Jeroen Van Der Laak, Professor of Computational Pathology at Radboud University, unpacked the advantages of using AI in pathology. He argued that AI is not a want but rather a necessity when it comes to digital pathology. Pathology refers to diagnostics based on tissue and pathologists traditionally examined tissue sections through a microscope and discerned the relevant information needed for treatment decisions. Nowadays the tissue sections are more frequently digitised.
AI can be used in the workflow of a pathologist to improve accuracy and efficiency in pathology diagnostics. It is especially useful at detecting small clusters of tumour cells which may be missed by the human eye.
Van der Laak commented that he views tissue sections as some types of biomarkers because they give insights into cancer grading which strongly correlates with prognosis. Traditional breast cancer grading is based on three features (tubule formation, nuclear pleomorphism, mitotic count) but suffers from low reproducibility and arbitrary thresholds. To counter this issue, Van der Laak and his team are redefining grading thresholds with AI for different breast cancer subtypes to improve prognostic value and identify patients who may benefit from tailored therapies more easily.
AI can help overcome the inherent variability between pathologists by providing objective, reproducible measurements. The AI tools that Van Der Laak is developing are continuous quantitative measures for tubule formation scoring, nuclear pleomorphism and mitosis counting.
Large-scale data projects like Big Picture are being developed to support AI research in pathology by providing millions of digital slides and metadata. However, there are regulatory and ethical considerations that must be taken into account. 
Overall, Van Der Laak gave a persuasive case for integrating AI into pathology workflows. AI enhances diagnostic accuracy while providing a more nuanced and personalised approach to cancer grading and treatment selection.
