An Expert System for Diagnosis of Cancer Diseases

Traditional ways of diagnosing patients can take a lot of time and are sometimes undependable thus computerized technologies could be employed for more accurate and faster analysis. The third most diagnosed cancer in the world is a cancer called colorectal cancer (CRC). CRC results from unrestrained cell growth in sections of the large intestine, particularly the colon or rectum. Early diagnosis of CRC is important for the survival of the patient since it can result from changes in lifestyle and increase in age of the patient. The MATLAB Image Processing toolbox together with the QUPATH software was utilized for image preprocessing, segmentation and feature extraction. Image were extracted for fine tuning the Visual Geometry Group-16 (VGG16) Convolutional Neural Network (CNN), which was pretrained on the ImageNet database thus enabling it to learn domain specific features necessary to classify the whole slide images. The concluding evaluation notes that the ground truth annotations are not ideal and thus the ability of deep learning to overcome issues in the quality of the data is verified. While due to the nature of the domain, deep learning techniques may never be suited to replace the expertise of practicing pathologists, they promise to aid in tasks which can be tedious, painstaking and subject to a degree of error and subjectivity."