The increasing prevalence of corneal diseases, including infectious keratitis and immunological keratitis, poses significant public health challenges worldwide. Amidst the rise of artificial intelligence (AI) technologies, the integration of AI assistance has been proposed to help ophthalmologists diagnose these conditions more accurately. A recent study has shed light on how clinical experience affects the responses of ophthalmologists to misleading AI diagnostic guidance, underscoring the importance of human expertise even in the age of advanced technology.
Corneal blindness ranks among the leading causes of visual impairment globally, affecting all age groups. It is often categorized as ‘preventable blindness’ due to the potential for early diagnosis and treatment to avert severe vision loss. Enter CorneAI, the AI classification tool developed to categorize various corneal conditions using slit-lamp microscopy. While initially this tool can improve diagnostic accuracy, it also raises concerns about the reliability of AI outputs, particularly when they are misleading.
This cross-sectional study, conducted across multiple institutions affiliated with the Japan Cornea Society, involved 23 ophthalmologists—7 corneal specialists, 7 board-certified non-corneal-specialists, and 9 residents. Researchers focused on evaluating how AI's misguidance impacted their diagnostic accuracy concerning infectious and immunological keratitis, using 60 slit-lamp images categorically classified through the AI system. Modifications were made to the AI outputs, presenting correct classifications for 70% of the cases and incorrect for 30%, to deliberately test the ophthalmologists' responses.
The findings reveal important insights: the overall diagnostic accuracy of the ophthalmologists remained relatively stable whether or not they utilized AI assistance (75.2% versus 75.9%, P = 0.59). While the corneal specialists maintained their accuracy when confronted with incorrect AI outputs (60.3% versus 53.2%, P = 0.11), both board-certified specialists and residents experienced significant drops—54.5% to 31.6% (P < 0.001)—demonstrATING their vulnerability to misleading AI suggestions.
According to the authors of the article, "Even with the introduction of AI diagnostic support systems, the importance of ophthalmologist’s experience remains crucials." The research indicates the potential negative consequences of over-reliance on AI technologies, especially among less experienced clinicians who may accept AI outputs without adequate scrutiny. Resident physicians, for example, were particularly susceptible to the misguiding effects of AI recommendations. The study emphasizes the need for ophthalmologists to remain vigilant and retain their diagnostic acumen when utilizing AI.
While AI-developed tools like CorneAI offer transformative potential to support diagnostic accuracy with reported effectiveness exceeding 85% under optimal conditions, challenges remain. This study’s results suggest caution, urging continued emphasis on medical education and training to nurture diagnostic skills among younger practitioners. With the rapid advancements of AI technologies, it is imperative for future healthcare to harmonize AI interventions with human judgment for optimal patient outcomes.
To conclude, this study reveals the complex interplay between clinical experience and AI-guided diagnostics, asserting the necessity for ophthalmologists to cultivate their expertise. The research also highlights AI's dual capacity to support and mislead medical professionals, underlining the continued relevance of skilled observation and hands-on experience. With careful integration, AI systems like CorneAI have promising applications in improving patient care and preventing blindness caused by corneal diseases.