Patient specific cardiovascular disease modelling based on the computational fluid dynamics simulations: segmentation and hemodynamic model of a thoracic artery

In this study, a patient specific thoracic artery model is first segmented based on the MRI images and then a thoracic aneurysm disease model is simulated to assess blood flow changes, for maximum flow condition


Graphical Abstract
In this study, a patient specific thoracic artery model is first segmented based on the MRI images and then a thoracic aneurysm disease model is simulated to assess blood flow changes, for maximum flow condition, and generated stress during the circulation.

Aim
The aim of this study is to obtain a patient specific thoracic artery model and perform fluid dynamic simulations based on the predefined disease conditions.

Design & Methodology
MIMICS software was utilized to segment the artery model.Subsequently, target artery model was imported into SolidWorks to perform simulation scenarios based on the aneurysm progression.

Originality
Patient specific thoracic artery segmentation is presented step by step.Finally, the STL artery model file was utilized in simulations with predefined conditions to observe effects of aneurysms in advance.

Findings
Blood flow velocity values were increased during the systole up to 0.2m/s and decreased during the diastole down to 0.02m/s on each region of the vessel.Maximum 0.0002MPa vonMises stress generated in more progressed aneurysm disease model An artery is composed of several tissue layers namely: i) Tunica intima, the innermost layer of an artery, ii) Tunica media, enclosing surface of the intima and iii) Tunica externa or Tunica adventitia, the outermost layer of an artery, that naturally aligned, biomechanically elastic and strong enough to stand the required blood pressure during circulation in body [5].The intima layer is formed with endothelial cells that directly affect the vessel physiology owing to the interaction between the blood flow and endothelium.Changes in the flow may cause a vascular pathogenesis.The intima layer is surrounded with the media layer formed with smooth muscle cells (SMCs).Thickness of the media layer may vary according to the vessel type.The media layer is the thickest layer of an artery while it is thinner than the externa layer in veins.
The elasticity of a vessel is also related to thickness ratio of both media and externa layers or vessel diameter.Therefore, more elastic behavior is expected in larger arteries with regards to pulsatile blood flow (Womersley flow) control.The externa layer is mostly comprised of collagen that guides to SMCs by means of micro fibers to achieve the required cellular alignment in a vessel structure.Furthermore, collagen layer protects the inner layers from the environmental risks in a body.Each tunica layer is supported with internal and external elastic membranes and elastic fibers.Flow dynamics of a vessel is important for diagnosis or prognosis of a CVD [6].Since blood directly interacts with endothelial layer during the circulation, a morphological change may be emerged by means of triggered chemical pathways owing to the deformation response of the endothelium against the generated shear forces (Wall Shear Stress -WSS) in a vessel.In this case, it may be vital to guide and determine the symptoms of an onset or a progression of CVDs [7,8].Therefore, physiological responses of healthy and diseased vessels are compared with each other to achieve a meaningful difference in terms of flow conditions and shear stress distribution [9].However, hemodynamic parameters may not be measured directly and even blood flow hemodynamics require detailed assumptions to create a realistic physiological model [9,10].
Recently, computational methods are increasingly being used in the fields of neurosurgery and cardiovascular surgery, atherosclerosis and aneurysm cases, to obtain realistic hemodynamic simulations [11,12].Besides, these simulations may be utilized in surgical planning [13,14].Thus, efficiency of the computational simulations is improved as well as patient's quality of life by means of the developed patient-specific disease models [15].In this study, a patient specific thoracic artery model is first segmented based on the MRI images and then a thoracic aneurysm disease model is simulated to assess blood flow changes, for maximum flow condition, and generated stress during the circulation.(tube current), 125mAs (time x milliamperes) with a collimation of 64 x 0.5mm and a rotation time of 0.5s in the raw data acquisition step.In addition, a contrast media (100mL) and a saline solution (40 mL) were injected with respectively at the same injection rate.Image reconstruction process was performed at: 7mm for section thickness, 0.3mm for overlapping steps and 700mm2 for FoV (Field of View) with an automated exposure control (2.5mm overlap).Thus, the raw data acquisition process was completed.MIMICS software (v19) was utilized in image segmentation step and a reference model of the aorta was automatically reconstructed by means of the default tools and functions of the software.Noise reducing filters such as binomial blur and mean tools were also used to reduce undesired noise from the region of interest on each twodimensional (2D) slice.Segmentation process was performed using the Thresholding tool at optimized value for target soft tissue (144 -Hounsfield Unit, HU as lower threshold, and 446 HU as higher threshold for this study).These parameters were ideal to obtain exact contour of the target model in a maximum allowable noisy form for this case [22].Segmentation step was completed after each image section was masked and highlighted by a different color.Subsequently, three-dimensional (3D) calculation function was used at high quality to reconstruct the 3D processable raw model.The surface geometry of the model was manually cleared via 3-Matic software (v11, Materialise) according to the raw image geometries in MIMICS.Polygon Area Mark tool under the Mark -Area Mark menu was used to select each noisy surface and marked surfaces were deleted.The gaps on deleted surfaces were determined as Bad Contours and then fixed respectively using the Fill Hole Freeform tool under the Fix menu.The Fill Hole Freeform process was performed at High triangulation quality and formed in Tangent shape to achieve desired detail on processed surfaces.Finally, 3D reconstructed model was saved as Standard Tessellation Language (STL) file (although STL is a common file extension, it is not possible to perform simulation analysis to target model in all cases).Therefore, thoracic aorta model was created by means of SolidWorks (Dassault Systemes, v2016) according to the obtained manual measurements in MIMICS.After manual modelling, flow simulation tool was utilized to obtain blood flow simulation during the systole and diastole phases (0-0.8s).Finally, an aneurysm condition was simulated on the same aorta model to obtain blood flow properties using the predefined parameters as listed in Table 1.Pressure was determined according to Systole and Diastole conditions (120-80 mmHg in terms of Pa). b.Body temperature (in terms of K). c. Velocity was determined according to Systole and Diastole conditions during a pulsatile waveform (in terms of m/s).d, e, f.Default values.

