Quantitative characterization and neural network-based evaluation of colonoscopic images

Extracting features from the colonoscopic images is essential for getting the quantitative parameters, which characterizes the properties of the colon. The features are employed in the computer-assisted diagnosis of colonoscopic images to assist the physician in detecting the colon status. Present methods mostly use manual approaches. A novel scheme is developed to extract new texture-based quantitative features from the texture spectra in the chromatic and achromatic domains of colonoscopic images. The texture spectra are obtained from the texture unit numbers, which contain local and global texture information of the image. These features are evaluated using supervisory Backpropagation Neural Network (BPNN) with various training algorithms, viz., resilient propagation (RPROP), scaled conjugate gradient (SCG), and Marquardt algorithms. The evaluation is based on training time, training epoch, and accuracy on classifying the colon status. The preliminary results obtained by the proposed approach support the feasibility of the technique.