Date of Award

4-10-2017

Publication Type

Master Thesis

Degree Name

M.Sc.

Department

Physics

Keywords

Algorithm, Blind-Deconvolution, Edge detection, Profile Extraction, Signal Processing, Ultrasound

Supervisor

Maeva, Elena

Supervisor

Maev, Roman

Rights

info:eu-repo/semantics/openAccess

Abstract

Ultrasound Imaging has been favored by clinicians for its safety, affordability, accessibility, and speed compared to other imaging modalities. However, the trade-offs to these benefits are a relatively lower image quality and interpretability, which can be addressed by, for example, post-processing methods. One particularly difficult imaging case is associated with the presence of a barrier, such as a human skull, with significantly different acoustical properties than the brain tissue as the target medium. Some methods were proposed in the literature to account for this structure if the skull's geometry is known. Measuring the skull's geometry is therefore an important task that requires attention. In this work, a new edge detection method for accurate human skull profile extraction via post-processing of ultrasonic A-Scans is introduced. This method, referred to as the Selective Echo Extraction algorithm, SEE, processes each A-Scan separately and determines the outermost and innermost boundaries of the skull by means of adaptive filtering. The method can also be used to determine the average attenuation coefficient of the skull. When applied to simulated B-Mode images of the skull profile, promising results were obtained. The profiles obtained from the proposed process in simulations were found to be within 0.15 λ ± 0.11 λ or 0.09 ± 0.07 mm from the actual profiles. Experiments were also performed to test SEE on skull mimicking phantoms with major acoustical properties similar to those of the actual human skull. With experimental data, the profiles obtained with the proposed process were within 0.32 λ ± 0.25 λ or 0.19 ± 0.15 mm from the actual profile.

Download

Share

COinS