MATERIAL AND METHOD
g.
Manually measured in MIMICS.

RESULT AND DISCUSSION
According to the results, blood flow velocity values were obtained as desired on each thoracic aorta model, increased during the systole and decreased during the diastole on each region of the vessel.Besides, the velocity values in aneurysm disease model were also increased with regards to the geometrical narrowing.Blood flow velocities of each vessel model are illustrated in Figure 2. Finite element analysis (FEA) results of a more progressed aneurysm disease model, based on the flow simulation (blood pressure) analysis and vessel parameters that stated in Table 1, are illustrated in Figure 3 (T=0.2ssystole condition only and vessel diameter was decreased down to 10mm, 25mm default).Blood flow velocities were obtained as a) 0.05m/s (T=0s), b) 0.01m/s (T=0.1s),c) 0.039m/s (T=0.2s),d) 0.044m/s (T=0.3s),e) 0.005m/s (T=0.4s),f) 0.002m/s (T=0.5s),g) 0.006m/s (T=0.6s),h) 0.006m/s (T=0.7s),i) 0.0065m/s (T=0.8s) and j) 0.07m/s (T=0.2s) at ADS regions.On the other hand, a vonMises stress at 2.89MPa was generated with a material displacement of 0.007mm on the arch of aorta model (vessel diameter was decreased down to 10 mm at ADS region) at Systole phase (T=0.2s).

