If you have a biopsy for suspected breast cancer, a pathologist views your tumour sample through a microscope, to see what type of tumour you have, how aggressive it is, and other important features. The process is time-consuming, sometimes subjective and can result in errors. Pathologists agree that new, digital “whole slide imaging” systems will make interpreting biopsies much faster and more reliable. The challenge is to develop the right algorithms to make these expert systems as accurate as possible.
Last year, Associate Professor Ramakrishnan Mukundan and his team from the School of Computer Science at Canterbury University came 2nd out of 104 entries in a global competition to develop machine-learning algorithms for the pathology of HER2+ breast cancer. We’ve awarded him $100,000 to refine and improve his algorithm and then to develop an automated image analysis system that will improve the accuracy and efficiency of pathology in NZ.