چكيده به لاتين
A common method to investigate Cardiac and Vessels functionality is echocardiography imaging. This imaging modality has been a growing clinical test compared to Magnetic Resonance Imaging (MRI) since ultrasound waves are not harmful to the body, are non-ionizing, low-cost, and portable, making them the first choice for most clinicians.
To take the most advantage of this modality, there are shortcomings to overcome. The major challenges are low image quality compared to MRI and low frame rate due to constant sound speed in living tissues. These make interpretations difficult and further enhancements seem to be necessary.
In the present work, methods have been proposed to overcome the limitations for better utilization of ultrasound images. These methods help to perform speckle reduction, frame rate enhancement, better speckle tracking, and tissue segmentation more precisely.
To this end, temporal super-resolution of 2D/3D echocardiography frames has been performed by interpolating Intensity Variation Time Curves (IVTCs) with the use of B-spline functions as a continuous representation of discrete pixel samples through time.
It has been further shown that the proposed frame rate enhancement can help better track speckle patterns in 3D echocardiography images through time by making speckle structures moving more slowly, which makes common speckle tracking algorithms act more precisely in addition to running faster by making searching span more limited.
As an enhancement to image quality, a separation method has been proposed to decompose an image into a texture and a cartoon image. These subcomponents are shown to be useful to reduce speckle inside the ventricle and exaggerate endocardial walls so that image quality criteria have been grown and endocardial segmentation has been performed more precisely.
To decompose the image into subcomponents, an optimization function has to be solved, minimizing cartoon image variation norm, reconstruction error, and the number of representation coefficients of the texture image. To optimize the function, a distributed optimization algorithm has been utilized called the Alternating Direction Method of Multipliers (ADMM) which makes it possible to perform large optimization computations performed on different processing systems.
To show the effectiveness of these methods, numerical evaluations along with qualitative results have been provided. Numerical parameters evaluate image quality, structure preservation, error with the original frame, and segmentation accuracy for both segmentation and tracking tasks. Therefore, it has been shown that the proposed methods are effective in improving the performance of ultrasound imaging in terms of quality improvement and so on.
