A General Bézier Tree Encoding Counterfactual Framework for Retinal-Vessel-Mediated Disease Analysis

2026-05-13

AI Summary

This work addresses the limitation of existing generative counterfactual methods, which intervene only at the image-level disease label and fail to explicitly disentangle the anatomical structure of retinal vasculature. The authors model the vascular network as interconnected cubic Bézier curves to construct a disease-agnostic structural representation, enabling parameterized causal interventions on geometric attributes—such as tortuosity and vessel caliber—within a diffusion-based generative framework. This approach achieves, for the first time, atomically perturbable encoding of vascular topology, ensuring causal isolation between geometric features and pixel-level confounders while preserving the fundus background. In cohorts spanning diabetic retinopathy, ischemic stroke, and Alzheimer’s disease, counterfactual interventions elicit dose-responsive shifts in classifier predictions, outperforming pixel-ablation baselines by over an order of magnitude in response strength and effectively ruling out out-of-distribution artifacts, thereby supporting unified cross-disease causal validation.

Abstract

The geometry of the retinal vessel is a key biomarker of vascular diseases, yet clinical evidence remains primarily observational. Existing generative counterfactuals intervene only at the image-level disease label, failing to isolate explicit anatomical structure. To address this limitation, we propose the Bézier Tree Encoding Counterfactual Framework (BTECF). By abstracting vascular networks into interconnected cubic-Bézier segments, BTECF establishes a disease-agnostic representation in which structural topology is explicitly preserved and atomically perturbable. Coupling this encoding with a diffusion-based generator enables parameter-level do-interventions on explicit geometric axes (e.g., tortuosity, caliber) while preserving background fundus textures. We validate BTECF on diabetic retinopathy, together with independent cohorts for ischemic stroke and Alzheimer's disease. Isolated counterfactual interventions produce dose-responsive shifts in classifier predictions; a matched pixel-drop control attenuates this response by an order of magnitude or more, ruling out out-of-distribution generation artifacts. By enforcing causal isolation between vessel topology and pixel-level confounders, BTECF provides a unified generative paradigm for hypothesis verification across systemic diseases. To support reproducibility, the code will be publicly released upon acceptance.

Problem

• retinal vessel
• counterfactual
• geometric structure
• causal analysis
• vascular diseases

Innovation

• Bézier Tree Encoding
• counterfactual generation
• retinal vessel geometry
• causal intervention
• diffusion-based generator