การเปรียบเทียบระบบคัดกรองภาวะเบาหวานขึ้นจอประสาทตาด้วยปัญญาประดิษฐ์ (DeepEye) และการคัดกรองแบบดั้งเดิมในด้านระยะเวลาการคัดกรองและความแม่นยำในการวินิจฉัย ณ โรงพยาบาลเจ้าพระยายมราช จังหวัดสุพรรณบุรี: การศึกษาแบบสุ่มและมีกลุ่มควบคุม
Keywords:
Artificial Intelligence, Diabetic Retinopathy, DeepEye, Screening Time, Diagnostic Accuracy, ปัญญาประดิษฐ์, เบาหวานขึ้นจอประสาทตา, ระยะเวลาคัดกรอง, ความแม่นยำในการวินิจฉัยAbstract
Background: Diabetic retinopathy (DR) is a major cause of global vision loss. While conventional screening often faces personnel shortages and high workloads, integrating artificial intelligence (AI) has emerged as a crucial strategy to enhance healthcare efficiency.
Objective: To compare the screening time and diagnostic accuracy between an AI-based system (DeepEye) and conventional screening in a real-world clinical setting.
Methods: A randomized controlled trial (RCT) was conducted with 212 diabetic patients at Chaophraya Yommarat Hospital. Participants were assigned to either an AI screening group or a conventional group (n=106 each), using ophthalmologist assessment as the gold standard.
Results: The AI system demonstrated significantly higher sensitivity at 96.88%, compared to 58.33% in the conventional group. Conversely, the conventional group exhibited higher specificity (92.68% vs. 63.51%). Regarding screening time in non-mydriatic cases, the AI group had a longer mean duration than the conventional group (74.24 vs. 60.34 seconds, p = 0.001), primarily due to manual data handling processes.
Conclusion: The AI system (DeepEye) proved highly effective as a screening tool with superior sensitivity, reducing the risk of missing referable DR cases. Despite slight increases in technical processing time, AI functions as a robust workflow optimization tool that alleviates the diagnostic burden on healthcare personnel.
