Less is More in Semantic Space: Intrinsic Decoupling via Clifford-M for Fundus Image Classification

TL;DR

This paper introduces Clifford-M, a lightweight geometric model that efficiently captures multi-scale structures in fundus images, outperforming larger CNNs without explicit frequency decomposition.

Contribution

Proposes Clifford-M, a novel geometric interaction backbone that replaces frequency-splitting modules, enabling efficient multi-scale feature fusion in fundus image classification.

Findings

Clifford-M achieves 0.8142 mean AUC-ROC on ODIR-5K without pre-training.

It attains 0.7425 macro AUC on RFMiD without fine-tuning.

Outperforms larger CNN baselines with fewer parameters.

Abstract

Multi-label fundus diagnosis requires features that capture both fine-grained lesions and large-scale retinal structure. Many multi-scale medical vision models address this challenge through explicit frequency decomposition, but our ablation studies show that such heuristics provide limited benefit in this setting: replacing the proposed simple dual-resolution stem with Octave Convolution increased parameters by 35% and computation by a 2.23-fold increase in computation; without improving mean accuracy, while a fixed wavelet-based variant performed substantially worse. Motivated by these findings, we propose Clifford-M, a lightweight backbone that replaces both feed-forward expansion and frequency-splitting modules with sparse geometric interaction. The model is built on a Clifford-style rolling product that jointly captures alignment and structural variation with linear complexity, enabling efficient cross-scale fusion and self-refinement in a compact dual-resolution architecture. Without pre-training, Clifford-M achieves a mean AUC-ROC of 0.8142 and a mean macro-F1 (optimal threshold) of 0.5481 on ODIR-5K using only 0.85M parameters, outperforming substantially larger mid-scale CNN baselines under the same training protocol. When evaluated on RFMiD without fine-tuning, it attains 0.7425 +/- 0.0198 macro AUC and 0.7610 +/- 0.0344 micro AUC, indicating reasonable robustness to cross-dataset shift. These results suggest that competitive and efficient fundus diagnosis can be achieved without explicit frequency engineering, provided that the core feature interaction is designed to capture multi-scale structure directly.