While the digitisation of cellular pathology will lead to workflow improvements and more quantitative reporting, it opens new and exciting opportunities for integrating genetic and molecular information with traditional pathology. The ability to associate the expression of bespoke biomolecules with cell and tissue architecture patterns holds the promise of advancing clinical decision making.

With our development of imCMS, an image-based predictor of consensus-based molecular subtypes of colorectal cancer, we have provided one example that illustrates the concept of learning morpho-molecular correlations. This deep learning algorithm provides a cost-effective tool to associate complex features of tissue organisation with molecular and outcome data and to resolve unclassifiable or heterogeneous cases. In addition, I will provide additional examples that demonstrate the ability to predict genetic traits from morphological features and recent results of predicting response to radiotherapy using morphological features.

AI will shape the role of pathology in modern multidisciplinary oncology. More standardised AI augmented pathology reports hold the promise of changing the communication between pathologists and oncologists. In addition, there is the potential of providing more accessible information to patients.