Full text
| |
| Author(s): |
Fernando Mitsuyama Cardoso
Total Authors: 1
|
| Document type: | Doctoral Thesis |
| Press: | São Paulo. |
| Institution: | Universidade de São Paulo (USP). Escola Politécnica (EP/BC) |
| Defense date: | 2015-02-24 |
| Examining board members: |
Sergio Shiguemi Furuie;
Julio Cezar Adamowski;
Antonio Adilton Oliveira Carneiro;
Eduardo Tavares Costa;
Pai Chi Nan
|
| Advisor: | Sergio Shiguemi Furuie |
| Abstract | |
Ultrasonography (US) is used by physicians to help on diagnosis and interventions. It provides tomographic views of inner organs such as pancreas, aorta, inferior vena cava, liver, gall bladder, bile ducts, kidneys and spleen. The physician may utilize US to perform only a visual assessment or may also compress the tissue to analyze its dynamics, since lesion elasticity may be related to dangerousness. Consequently, several computational procedures have been developed in order to provide information about the elastic properties of the tissue. However, a thorough and objective evaluation of US computational procedures may be hindered by the difficult access to US images with the desired features and the lack of gold-standard parameters. Therefore, we developed a tool that is able to create numeric phantom that mimics the compression induced by physician with the transducer. The tissue deformation was based on finite elements method and the displacement of the scatterers were calculated using linear isomorphism. After the scatterer displacement, Field II was used to simulate the speckle noise. Thus, it is possible to create a sequence of US images with realistic deformation. This technique was implemented in Matlab and is available for free download. The phantom deformation was validated by measuring the strain contrast from double-layered phantoms. Special attention was given to cardiovascular diseases due to their impact on Brazilian and world populations. During the last decades, the prevalence of cardiovascular pathologies has increased progressively, and has become a serious public health problem. They are among the major causes of death, hospitalizations and health expenses. In interventionist practice, intravascular ultrasound (IVUS) is used to obtain information about blood vessels and eventual pathologies. Therefore, we also created numeric IVUS phantoms. The simulation of the blood vessel was also based in finite elements method with linear isomorphism. However, a reliable IVUS simulation must consider the catheter path inside the blood vessel, because it determines the position of the transducer. Hence, we developed a new method, based on equilibrium of forces, to determine the minimum energy position of the catheter. The method was validated by comparing its position with the position of a real stainless steel IVUS guidewire and presented root mean squared error and Hausdorff mean smaller than 1 mm for both. We used two different techniques to track and estimate deformation of different structures in the simulated US images, namely, Optical Flow and 2D Block Matching. We applied an innovative implementation of 2D block matching with sub-pixel linear interpolation and displacement propagation. Then, the estimated deformation from both methods were compared with the numeric gold-standard, and 2D block matching presented better results than optical flow. After the work with numeric phantoms, real US equipment was utilized to acquire B-mode images from a physical phantom. Then, we performed the movement estimation of the imaged tissue to analyze its morphological and dynamic properties. The results were compared to the elastography images provided by the US equipment. In accordance with the results from the numeric simulation, 2D block matching presented better results than Optical flow. Finally, we performed the two speckle tracking on a set of numerically simulated IVUS images, where the images were divided into two sets of frames. The first set, S1, contained all the frames from the IVUS sequence and the second set, S2, contained only the frames corresponding to a specific phase of the cardiac cycle. Thus, we analyzed the trade-off between the impact of the cardiac motion and low frame rate. For the points located at the edges of the object, optical flow had a good performance for both S1 and S2. In homogenous regions, however, optical flow was able to track the points only in S2, suggesting that it is better to work with low frame rate and reduced cardiac motion, as in EKG-triggered IVUS acquisition. 2D block matching presented poor results in all points of both S1 and S2. Besides the simulation of ultrasound acquisition with deformation and structure tracking, in this work, we also developed a new filtering technique that is able to remove the speckle texture without blurring the edges. The proposed filter presented the best results when compared to other nine filters from the literature. We also developed a metric that uses the speckle texture of the image to provide a parameter that may help the user decide the size of the window of the filter. (AU) | |
| FAPESP's process: | 11/01314-3 - Estimation of structures movement based on echographic images |
| Grantee: | Fernando Mitsuyama Cardoso |
| Support Opportunities: | Scholarships in Brazil - Doctorate (Direct) |