Review of computer vision algorithms in histopathology to decipher complex features in cancer

Artificial intelligence (AI) techniques are increasingly designed to augment human capabilities, offering significant improvements in speed and accuracy across various applications. While AI has demonstrated high precision in solving fundamental tasks like common object recognition and processing large datasets to extract features and make decisions, it continues to face substantial challenges in domain-specific fields. The assessment of histopathological data for precision medicine in tumor analysis is one such complex task. It requires a deeper understanding of biomedical principles and their integration into AI-driven solutions. This review captures the hidden and relevant aspects of the cancer ecosystem within whole slide images (WSIs) and the current challenges associated with quantifying tumor heterogeneity and its critical influence on the development of robust AI solutions. Several cancer types (lung, breast, colorectal, and prostate cancers) are discussed, including an overview of freely available datasets and software for data processing, with regard to the suitability of computer vision models for specific medical diagnostic tasks. Finally, we delineate how emerging trends in AI can be synergistically integrated with specialized medical knowledge to provide new perspectives for trustworthy, AI-based digital personalized medicine in histopathology.