ModelMix: A New Model-Mixup Strategy to Minimize Vicinal Risk across Tasks for Few-scribble based Cardiac Segmentation

TL;DR

ModelMix introduces a novel parameter interpolation strategy for few-scribble cardiac segmentation, effectively reducing vicinal risk across tasks and outperforming existing weakly supervised methods on multiple datasets.

Contribution

The paper proposes ModelMix, a new model-mixup approach that interpolates convolutional parameters to improve weakly supervised segmentation with scribbles.

Findings

Significantly outperforms state-of-the-art scribble supervision methods.

Effective across multiple open cardiac segmentation datasets.

Reduces resource requirements for dense pixel-level labeling.

Abstract

Pixel-level dense labeling is both resource-intensive and time-consuming, whereas weak labels such as scribble present a more feasible alternative to full annotations. However, training segmentation networks with weak supervision from scribbles remains challenging. Inspired by the fact that different segmentation tasks can be correlated with each other, we introduce a new approach to few-scribble supervised segmentation based on model parameter interpolation, termed as ModelMix. Leveraging the prior knowledge that linearly interpolating convolution kernels and bias terms should result in linear interpolations of the corresponding feature vectors, ModelMix constructs virtual models using convex combinations of convolutional parameters from separate encoders. We then regularize the model set to minimize vicinal risk across tasks in both unsupervised and scribble-supervised way. Validated on three open datasets, i.e., ACDC, MSCMRseg, and MyoPS, our few-scribble guided ModelMix significantly surpasses the performance of the state-of-the-art scribble supervised methods.