Computer Assisted Selection of Serrated Adenomas and Neoplastic Polyps - a New Clinical Draft

The aim of the study is to develop a computer program which is able to classify different entities of colorectal polyps on the basis of optical polyp features. In the end, the computer program shall differentiate between (i) hypeplastic polyps, (ii) adenomas and (iii) serrated adenomas .

In the first phase of the study a computer program will be established which aims to distinguish between the above mentions entities on the basis of optical features derived from still images. A machine learning apporach will be used for creating the program. Afterwards, in a second phase of the study, still images of 100 polyps (not used in the first phase) will be presented to the computer program. Quality of the computer program will be tested by calculating the accuracy for differentiating the three different polyp types. The gold standard for true polyp diagnoses will be based on histopathological diagnoses of the polyps. The same pictures of 100 polyps will also be presented to human experts. Experts will also predict histopathological diagnoses on the basis of optical polyps featurs. Accuracy of computer-decisions and human expert predictions will be compared. The establishment of a well- functioning computer program is the primary aim of the study.