In Pathology, AI Won’t Replace Doctors—But Doctors Who Ignore It May Fall Behind

Dr. Cheng Chee Leong, head of the department of anatomical pathology at Singapore General Hospital, explains how artificial intelligence is becoming essential in helping pathologists manage the growing complexity and volume of patient cases. As Singapore’s population ages, patients are presenting with multiple, overlapping conditions, making diagnoses more challenging. This shift has increased the workload significantly—what once required analyzing four parameters in a prostate biopsy now demands evaluation of many more, with some cases involving 20 to 30 tissue specimens, a 10 to 20-fold increase in effort. With a persistent shortage of pathologists, the traditional model of relying solely on human labor is no longer sustainable. This is where AI steps in—not to replace doctors, but to enhance their capabilities. AI tools can quickly scan digital slides and highlight areas of interest, improving both speed and diagnostic confidence. They help pathologists focus on the most critical parts of a case, reducing fatigue and the risk of oversight. Dr. Cheng has been involved in medical informatics for over two decades, including a project from 2020 to 2021 with AI Singapore to develop algorithms that distinguish between two similar breast lesions—fibroadenomas and phyllodes tumors. The goal was to improve diagnostic accuracy and guide treatment decisions, demonstrating that AI has been part of pathology innovation for years. However, AI is not perfect. It relies heavily on the quality and consistency of training data. If tissue samples are processed differently—such as those from overseas labs with varying staining techniques or coloration—AI performance can degrade. It may misinterpret folded tissue as abnormal or struggle with unusual patterns that don’t fit its training data. Unlike human experts, AI lacks the adaptability to handle novel or ambiguous cases creatively. Human judgment remains crucial. Pathologists integrate findings from slides with clinical data, such as electronic health records and radiology reports, to form a complete picture. They can recognize when AI is wrong and adjust accordingly. The ideal model is a “human-in-the-loop” approach, where AI supports, but does not override, clinical expertise. Dr. Cheng believes that while AI will not replace doctors in the foreseeable future, those who don’t use AI will fall behind. The future of pathology demands new skills—understanding AI limitations, interpreting its outputs critically, and working collaboratively with intelligent systems. Over time, as AI is trained on more diverse and high-quality data, it may eventually match or surpass human performance in certain tasks. But for now, the partnership between human expertise and artificial intelligence is the key to delivering accurate, timely diagnoses in an increasingly complex healthcare landscape.