Master thesis presentation - Nils Karlsson
TITLE: Aortic 4D modeling: Real-time adaptive model using MR images
EXAMINER: Prof Mikael Persson, Chalmers S2
START: 2011-09-02 15:30
END: 2011-09-02 16:30
EVENT LOCATION: Lecture hall Kammaren, Sahlgrenska University hospital
Link to PDF document.
ABSTRACT:
A method of modeling the aorta in real time using images from an MRI scanner has been implemented within an existing interventional MRI interface and empirically evaluated. Classically, the generation of such models is performed following complete acquisition of 3D image data. In contrast the proposed algorithm makes use of data that is sparse; i.e. arbitrary MRI images. The algorithm is based on recent strategies for generating these kinds of models based on implicit segmentation. It was designed in the context of interventional MRI with the goal of being a visual aid; e.g. to aid physicians during aortic interventions such as stent placement. For the generation of the model, images from an MRI scanner are continuously acquired and used to optimize the model.
The model uses an initial model of the aorta that is then optimized with implicit segmentation. The implicit segmentation is based on histograms of the pixel values inside and outside of the aortic model and on edge data; i.e. gradients in the image. Testing of the algorithm was carried out using healthy volunteers and both data acquired off-line and in a real-time setting. In particular the model was used to measure the radius along the centerline of the aorta for 8 positions. The results demonstrate that the measurement error ranged from 0.31-1.52mm.
