It is becoming increasingly difficult to escape conversations about artificial intelligence (AI), particularly debates over how useful and effective the technology really is.
Science and healthcare are among the fields where AI could have a major impact. The technology already helps researchers analyze medical scans, develop drugs, and predict future health outcomes.
AI Takes on a Bigger Role in Eye Care
Now, researchers led by the Beijing Visual Science and Translational Eye Research Institute (BERI) in China have taken the use of AI a step further by allowing it to help run an eye clinic under human supervision.
Researchers designed the AI-TEC (AI-Agent Augmented Tsinghua Eye Clinic) specifically around AI instead of simply adding AI tools to traditional medical systems. Human doctors remained involved, but the facility uses AI extensively throughout the patient-care process.
According to a study published in Nature Medicine, AI supported almost every stage of care, from the initial pre-consultation and examination to patient follow-up. This included analyzing the eye scans that play a crucial role in diagnosis and treatment.
The researchers say their work offers early insights into how AI can move beyond simply improving algorithms and instead transform clinical workflows and integrate across entire healthcare systems, where its real-world value can ultimately be achieved.
AI Initially Struggles to Detect Eye Diseases Accurately
The AI-TEC trial produced several notable findings. At first, the AI had a relatively low accuracy rate when detecting eye diseases such as glaucoma and age-related macular degeneration from scans.
Its performance improved considerably after expert ophthalmologists provided 1,426 high-quality eye scans that had been accurately labeled according to the conditions they showed.
Interestingly, this new dataset proved more effective than the AI’s original training set, which contained nearly 27,000 lower-quality images with less detailed labeling.
The AI-TEC trial produced several notable findings. At first, the AI had a relatively low accuracy rate when detecting eye diseases such as glaucoma and age-related macular degeneration from scans.
Its performance improved considerably after expert ophthalmologists provided 1,426 high-quality eye scans that had been accurately labeled according to the conditions they showed.
Interestingly, this new dataset proved more effective than the AI’s original training set, which contained nearly 27,000 lower-quality images with less detailed labeling.
The use of the AI system increased again the following month, reaching 259 out of 1,126 examinations, or about 23 percent. The researchers attributed the rise to improvements that made the system faster and easier for staff to operate, including fewer clicks and less manual data entry.
The researchers noted that their early experience showed that introducing AI-based healthcare is not simply a matter of developing a powerful algorithm. Instead, it requires an entire ecosystem involving high-quality data, smooth workflow integration, clinician participation, proper oversight and monitoring, and clear evidence of clinical benefits.
Rapid Clinical Feedback Helps AI Improve Faster
They also emphasized that feedback from healthcare professionals needs to reach the AI system quickly. Delays of several weeks can limit the system’s ability to learn and improve effectively.
Overall, the researchers identified three key lessons: high-quality data, efficient system operation, and frequent, rapid feedback from clinicians.
The researchers stress that close collaboration between doctors and AI specialists is essential when developing and implementing AI-driven healthcare systems
Although AI-first clinics are still in their early stages, this small-scale trial suggests that they can work effectively when the necessary conditions are in place.
However, significant challenges remain. One major issue is the difference in how AI and doctors approach diagnosis. AI typically focuses on identifying the final result, such as whether a patient has an eye disease, while doctors usually begin by considering the patient’s symptoms, such as blurred vision.
Therefore, an AI system that performs highly accurately when analyzing medical scans does not automatically guarantee better patient care. The researchers suggest that future studies should look beyond AI performance scores and place greater emphasis on how the technology actually improves clinical outcomes and healthcare delivery.
Read the original article on: sciencealert
Read more:AI-powered smart patch could transform the future of health monitoring