CONCLUSION
Although computational fluid dynamics with the liquidmaterial interaction may help to obtain physiological information and biological response under the specified conditions that may not be measured directly, simulations require more precise definitions to achieve more realistic results.Since medical imaging technologies may allow to export target anatomical model in a STL file extension and although STL is a common file extension, it is not possible to perform a dynamic simulation analysis to target model in all cases.Therefore, the target model may require to be modeled manually or a 3rd party software such as Comsol, Abaqus or Fluent may be utilized to overcome the lack of STL file support especially in dynamic analysis [23,24].On the other hand, researchers may perform static analysis on an STL file by means of a tool such as Power Surfacing or may apply directly to the generated mesh surfaces of the model in SolidWorks [25].In addition, FEA of a blood vessel may require to define biomechanical properties of each vessel tissue layer separately.Thus, more accurate shear rates, regarding to the blood flow, may be obtained [26].Furthermore, some parameters such as gravity or temperature may also be involved.
According to results of simulation, blood flow velocity is affected by ADS regarding to the decrease on the vessel diameter (Normal:25 mm, ADS: 20 mm).The velocities are almost doubled for Systole phase (T=0.2s) at ADS regions (Normal: 0.039m/s, ADS: 0.07m/s).Material displacements are mostly observed at the arch of aorta and ADS regions due to the curvature changes on laminar blood flow path.Although blood flow velocities may vary regarding to the vessel geometry or predefined parameters such as blood and flow conditions, similarities of the flow velocity are confirmed from the previous studies [27].In addition, Reynolds number (Re), a ratio of the inertial forces to viscous forces, can be calculated to determine the blood flow type (Laminar or Turbulent) for each pulsatile blood flow phase by using determined blood flow velocities at ADS regions.High Re up to 4000 indicates that inertial forces are dominant while low Re up to 2300 indicates viscous forces are significant.As Re may vary during the pulsatile blood flow, it is important to calculate the WSS affected by turbulent flow, according to natural geometry of the vessel or an implanted stent, to predict biomechanical stress on the vessel walls [28].It should be noticed that vessel walls are assumed as rigid and blood is assumed as fully Newtonian to simplify calculations as well as removing branched vessel geometries [29][30][31].However, it may not give the real accurate results since these conditions are accepted as ideal.Disease modeling may be guide to obtain physiological responses against a patient specific progression scenario or predict symptoms of a CVD.Thus, development of a disease, such as a development stage or a rupture of an aneurysm or even a hypertension, may be simulated to calculate the damage [32].Furthermore, simulation analysis may be useful for surgical planning and even preventive management as well as it may be utilized in drug research responses or even aging effects (European Commission supported projects, Research and Innovation in the field of ICT for Health, Wellbeing & Ageing Well: Cardioproof, EurValve, SmartTool) [33].
: torasik arterin segmentasyonu ve hemodinamik modeliYazar(lar) (Author(s)): Levent AYDIN 1 , Serdar KUCUK 2 , Ozgur CAKIR 3 ORCID 1 : 0000-0003-2926-2824 ORCID 2 : xxxx-xxxx-xxxx-xxxx Patient Specific Cardiovascular Disease Modelling Based on The Computational Fluid Dynamics Simulations: Segmentation and Hemodynamic Model of a Thoracic Artery Highlights  A patient specific thoracic artery model is segmented based on the MRI images,  A thoracic aneurysm disease model is simulated to assess blood flow changes and stress during the circulation.
h. Calculated using the Mass Flow's equation (m = ρ x V x A where ρ: Density of the flowing liquid or gas, in kg/m 3 , V: Flow speed, in m/s, A: Flow area, in m 2 and m: Mass flow rate, in kg/s).3D modelling and flow simulation of the reference thoracic aorta model, including the aneurysm condition, are illustrated step by step in Figure 1.

Figure 1 .
Figure 1.Step by step 3D modelling and flow simulation of the patient specific thoracic aorta and aneurysm disease model.a) An image of the 3D raw model, based on non-filtered DICOM images in Sectra workspace, a 3D DICOM viewer of the default system software that does not allow 3D model exporting or processing, b) 3D reconstruction of the raw images in a noisy form via MIMICS, c) Manual noise cleaning in 3-Matic, d) Manual surface reconstruction and fixing the geometrical errors of 3D raw model in 3-Matic, e) Manual vessel diameter measurements in MIMICS, f) 3D modelling of the thoracic artery in SolidWorks based on the manual measurements in MIMICS, g) Modelling the aneursym disease scenario (ADS) in SolidWorks and h) Flow simulation analysis of the thoracic aorta model, including aneurysm condition, during the each phases of the pulsatile blood flow.

Figure 3 .
Figure 3. Blood flow velocity based FEA results (T=0.2ssystole condition only and vessel diameter was decreased down to 10mm).a) Material displacement values under the applied fluid dynamics and b) Generated vonMises stress against the blood flow.

Table 1 .
Flow simulation parameters a.