Computer Vision and Machine Learning for Digital Pathology

This project aims to develop and validate algorithms and software that can be used to efficiently extract useful information from histopathology images now produced in very large quantities by hospital pathology departments. Microscopic examination of very thin sections of stained tissue by expert histopathologists is used in diagnosis, treatment selection, and in research to help further our understanding of diseases such as cancers. However, there is an ever-growing need for automated methods for quantitative analysis of images of these tissue sections so that the information they contain can be extracted and mined at scale, and ultimately used to improve diagnosis and treatment. The student will develop and apply deep learning and computer vision algorithms to automatically detect and segment structures of interest in histology images so that they can be quantified in a reproducible and high-throughput way. The student will benefit from inhouse
software and algorithms developed in our previous related research, e.g., in breast tumour segmentation [1], gland segmentation in colon cancer [2], and discrimination of dysplastic changes in colorectal polyps [3].
This is an interdisciplinary collaboration between Computer Vision & Image Processing (CVIP) in the School of Science & Engineering, and Pathology at Ninewells Hospital and Medical School. The PhD student will benefit from access to both technical and clinical expertise. The CVIP research group has considerable experience developing novel methods for biomedical image analysis applications and has won several recent international contests in this area. Our graduated PhDs find work in other prime academic research groups (e.g. UCL, Edinburgh, Toronto) and companies (e.g. Toyota, Toshiba, OPTOS plc).