Automatic small bowel tumor diagnosis by using multi-scale wavelet-based analysis in wireless capsule endoscopy images

The set of features proposed in this paper to code textural information is based on statistical modeling of second order textural measures extracted from co-occurrence matrices. To cope with both joint and marginal non-Gaussianity of second order textural measures, higher order moments are used. These statistical moments are taken from the two-dimensional color-scale feature space, where two different scales are considered. Second and higher order moments of textural measures are computed from the co-occurrence matrices computed from images synthesized by the inverse wavelet transform of the wavelet transform containing only the selected scales for the three color channels. The dimensionality of the data is reduced by using Principal Component Analysis.The proposed textural features are then used as the input of a classifier based on artificial neural networks. Classification performances of 93.1% specificity and 93.9% sensitivity are achieved on real data. These promising results open the path towards a deeper study regarding the applicability of this algorithm in computer aided diagnosis systems to assist physicians in their clinical practice.The innovation of wireless capsule endoscopy (CE) has revolutionized the investigation and management of patients with suspected small bowel disease [1]. Since its introduction, in the year 2000, a new chapter in the small bowel examination was opened, as this new technology allows the visualization of the entire gastrointestinal (GI) tract, reaching places where conventional endoscopy is unable to. In fact, conventional endoscopy presents some important limitations in the diagnosis of small bowel problems, since it is limited to the upper GI tract, at the duodenum, and to lower GI tract, at terminal ileum. Therefore, prior to the wireless capsule endoscopy era, the small intestine was the conventional endoscopy's last frontier, because it could not be internally visualized directly or in it's entirely by any method [2]. The