Developing an AI based approach to histological tissue type classification

Histological tissue type classification is a profound research topic. However, most of the research in this area is confined to either to differentiate cancerous tissue from non-cancerous tissue or to classify connective, epithelial, muscleand nervous tissue rather than classifying an organ specific tissue from another. Many more artificial intelligence algorithms were developed to accelerate the research in former type of classification rather than latter.Prostate cancer is one of the most common types of cancer. This usually grows slowly and may take years to grow. Various algorithms have already been developed to grade this cancer. The very first step in this pipeline is to verify if the biopsy sample is a prostate tissue and this master thesis focuses on this classification. This research aims to explore, examine, develop and evaluate different state-of-the-art machine learning algorithms to classify a prostate tissue from non-prostate tissue.Through the experiments, this thesis shows that a deep convolutional neural networks work with stable performance and have been generalized with relatively small dataset. The evaluation metrics that were employed to evaluate different models were area under the curve (AUC) and F1 score