Gradient-Based Latent Decomposition Reveals Mechanisms of Feature Degradation in Weakly Supervised Mammography

arXiv:2607.24835v1 Announce Type: new Abstract: Weakly supervised hierarchical models exhibit a persistent asymmetry: coarse lesion-type features are preserved under reconstruction while fine-grained malignancy cues degrade---a pattern with direct consequences for the clinical reliability of breast cancer screening pipelines. We introduce gradient-based orthogonal latent decomposition for hierarchical Variational Autoencoders~(H-VAEs) to mechanistically explain this asymmetry. The latent space is partitioned into a task-aligned component~(z_1), shaped by coarse supervisory gradients, and an orthogonal residual~(z_{ext{res}}) capturing remaining representational capacity. On3,550 mammographic Regions of Interest(ROIs) from CBIS-DDSM, onlysim4.4% of latent magnitude aligns with supervisory gradients, leavingsim95.6% in the orthogonal residual upon which fine-grained pathology prediction primarily depends. The model achieves Stage-1AUC0.866 and Stage 2AUC0.552, with a reconstruction stability gap of Delta_{ext{diag}}=5% (p=0.005) and a classification gap of Delta_{ext{AUC}}=0.314 (p{<}0.001). Latent ablation confirms that features for both tasks reside heavily inz_{ext{res}}, structurally explaining why reconstruction degrades pathology stability disproportionately. Comparisons with Multi-Instance Learning(MIL) and Multi-Task Learning~(MTL) confirm generalization across architectures and modalities. These findings reveal that in high-dimensional spaces, a single coarse supervisory signal isolates only a sparse 1D latent direction, forcing critical fine-grained features into the vulnerable residual subspace.

Source: arXiv cs.CV | 2026-07-29