SigVLP: Sigmoid Volume-Language Pre-Training for Self-Supervised CT-Volume Adaptive Representation Learning

TL;DR

SigVLP introduces a novel volume-language pre-training method for medical imaging that uses rotary position embeddings to handle variable-sized CT volumes, enhancing alignment and performance on downstream tasks.

Contribution

The paper proposes a new approach using Rotary Position Embeddings for flexible, size-agnostic volumetric representation learning in medical imaging, enabling better text-volume alignment.

Findings

Improved zero-shot classification accuracy

Enhanced segmentation and retrieval performance

Robustness to variable input sizes

Abstract

Large-scale, volumetric medical imaging datasets typically aggregate scans from different vendors and devices, resulting in highly variable resolution, slice thicknesses, and numbers of slices per study. Consequently, training representation models usually requires cropping or interpolating along the z-axis to obtain fixed-size blocks, which inevitably causes information loss. We propose a new training approach to overcome this limitation. Instead of absolute position embeddings, we interpret volumes as sequences of 3D chunks and adopt Rotary Position Embeddings, allowing us to treat the z-axis as an unconstrained temporal dimensions. Building on this idea, we introduce a new vision-language model: SigVLP. In SigVLP, we implement Rotary Position Embedding as the positional encoding method, which is applied directly within the attention operation, generating input-conditioned sine and cosine weights on the fly. This design ensures consistent alignment between query and key projections and adapts to any input sizes. To allow for variable input size during training, we sample Computed Tomography volumes in chunks and pair them with localized organ-wise textual observations. Compared to using entire reports for conditioning, chunkwise alignment provides finer-grained supervision, enabling the model to establish stronger correlations between the text and volume representations, thereby improving the precision of text-to-volume alignment. Our models are trained with the Muon optimizer and evaluated on a diverse set of downstream tasks, including zero-shot abnormality and organ classification, segmentation, and retrieval tasks.