Kumi, Micael Polanco, Sebastian Bawab, Stacie Ringleb, Carl St. Remy, James Bennett, Michel Audette
V The use of deep learning (DL) neural networks (NN) for medical image analysis is dependent on available datasets and associated ground truths. Often, pathological image datasets, especially for the spine, are disproportionately challenging to obtain. To assist DL NN training and contribute to the pathological dataset landscape, this study presents a methodology for the generation of scoliotic computed tomography (CT) volume data. The CT data have associated ground truth segmentations of bone and select soft tissues and are produced from biomechanically based, finite element simulations of spine models. Cervical, thoracolumbar, and lumbar spine finite element (FE) models are deformed by FE analyses prior to their conversion into CT volumes, which are characterized by authentic Hounsfield units. Volumes are tested in a pre-trained vertebral segmentation DL NN to prove compatibility with image analysis methods. Further, an osseoligamentous, lumbar FE model, provides volumetric, ground truth segmentations of soft tissue structures; this affords DL NNs the opportunity to expand segmentation predictions to anatomy, like ligaments, that remain inconspicuous in CT imaging.
> speaker: Austin Tapp
* institution: Old Dominion University
* country: United States of America
* city: Norfolk
* local time: 11:20 EDT
* 15:40 UTC to 16:20 UTC Break (40 minute)
V 16:20 UTC to 17:30 UTC Sessions
> Chair: Poul Nielsen
V 16:20 UTC Podium 11
*
Characterizing the biomechanics of cerebral aneurysms by geometrically nonlinear Kirchhoff– Love shells of nonuniform thickness
V Nicolás Muzi, Francesco Camussoni, Luis G. Moyano, Daniel Millán
V Rupture of intracranial aneurysms is the most common cause of spontaneous subarachnoid bleeding, related to high morbidity and mortality rates. However, intracranial aneurysms have a higher prevalence than that due to their spontaneous rupture rate, exacerbated by the risks associated with occlusion intervention, which motivates the development of technological tools to support clinical diagnosis and endovascular occlusion intervention planning. In particular, the aneurysm dome is sensitive to applied loads in the contiguous surroundings to the aneurysm neck. Indeed, this region shows high complexity due to the arterial wall nature of the pathology. This work presents preliminary statistical analysis results of a thin shell model, with varying material and geometrical parameters, under a localized load emulating the effect of a microcatheter pressing the neck area. In a selection of 34 cases, we show that dimensionality reduction techniques such as Isomap can help determine non-trivial regions of interest under concentrated loads, leading to more general machine learning classification models for sensitive area identification.
> speaker: Nicolás Muzi
* institution: National Scientific and Technical Research Council
* country: Argentina
* city: San Rafael
* local time: 13:20 ART
V 16:40 UTC Keynote 2
*
Personalized modeling of alzheimer’s disease
V Amelie Schafer, Mathias Peirlinck, Kevin Linka, Ellen Kuhl
V More than 50 million people are living with dementia today and this number is expected to triple by the middle of the century. Recent studies have reinforced the hypothesis that the prion paradigm, the templated growth and spreading of misfolded proteins, could help explain the progression of a variety of neurodegenerative disorders. However, our current understanding of prion-like growth and spreading is rather empirical. Here we show that a physics-based reaction-diffusion model can explain the growth and spreading of misfolded tau proteins in Alzheimer’s disease. We combine the classical Fisher-Kolmogorov equation for population dynamics with anisotropic diffusion and simulate tau pathologies across finite element models of the human brain. To reduce the computational complexity, we replace the continuous brain model through a discrete connectivity-weighted Laplacian graph created from 418 brains of the Human Connectome Project. We show that our brain network model correctly predicts the key characteristic features of whole brain models at a fraction of their computational cost. Combined with Bayesian statistics, it allows us to infer personalized kinetic rate constants and diffusion coefficients from a longitudinal study of 76 human brains. Our model has important clinical implications, from predicting the personalized timeline of neurodegeneration to probing the efficacy of pharmacological intervention.
> speaker: Ellen Kuhl
* institution: Stanford University
* country: United States of America
* city: Stanford
* local time: 09:40 PDT
V 17:40 UTC Conclusion
* Karol Miller