Search papers, labs, and topics across Lattice.
This paper introduces DisMix, an innovative order-aware mixup framework specifically designed for ordinal classification tasks in medical imaging. By utilizing a dual-codebook VQ-VAE to disentangle ordinal and non-ordinal features, DisMix enables meaningful interpolation of ordinal codes while preserving appearance diversity through non-ordinal code variation. The method outperforms six existing mixup baselines across four medical imaging datasets, demonstrating robust performance even in scenarios with limited data and variability in clinical grading.
DisMix achieves superior performance in medical imaging by preserving the integrity of ordinal disease grading while enhancing data diversity through innovative feature disentanglement.
Image mixup is a widely adopted data augmentation strategy, yet it is ill-suited for ordinal classification tasks such as medical disease grading, where labels encode a progression of severity. By indiscriminately blending disease-severity cues (ordinal) with appearance-level variation (non-ordinal), standard mixup produces samples that distort the very ordinal structure that underpins clinical severity grading. We introduce DisMix, an order-aware mixup framework for ordinal classification. DisMix disentangles ordinal and non-ordinal features via a dual-codebook VQ-VAE, allowing each subspace to be mixed independently: ordinal codes are interpolated to produce meaningful intermediate ranks, while non-ordinal codes are varied to introduce appearance diversity without corrupting the ordinal signal. Across four medical imaging datasets, DisMix shows the best aggregate performance among six image mixup baselines paired with six ordinal classifiers and remains effective under data scarcity and clinical grading variability.
