Abstract
Pathologists are quickly reaching a state where they are able to leverage advances in computation, digital pathology, big data analytics and deep learning. When coupled with artificial intelligence (AI) tools this creates many opportunities to improve clinical practice and patient outcomes. There is tremendous potential for AI in all fields of pathology that extend from the womb to the tomb. AI can triage cases, enhance routine tasks, better diagnose disease, provide prognosis and make new discoveries. Not surprisingly, several vendors have shifted towards developing AI tools and we have witnessed a spurt of AI start-up companies targeting pathology. Pathologists are important at all stages of an AI tool creation including algorithm design, development, validation, regulatory approval and integration. Academic industry partnerships have consequently becoming key in the innovation cycle in AI. However, only a handful of labs around the world have started to deploy AI technology. There are four important questions we need to consider before adopting AI in pathology. These include: (1) What are the appropriate tasks for AI? (2) What are the right data for developing AI tools? (3) What is the right evidence standard for accepting AI? (4) What is the right method for deploying AI in routine practice? The risks of prematurely using AI include poor algorithm performance and misdiagnoses due to biased algorithms, failed implementations with unintended consequences, workflow disruption, driving up health care cost, liability concerns, as well as deskilled pathologists and trainees. Guidelines will help pathologists facilitate how to best use AI technology in clinical practice.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
