Texture-based quantitative characterization and analysis of colonoscopic images

The importance of computer-assisted diagnosis in colonoscopy is to assist the physician in detecting the colon status by characterizing the features from the colonoscopic image. A novel approach is developed to extract new texture-based quantitative features from the texture spectra in the chromatic and achromatic domains. The texture spectra are obtained from the texture unit numbers, which contain local and global texture information of the image. These features are fed into a supervisory backpropagation neural network (BPNN) and classification is performed. The results are compared with those obtained using non-supervisory neural networks, namely, probabilistic neural network (PNN), learning vector quantization (LVQ), and self organizing method (SOM). The performance of the neural networks for classifying the colon status is compared. The preliminary results obtained by the proposed approach support the feasibility of the technique.