TC-SSA: Token Compression via Semantic Slot Aggregation for Gigapixel Pathology Reasoning

TL;DR

This paper introduces TC-SSA, a learnable token compression method that aggregates patch features into semantic slots, enabling efficient and accurate gigapixel pathology analysis with significantly reduced tokens.

Contribution

The paper presents a novel semantic slot aggregation framework that effectively compresses visual tokens in large pathology images, maintaining diagnostic accuracy while reducing computational load.

Findings

Achieves 78.34% overall accuracy on SlideBench with only 1.7% of original tokens.

Outperforms sampling-based methods under similar token budgets.

Generalizes well to multiple MIL classification tasks with high AUC scores.

Abstract

The application of large vision-language models to computational pathology holds great promise for diagnostic assistants but faces a critical computational bottleneck: the gigapixel scale of Whole Slide Images (WSIs). A single WSI typically contains over 105 patches, creating sequence lengths that exceed the constraints of standard Transformer architectures. Existing solutions often resort to spatial sampling, which risks discarding diagnostically critical evidence. To address this, we propose TC-SSA (Token Compression via Semantic Slot Aggregation), a learnable token compression framework that aggregates patch features into a fixed number of semantic slots. A gated routing module assigns patches to slots using sparse Top-2 routing, followed by weighted aggregation, enabling global slide coverage under a strict token budget. The resulting representation retains diagnostically relevant information while reducing the number of visual tokens to 1.7% of the original sequence. On the SlideBench(TCGA), our model achieves 78.34% overall accuracy and 77.14% on the diagnosis subset, outperforming sampling-based baselines under comparable token budgets. The method also generalizes to MIL classification, reaching AUC of 95.83% on TCGA-BRCA, 98.27% on TCGA-NSCLC and 79.80% on PANDA. These results suggest that learnable semantic aggregation provides an effective trade-off between efficiency and diagnostic performance for gigapixel pathology reasoning.