Open Collections

Digitization and analysis of mammographic images for early detection of breast cancer

University of British Columbia

X-ray mammography is the proven method for early detection of breast cancer. Digital processing and analysis of mammographic images can potentially assist in improved performance of radiologists in earlier detection and recognition of abnormalities. In this work a novel image acquisition system based on an area scanning CCD array has been developed for the digitization of mammograms at high spatial and photometric resolutions. The system characteristic parameters were measured. The quality of the resulting images in terms of sharpness and noise content is comparable with that obtained by the more expensive and slower drum laser-scanning microdensitometer. The clinical application of soft-copy display of digitized images are evaluated. To further improve the quality of the images, restoration algorithms were applied to restore the images from the degrading effects of the system’s blur and noise. Performance of three filtering techniques was compared. A new method for the reduction of boundary truncation artifacts in image restoration was suggested and studied. The process of radiographic image formation was modeled and two locally adaptive smoothing filters were employed to counter signal-dependent radiographic noise before application of restoration filters. The results of the restored images show a marked improvement in detectability of smallest particles of microcalcifications when judged by a human observer. Image segmentation routines were developed to separate microcalcifications from the background parenchymal pattern. Performances of two algorithmic approaches to segmentation and two artificial neural networks were compared. Over 100 numerical features were automatically extracted from the clusters of microcalcifications. These features were evaluated for their ability to separate the benign and malignant formations. Using a database of 68 digitized mammograms a discriminant function was calculated. The sensitivity and specificity of this approach in recognition of malignant microcalcification clusters is shown to be comparable to that of trained radiologists.

Thesis/Dissertation

application/pdf

2009-06-04

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/8738

Electrical and Computer Engineering

University of British Columbia

UBCV

DSpace