Pakistan Journal of Pharmaceutical Sciences

Abstract: To assess the effectiveness of a Deep Neural Network (DNN)-based ophthalmic disease diagnosis framework in facilitating personalized medication treatment plans and compare its performance with conventional physician judgment methods. This study employed a prospective, single-center, randomized controlled clinical trial design to treat 500 patients with common ophthalmic diseases. Participants were randomly divided into two groups: the DNN-aided experimental group, which received medication treatment plans generated by the DNN model, and the control group, which received standard physician judgment-based treatment plans. The primary outcome measures included medication selection accuracy, clinical treatment efficacy (measured by BCVA and CMT), patient treatment compliance and adverse reaction management. The DNN-aided treatment plans led to a significant increase in medication selection accuracy and improved treatment quality. The experimental group showed higher BCVA and CMT scores compared to the control group. Additionally, patient compliance in the experimental group was notably higher, indicating that the DNN-generated treatment plans positively influenced patient confidence and adherence to treatment. While there was no statistically significant difference in the rate of adverse reactions between the two groups, the experimental group demonstrated a trend toward lower rates, suggesting that DNN-based treatment plans might reduce treatment-related risks. The DNN-based ophthalmic disease diagnosis model demonstrated its potential to enhance medication selection accuracy, treatment efficacy and patient compliance while reducing adverse reactions. As artificial intelligence technology continues to evolve, DNN models are expected to play an increasingly vital role in individualized ophthalmic disease management, offering more precise and personalized treatment strategies.