Please use this identifier to cite or link to this item: http://hdl.handle.net/123456789/14488
Title: PROCESSING OF ECHOCARDIOGRAPHIC IMAGES WITH REFERENCE TO MITRAL REGURGITATION
Authors: Saini, Kalpana
Keywords: MR Common Heart Valve Disorder
Here After
De-Speckled Images
Modified Non-linear
Issue Date: Jun-2013
Publisher: Dept. of Electrical Engineering iit Roorkee
Abstract: The present thesis deals with the research work carried out for the automatic severity evaluation of the mitral regurgitation (MR). MR is a common heart valve disorder. In present days, the clinicians carry out this task by extracting the regurgitant jet area from the echocardiographic images, manually. Since, the echocardiographic images primarily suffer from the multiplicative speckle noise, efforts have been made to clean the echocardiographic images to enable clinicians to obtain higher MR assessment accuracies. Here after, in the thesis, only the word images shall be used to refer to the echocardiographic images. Digital image processing has been done in the MATLAB environment. To facilitate the manual analysis of the MR images by the clinicians the research efforts have been worked out to provide clinicians with the much better contrasted enhanced and almost completely de-speckled images with the view to improve accuracy of his assessment because of the improved visual enhancement. The contrast enhancement has been achieved using the proposed modified log transformation technique and the images has been de-speckled using the newly developed two techniques; the Hybrid Speckle Reducing Anisotropic Diffusion (HSRAD) filter and the Modified Non-linear Complex Diffusion (MNCD) filter for cleaning the images of the multiplicative speckle noises. For the contrast enhancement the log transformation has been modified to give better contrast enhanced images than the one which are obtainable using conventional log transformation. Results demonstrate that the proposed modified version of log transformation gives better image enhancement one obtainable using conventional log transformation technique. The results are better balanced contrasted images bearing much higher quality indices proving to be more useful for manual processing in assessing the MR severity. For de-speckling the two newly developed algorithms have been proposed which produce much better results when compared with the results obtainable using existing filters which are also employed on echocardiographic images besides other types of images. These two new filters allow clinicians to carry out flawless diagnosis. One of the proposed de-speckling filters is a hybrid filter, namely HSRAD filter is a combination of speckle reducing anisotropic diffusion (SRAD) filter and the relaxed median filter. Processing of echocardiographic images with reference to mitral regurgitation iv Experiments are carried out on a number of images. Results show significant reduction in speckle noise while retaining the edges at the same time. The proposed filter removes speckle noise as well as the impulse noise and enhances the edges. The second proposed de-speckling filter MNCD filter is developed to remove speckle noise more effectively from the images and provides much better enhancement even of the finer details of the image. The performance of the MNCD filter has been found to be much better than the existing de-speckling filters with particular reference to echocardiographic images. The results are compared with the results of the other existing filters. The clinicians’ assessment of MR, carried out manually, is thus, taken as the ground truth to compare with the results obtained from the following designed and developments of algorithms for ‘Automatic MR Severity Evaluation’ techniques. The thesis presents the three novel approaches for automatic MR severity grade evaluation. The first approach is the boundary detection of heart’s chambers. Two new algorithms viz. the Fast Region Based Active Contour Model (FRACM) and the New Selective Binary and Gaussian Filtering Regularized Level Set (NSBGFRLS) model have been proposed for the purpose of above stated boundary detection. The second approach is the detection of MR jet area which employs the two newly proposed techniques; A Region Growing based Mosaic Jet Segmentation (RG-MJS) and the Combined Wavelet and Watershed Transformation based Mosaic Jet Segmentation (CWW-MJS), for carrying out automatically the MR mosaic jet segmentation and evaluation of its area. The third approach has been the Vena Contracta Width (VCW) detection. A technique which automatically finds out the location of VC and calculates the VCW has been developed and proposed. The results evaluated automatically using the proposed technique and assessed manually by the clinicians are compared. The comparison corroborates results within very close limits. In case of significant MR, left ventricle has to accommodate both the stroke volume and regurgitant volume with each heart beat so it leads to volume overload of the left ventricle. The left ventricle dilates and becomes hyper-dynamic for compensation. The left atrial and pulmonary venous pressures increase sharply in case of acute severe MR, leading to pulmonary congestion and pulmonary edema. A gradual increase in left atrial size and compliance compensates in chronic MR, so that left atrial and pulmonary venous pressures do not increase until late in the course of the disease. An increase in after load, contractile dysfunction, and heart failure occur in case of progressive left ventricular dilation. This entails the detection of boundaries of heart’s chambers, for which two new models, viz. the Fast Region Active Contour Model (FRACM) and Abstract v the Novel Selective Binary Selective Regularized Level Set (NSBGFRLS) have been developed and presented in the thesis. The proposed models the FRACM and the NSBGFRLS are the much faster algorithms than the existing algorithms to detect the boundaries of the heart chambers. The performance of these two boundary detection models has been experimented and the results are tabulated, plotted and compared with the performance of other existing models which are also employed for boundary detection of echocardiographic images. The performance of the models has been found to be much superior to other existing models. The second approach to detect the severity grades of MR is to evaluate the value of the MR jet area. The objective is to segment out the color mosaic pattern from the color Doppler image. At present the clinicians draw this area and assess the value of the regurgitation jet area manually. If this area is found out automatically then it will save a lot of time of the clinicians and allow higher measurement accuracies. The two novel methods, RG-MJS and CWW-MJS have been proposed for the automatic evaluation of the mosaic MR jet area. The methods have been applied on some ten patients who were suffering from varied grades of MR severity from mild to severe. The RG-MJS method is recommended for those patients who are suffering from moderate or severe MR. Whereas the method CWW-MJS is recommended for all the patients who may be suffering from MR of any one of the three severity grades viz. mild, moderate and sever. The results of experiments have been compared with the results of manual assessment of MR assessed by the clinicians as the ground truth. The third approach to detect the severity grading of MR is to calculate the vena contracta width (VCW) of the MR jet. The narrowest part of the jet that appears just downstream from the orifice is referred to as the vena contracta. High velocity and laminar flow are the characteristics of the vena contracta. A mathematical expression has also been proposed for the calculation of VCW. The VCW is smaller than the anatomic regurgitant orifice to some extent due to boundary effects. It is difficult to locate this narrowest part exactly in TTE or TEE images. It entails to develop a method which should process the MR images to locate and find out the numerical value of VCW. A simpler method than the complexity of the mathematical derivation has been developed and proposed for getting the numerical value of the vena contracta width. The proposed method also automatically finds out the location of the VCW for which it does not require zooming out of the vena contracta width portion of the MR jet images. Processing of echocardiographic images with reference to mitral regurgitation vi Clinical values have been taken as ground truth for the jet area and vena contracta width comparisons. MR severity comparison has been done between clinical values and values obtained from proposed methods. Results are very much in agreement and within very close limits. The results of experiments are very encouraging and the proposed developments find clinical applications. To demonstrate application a Graphical User Interface (GUI) which is supported by all the existing and the proposed algorithms has been designed. With the help of the proposed GUI the clinicians or any other user will be able to process images just by selecting the name of the method from the GUI panel; without going in to complexity of the mathematical treatment of the techniques. The GUI facilitates processing of images, provides comparison between different methods, and also displays the measured values of MR jet area and VCW in terms of numerical digits.
URI: http://hdl.handle.net/123456789/14488
Appears in Collections:DOCTORAL THESES (Electrical Engg)

Files in This Item:
File Description SizeFormat 
G24346-KALPANA-T.pdf9.48 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.