文摘
Image enhancement technologies, such as chromoendoscopy and digital chromoendoscopy were reported to facilitate the detection and diagnosis of colonic polyps during endoscopic sessions. Here, we investigate the impact of enhanced imaging technologies on the classification accuracy of computer-aided diagnosis systems. Specifically, we determine if image representations obtained from different imaging modalities are significantly different and experimentation is performed to figure out the impact of utilizing differing imaging modalities in the training and validation sets. Finally, we examine if merging the images of similar imaging modalities for training the classification model can be effectively applied to improve the accuracy.