Grant: Collaborative Health Research Projects (NSERC partnered)

The evaluation of biopsy samples from tumours has been used for decades by pathologists to assist in the diagnosis, treatment planning and prognosis estimation for cancer patients. In order to improve outcomes for patients with cancer we need to develop improved and more accurate methods for histological analyses. We propose the development of techniques to enable three-dimensional (3D) histopathology of cancer biopsies powered by machine learning-based image analysis for prognosis prediction. We will develop a method to process human lung tumour biopsies that enable detailed sub-cellular 3D imaging, develop machine learning based algorithms to segment and quantify biological features from the 3D images, and use the features measured from the 3D images to train a model to predict patient outcomes (overall survival, treatment efficacy) from the information measured from the 3D imaging. This approach will allow us to measure key and new information about a patient's tumour that cannot be accurately measured using current methods. We have built a team of key interdisciplinary academic and industrial collaborators (Drs. Chan, Martel, Tsao, and Hwang, and Google Canada and Huron Digital Pathology) with expertise in imaging, machine learning, cancer pathology and bioengineering. Through these partnerships we have the ability to translate the technology generated into commercial products for the analysis of 3D histopathology images and tools to assist in predicting patient outcomes from the large amount of patient data generated from 3D imaging. Finally, we have developed a training plan that teaches researchers from different academic backgrounds to work together to solve pressing clinical issues and to translate the technological developments.