Artificial intelligence is becoming increasingly powerful at reading pathology slides. But a new Perspective in Science Bulletin argues that accuracy alone is not enough. For AI to truly help pathologists, it must work safely and transparently in real clinical settings.
The article, “Toward clinically integrated computational pathology: advancing diagnostic insight and practice transformation,” discusses how computational pathology is moving from technical innovation toward clinical integration. Pathology remains essential for diagnosis and cancer treatment decisions. Yet hospitals are facing rising case numbers, more complex biomarker testing, and growing pressure to deliver faster results. Computational pathology may help by turning whole-slide images into quantitative information for diagnosis, risk assessment, biomarker evaluation, and treatment planning. The authors emphasize that useful AI tools must be interpretable, reportable, and compatible with existing pathology workflows. They should show what was measured, where the evidence came from, how reliable the result is, and whether further testing is needed.
Researchers also highlights the need for real-world validation. Computational pathology systems should be assessed not only by accuracy, but also by their effects on turnaround time, diagnostic consistency, case prioritization, unnecessary testing, clinical decision-making, and implementation costs. Near-term opportunities include biomarker prescreening, confirmatory test prioritization, quantitative scoring, treatment response assessment, and intraoperative support. Looking ahead, the authors suggest that successful systems should be built around clear clinical tasks and remain under pathologist supervision.
The Perspective concludes that AI is unlikely to replace pathologists. Its greater value may lie in helping convert tissue morphology into computable, interpretable, and actionable evidence for better clinical decisions.
Science Bulletin