Blood flow at the bifurcation regions of the arteries which are connected to the main artery. The aim is to investigate the effect of blood flow on plaque development and the effect of this plaque on blood flow. So, it may be possible to simulate the experimental studies which may be expensive and long time consuming and sometimes impossible. It might be possible to slow down or stop the growth of paque by determining the main affects that help plaque to develope. Adding a fluid structure interaction (FSI) model to the simulation helps the simulation be more realistic.

                   

(a) Velocity profile at artery cross section       (b) Artery pressure distribution                 (c) Streamlines

The mechanical properties at main artery will be investigated using fluid structure interaction model. Large arteries will be included in to this model using only a flow simulation. So the arterial network simulation will be performed. The data obtained using these simulations will be used to study the biomechanical properties of human arterial system.

Sample carotid artery model

 

Most CFD codes need a mesh data to simulate the flow. In this project 3D surfaces of several arteries were obtained using CT and MRI data. These data will be used to obtain appropriate surface and volmume meshes which will be used in flow simulations.

 

           

The computer models of physical objects will be a closer abstraction to the objects they representing if both of their visual and behavioral aspects are well defined. Visual reality presented by these modes dictates how model will be perceived and viewed by the user and behavioral aspects embedded within the model determines its physical interactions to its environment [1]. Often user’s interaction is needed as inputs to these models which may be catered for by haptic devices that are capable of mimicking real world experiences. The above mentioned aspects should be made present in a single framework if users “real time” and “real world” interactions are desired. The research to date on this subject has been concentrated within the field of computer graphics. Within the last decade an increased activity in this research resulted in “physically based models” which have found their ways in computer animation, virtual reality, games, and virtual surgery applications.

                   

Click here for details >>

In this thesis, we propose a trainable filter which recovers most of chest data modified by chest tube attenuations. It describes how you may remove the chest tube figure (as an example of artificial object might be presented at the radiography) from Postero-Anterior (PA) Chest Radiography.

We decide to focus on this issue, because the chest radiography is the most common radiological modality in the practice. Especially, it is very popular for scanning and screening purposes. Unfortunately, at the same time it is one of the most difficult radiological modalities. The overlapping tissues cause a highly complex projection. In addition, artificial objects such as catheters, chest tube, pacemaker, and/or even cloths might be presented at this projection image. It is obvious that the anomaly detection algorithm should not be confused by these objects.

To achieve this goal, we study to train a Convolutional Neural Network (CNN) that gives an artificial x-ray without a chest tube as output for a given x-ray image which contains the chest tube as an input.

The results show that our model can remove chest tube figures from radiographies. We believed that our model gives promising results as a starting point for an artificial object removing problem. We show that a module to protect the Computer Aided Diagnosis (CAD) system’s accuracy from errors caused by foreign objects is possible.

 

Click here for next page >>