Scholarlink

<p>Diabetic retinopathy (DR) where blood sugar levels damage the vessels feeding blood to the tissue behind the retina remains a progressive complication disease. To create the possibility of detection and prevention of this condition, we propose a predictive and preventive multidimensional framework for prognosis of DR by constructing a mechanistic model of patient-specific physiological deficiencies and projecting their long-term retinal consequences. The framework mimics the deficiencies in a person’s body (like dysregulated glucose control, vascular fragility, and insulin resistance) as actionable embedded feature vectors via physiological simulation layer. These vectors then become the foundation for continuous-time disease acceleration modeling via Neural ordinary differential equations layer. In the final segment of the framework, Capsule Networks with dynamic routing layer analyses interventions by preserving hierarchical cause–effect relations enabling the system to map therapeutic adjustments into measurable retinal outcomes. Evaluated on UK Biobank, EyePACS, and Messidor public datasets, it reduces missed early cases by 2.4×, achieves 1.26× stronger trajectory alignments, and delivers 1.48× more faithful intervention-response mapping.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

PHYS-ODE-CAPS: a physiological simulative multi-dimensional framework for prognosis of retinal microvascular damage in diabetic retinopathy (DR)

  • 期刊论文

  • 发表日期：

    2025年12月27日

查看全文
  • Saumendu Bose,
  • Vandana Dixit Kaushik

摘要

Diabetic retinopathy (DR) where blood sugar levels damage the vessels feeding blood to the tissue behind the retina remains a progressive complication disease. To create the possibility of detection and prevention of this condition, we propose a predictive and preventive multidimensional framework for prognosis of DR by constructing a mechanistic model of patient-specific physiological deficiencies and projecting their long-term retinal consequences. The framework mimics the deficiencies in a person’s body (like dysregulated glucose control, vascular fragility, and insulin resistance) as actionable embedded feature vectors via physiological simulation layer. These vectors then become the foundation for continuous-time disease acceleration modeling via Neural ordinary differential equations layer. In the final segment of the framework, Capsule Networks with dynamic routing layer analyses interventions by preserving hierarchical cause–effect relations enabling the system to map therapeutic adjustments into measurable retinal outcomes. Evaluated on UK Biobank, EyePACS, and Messidor public datasets, it reduces missed early cases by 2.4×, achieves 1.26× stronger trajectory alignments, and delivers 1.48× more faithful intervention-response mapping.