PUMCH Ophthalmology Team: "KongMing Model" Predicts Prognosis of Anti-VEGF Therapy for nAMD, Published in Lancet Digital Health

PUMCH Ophthalmology Team: "KongMing Model" Predicts Prognosis of Anti-VEGF Therapy for nAMD, Published in Lancet Digital Health

🚨 Clinical Dilemma and Challenges

The prognosis of anti-VEGF therapy for neovascular age-related macular degeneration (nAMD) exhibits significant individual heterogeneity.
Objectively predicting therapeutic efficacy prior to treatment remains a core challenge in formulating individualized intervention strategies.

💡 Innovative Multi-Task Deep Learning Architecture
Prof. Youxin Chen's team at Peking Union Medical College Hospital (PUMCH), in collaboration with Prof. Bin Sheng's team at Shanghai Jiao Tong University, developed the "KongMing" deep learning system based on data from 18 hospitals across 12 provinces.

As illustrated in the attached flowchart, this lesion-aware Transformer architecture achieves triple predictions: visual function change classification, visual acuity value regression, and post-treatment OCT image generation.

⚡ Objective Model Performance Validation
Referencing the attached ROC curves and scatter plots, the KongMing model demonstrated excellent performance in both internal and external validations across three key timepoints: post-single injection, post-three loading injections, and 1-year post-3+PRN.

Its accuracy in predicting visual acuity not only outperformed ophthalmologists but also precisely identified poor prognostic biomarkers such as intraretinal fluid.

📈 Clinical Translation and Application Value
This model provides an objective, digital decision-support tool for the individualized treatment of nAMD.

It assists clinicians in early identification of poor responders, enabling timely regimen adjustments and optimizing the allocation of medical resources.