Automated classification of colorectal adenoma-dysplasia-carcinoma in tissue biopsies using artificial intelligence based image evaluation

Background: Segmentation and rule based image analysis techniques can work well on standardized specimen. In case of atypical alterations with specimen inequalities these methods can fail in detecting colorectal adenoma, carcinoma. In recent years, new artificial neural networks based image evaluation techniques raised, which should work with a higher flexibility towards the real-life histological sample, quality issues. Aims of the study: Develop a tissue component segmentation algorithm (epithelia, lamina propria, muscles, connective tissue, fat) based on pixel derived parameters using artificial neural networ…