Efficient WSI classification with sequence reduction and tranformers pretrained on text in Scientific Reports

Pisula, J.I., Bozek, K. Efficient WSI classification with sequence reduction and transformers pretrained on text. Sci Rep 15, 5612 (2025). https://doi.org/10.1038/s41598-025-88139-5

Language models have shown remarkable success in various tasks beyond NLP, such as computer vision and protein folding. In this work, we propose their use in Whole Slide Image (WSI) classification. To address the computational challenges associated very large slide resolution, we propose an input pooling layer, SeqShort, which summarises slides into shorter sequence representations. We use SeqShort to effectively classify WSIs in different digital pathology tasks using a deep, text pre-trained transformer model while fine-tuning less than 0.1% of its parameters, demonstrating that their knowledge about natural language transfers well to this domain.