Abstract
The focus of this dissertation is the feature extraction of ultrasonic liver images. The feature extraction is essential in a computer vision system for diagnosis of medical images. We propose a fractal feature vector based on M-band wavelet transform to classify ultrasonic liver images—normal liver, cirrhosis, and hepatoma. The proposed feature extraction algorithm is based on the spatial-frequency decomposition and fractal geometry. And, various classification algorithms based on respective texture measurements and filter banks are presented and tested. Classifications for the three sets of ultrasonic liver images reveal that the fractal feature vector based on M-band wavelet transform is trustworthy. A hierarchical classifier, which is based on the proposed feature extraction algorithm is at least 96.7% accurate in the distinction between normal and abnormal liver images and is at least 93.6% accurate in the distinction between cirrhosis and hepatoma liver images. Additionally, the criterion for feature selection is specified and employed for performance comparisons herein. In supervised classification, we also propose a modified computation of fractal dimension since the estimation of the fractal dimension is crucial in fractal geometry. The adopted estimation approach is based on box-counting. However, the scheme, which is easily disturbed by noise, produced many non-negligible plateaus that cause underestimate. A more robust and efficient computation of the fractal dimension is verified from experimental results.Finally, we applied the proposed multiresolution fractal feature vector to segment suspicious abnormal regions of ultrasonic liver images. Segmentation of various liver diseases reveals that the fractal feature vector based on multiresolution analysis is reliable. A quantitative characterization based on the proposed unsupervised segmentation algorithm can be utilized to establish an automatic computer-aided diagnostic system. As well, to increase the visual interpretation capability of ultrasonic liver image for junior physicians, an off-line learning system can be developed to investigate the visual criteria.