Patient safety in AI-powered diagnostic pathology

Abstract

Artificial intelligence (AI)-powered diagnostic pathology involves combining traditional histological techniques with computer-assisted AI technology. This process comprises several key steps: generating whole slide digital images; annotating and training algorithms; constructing robust datasets; testing and monitoring consistency with clinical expectations; validating results externally and overseeing the output of algorithms. All of these steps must adhere to quality standards and ensure patient safety.Current scientific evidence suggests that, while AI can enhance the accuracy of human diagnostics, it cannot replace humans as autonomous classifiers. Generative intelligence offers new, promising technological advancements. When applying these technologies in clinical practice, international healthcare institutions recommend clearly defining the application domains and implementing and monitoring safety measures.This critical review of current AI applications in diagnostic pathology underscores the paramount significance of patient-centred safety considerations. It also highlights the necessity of collaborative efforts among governments, academic institutions, international healthcare agencies, scientific societies, patient associations and algorithm developers to implement safety-oriented regulatory measures for AI-powered pathology.

Ähnliche Arbeiten