Files and links (1)
PDFOpen Access (License Unspecified), Open Access
Dissertation
Diversity and compounding for enhanced discrimination of breast masses in ultrasonic B-scan images
Doctor of Philosophy (Ph.D.), Drexel University
May 2002
DOI:
https://doi.org/10.17918/etd-35
Abstract
Ultrasonic tissue characterization aims at improving the ability of ultrasound to classify benign and malignant masses. The akagami distribution was recently proposed to model the statistics of the envelope of the backscattered echo from tissue. Its parameters demonstrated an ability to discriminate benign and malignant masses in breast tissue. However, there is a need to improve the performance of the parameters to reach clinically acceptable standards and also to eliminate any influence of variation in operator gain settings during scanning, time-gain compensation settings, location of the mass, depth and frequency dependent attenuation characteristics and size of the range cell on the parameters while performing the classification. In this research, diversity techniques have been investigated to achieve these goals. Analytical results explaining the improvement in the ability of the parameters to separate different scattering conditions have been derived and tested through computer simulation and experiments on tissue-mimicking phantoms. The technique is then applied to classify in vivo breast masses as benign or malignant. Frequency diversity and compounding are specifically applied to normalize both the parameters of the Nakagami distribution making them insensitive to any of the variations described above. A combination of normalized Nakagami parameters at the site from spatially diverse images of a mass, similar to spatial compounding, is performed to improve the ability to discriminate benign and malignant masses along with a theoretical explanation of why such a combination is likely to improve the classification performance. ROC analysis is undertaken to evaluate the performance of the parameters before and after diversity and compounding. Additional parameters conveying information about the sharpness of the boundary and scattering characteristics at the site are combined with the normalized Nakagami parameter after spatial compounding, improving the area under the ROC curve z A to 0.87 and reaching a sensitivity and specificity of 95% and 70% respectively. The performance of this parameter-based approach requiring minimal clinical intervention exceeded that of the radiologist encouraging its application for automated classification and also as an adjunct to x-ray mammography in reducing unnecessary biopsies.
Metrics
27 File views/ downloads
6 Record Views
Details
- Title
- Diversity and compounding for enhanced discrimination of breast masses in ultrasonic B-scan images
- Creators
- Vishruta Ajitkumar Dumane - DU
- Contributors
- P. Mohana Shankar (Advisor) - Drexel University (1970-)
- Awarding Institution
- Drexel University
- Degree Awarded
- Doctor of Philosophy (Ph.D.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Resource Type
- Dissertation
- Language
- English
- Academic Unit
- College of Engineering (1970-2026); Electrical (and Computer) Engineering [Historical]; Drexel University
- Other Identifier
- 35; 991014632228804721