Development of multi-compartment model of the liver using image-based meshing software

Abstract

Computer simulation of biological systems for in silico validation has the potential of increasing the efficiency of pharmaceutical research and development by expanding the number of parameters tested virtually. Then only the most interesting subset of these has to be probed in vivo. By focusing on variables with the greatest influence on clinical end points, valuable drug targets can be advanced more quickly. A large number of methods have been developed to rebuild a three-dimensional (3D) model of a liver, mostly to prepare a liver surgery. These models are often not accurate and most of the them don't take into account the fluidics inside the vessels. The aim of this work is to provide an accurate computational multi-compartement model of the healthy and the pathological liver with their network of blood vessels (vasculature) using a finite-element-modeling software. Computed tomography (CT) slices, in DICOM format, from two different patients were used to provide the datasets of transverse images for the modeling. Each dataset of images was segmented in order to extract the liver’s shape and define the vein and artery networks. On CT images, the contrast between the liver and the nearby organs (background) is very low because all these structures are a similar density. Thus, we used semi-automatic tools to determine liver contours. Manual segmentation was used as a last resort. Then, strong filtering (bilateral filter) and confidence-connected-region-growing algorithm were applied to rebuild from each - healthy and pathological - liver a multicompartment model including parenchyma, arteries and veins. The precision of the obtained vasculature model allowed anatomical classification of hepatic segments and the quantification of their volumes. Although our study demonstrated the difficulties in use of CT images for computational modeling of the liver, it also confirmed that semi-automatic tools can be used to develop anatomically accurate models of hepatic vasculature.

A Hybrid Method for Pancreas Extraction from CT Image Based on Level Set Methods

“…The proposed hybrid level set method was compared to confident connected region growing method (CCRG) [2], geodesic active contour method (GAC) [6], geodesic active without edge method (C-V) [7], fast-marching method (F-M) [9], and a combined edge-region level set method (CER) [12]. Our method which depends on standard line is referred to as FMDSL-SL.…”

“…Patients are commonly examined using abdominal computerized tomography, and pancreas extraction from CT image can improve a variety of clinical applications. However, medical images are often impacted by noise and distortion, which causes difficulties to apply conventional segmentation methods, such as edge detection methods and region growing methods [2] to extract pancreas.…”

“…The accuracy of the proposed method is compared to region growing method [2], GAC model [6], C-V model [7], fast-marching method [9], and combined level set method [12]. Three measures, (1) false-positive error (FPE), (2) false-negative error (FNE), and (3) the similarity index (SI), are used as evaluated standards of accuracy.…”

A Priori Knowledge and Probability Density Based Segmentation Method for Medical CT Image Sequences

“…R p ( ϕ ) is defined in [11] by $\begin{array}{c}{R}_p\left(\varphi \right)\triangleq {{\displaystyle false\int }}_{\Omega }p\left(\left|\nabla \varphi \right|\right)dx,\end{array}$ where p is a double-well potential function for the distance regularization term R p and is constructed as $\begin{array}{c}p\left(s\right)=\{\begin{array}{cc}\frac{\mathrm{normal1}}{{\left(\mathrm{normal2}\pi \right)}^{\mathrm{normal2}}}\left(\mathrm{normal1}-\cos \left(\mathrm{normal2}\pi s\right)\right),\hfill & \text{if}\hspace{0.17em}\hspace{0.17em}s\le \mathrm{normal1}\hfill \\ \hfill & \hfill \\ \frac{\mathrm{normal1}}{\mathrm{normal2}}{\left(s-\mathrm{normal1}\right)}^{\mathrm{normal2}},\hfill & \text{if}\hspace{0.17em}\hspace{0.17em}s>\mathrm{normal1}.\hfill \end{array}\end{array}$ …”

“…The level set evolution equation in a priori based distance regularity level set is finally defined by $\begin{array}{c}\frac{\partial \varphi }{\partial t}=\mu \mathrm{div}(d_p\left(\left|\nabla \varphi \right|\nabla \varphi \right)+\alpha s\left(I_m\right){\delta }_{\epsilon }\left(\varphi \right)\\ +\lambda {\delta }_{\epsilon }\left(\varphi \right)\mathrm{div}\left(g\left(I\right)\frac{\nabla \varphi }{\left|\nabla \varphi \right|}\right)),\end{array}$ where div⁡(·) is the divergence operator and d p is a function defined in [11]: $\begin{array}{c}{d}_p\left(s\right)\triangleq \frac{p^{\prime }\left(s\right)}{s}.\end{array}$ …”