On the current clinical applicability of artificial intelligence in pathology: An evidence-informed clinical...

The integration of artificial intelligence (AI) into pathology practice is anticipated to drive substantial advancements. Numerous AI-based applications have been reported but the extent of their current clinical applicability remains uncertain. An evidence-informed clinical readiness review was conducted using the assistance of ChatGPT 5.1 Plus Thinking (OpenAI, San Francisco, CA). The initial summary provided by the large language model (LLM) was supplemented with additional data collected through further questioning and "conversation" between the authors and the LLM. The results were verified for accuracy and summarized by topic. The LLM provided useful outlines but extensive and time-consuming additional searches by the authors were needed to collect clinically relevant best evidence. Deep learning models (DLMs) for assistive diagnosis of prostate and breast cancer are being implemented in selected laboratories. DLMs for Pap smear assessment, metastasis detection, molecular inference, and other tasks have achieved promising results in curated settings but need further clinical validation. A variety of chatbots have been developed that show promise to enhance communications with patients and improve laboratory logistics. To date DLMs have shown great potential to improve the integration of clinico-pathologic data and can improve the diagnosis of selected entities. DLMs that can diagnose a more substantial number of pathological entities and provide better standardization of results reporting, prospective multicenter validation, and more detailed cost benefit analysis of AI-based models are needed prior to widespread adoption into the daily clinical practice of pathology.