Use this URL to cite or link to this record in EThOS:
Title: Computer-aided ultrasonic tissue characterization
Author: Eastwood, Linda M.
ISNI:       0000 0001 3437 3559
Awarding Body: University of Aberdeen
Current Institution: University of Aberdeen
Date of Award: 1980
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Pulse-echo ultrasound is widely used in medical imaging. This thesis examines the problem of characterization of tissues from properties of the echo signals. The potential value of techniques from the field of statistical pattern recognition is examined. As a background to the problem, the generation and reception of ultrasound are discussed. A computer model of a plane piezoelectric transducer is presented. The structure and acoustic properties of tissue are reviewed, together with current theories for the physical basis of interactions of ultrasound in soft tissue. A computer implementation of a "plane parallel layers" model of tissue is used (in conjunction with computer models of transducer and associated electronics) to illustrate the relationship between tissue structure and backscattered ultrasound signals. The approaches to in-vivo ultrasonic tissue characterization reported in the literature are reviewed. To test tissue classifiers, backscatter signals from pieces of fresh, normal, bovine tissue (liver, pancreas and spleen) were used. The instrumentation designed and built for collection of backscatter data is described. A pulse-echo system, with associated analogue spectrum analyzer and digital transient recorder, was interfaced to a minicomputer for system control and data storage. A brief overview of pattern recognition methods is followed by more detailed consideration of feature extraction, and of the optimum combination of features to set up a tissue classifier. The feature sets examined involve local property statistics, spectral features, and second-order amplitude statistics. Parametric methods are used to set up a Bayesian classifier. When these techniques were applied to the bovine-tissue backscatter records, classification success of over 90% was achieved. The most efficient features were those related to the autocorrelation function of the backscatter record. The potential of these techniques for clinical application is discussed.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Physics, general