SPECKLE REDUCTION FOR THE ENHANCEMENT OF ULTRASOUND IMAGES

Abstract

Ultrasound imaging has been established as one of the most important techniques in the field of the medical diagnostic technology because of its non-invasing nature, portability, low cost and real time information. The image quality is of central im-portance in an ultrasound examination as the diagnosis of a disease by a radiologist is based on his interpretation of the medical images. However, ultrasound images suffer from an intrinsic artifact called speckle. Speckle degrades spatial and contrast resolution and obscures the underlying anatomy. It makes human interpretation and computer-assisted detection techniques difficult and inconsistent. Since speckle is a major shortcoming of ultrasound, reducing or eliminating speckle is necessary for the visual enhancement and auto segmentation improvement. In this thesis work speckle reduction techniques have been implemented to in-crease the visualization and to make the auto segmentation process easy and fast in medical ultrasound images. The adaptive filtering and anisotropic diffusion based techniques have been ex-amined and compared on the basis of their speckle suppression ability and feature preservation. Under the adaptive filtering, adaptive weighted median filter (AWMF) and aggressive region growing filter (ARGF) have been implemented. Both algo-rithms uses local statistics of the image for the filtering action. AWMF is an enah-nced median filter and it is based on weighted median. Aggressive region growing filter (ARGF) selects a filtering region size using an appropriately estimated ho-mogeneity value for region growth. In case of diffusion based techniques speckle reduction anisotropic diffusion (SRAD) has been applied. SRAD is based on the same minimum mean square error (MMSE) approach to filtering as Lee and Kuan filters. These filtering algorithms are applied on the simulated and the tissue mimicking phantom to have a quantitative analysis. To study the feasibility and usefulness of these methods, the algorithms are applied on the real ultrasound image taken from GE website and medpix database. To quantify the results obtained by these three techniques, evaluation indices (CNR, FOM, MSSIM) have been calculated. It is found that SRAD is superior to adaptive filtering based- techniques in in-creasing the visualization of the images while reducing the speckles. AWMF cannot remove the speckles effectively and also causes blurring with the loss of details. ARGF technique reduces the speckles effectively and its smoothening effect can be used as a preprocessing step for auto segmentation and image registration. Qualitative analysis for these three methods has been done by the assistance of the medical experts. The results obtained may be useful for radiologists, clinicians or experts , who may use it for clinical diagnosis. 3