Democratising Pathology Co-Pilots: An Open Pipeline and Dataset for Whole-Slide Vision-Language Modelling

TL;DR

This paper introduces a new dataset and model for whole-slide image vision-language tasks in pathology, enabling more transparent and generalisable co-pilots for pathologists.

Contribution

It presents Polysome for synthetic instruction generation, creates the HISTAI-Instruct dataset, and trains ANTONI-α, a VLM that surpasses existing models on WSI-level VQA tasks.

Findings

ANTONI-α outperforms MedGemma on tissue identification and neoplasm detection.

HISTAI-Instruct contains over 1.1 million instruction-response pairs.

Multiple versions of ANTONI-α trained with different data amounts are evaluated.

Abstract

Vision-language models (VLMs) have the potential to become co-pilots for pathologists. However, most VLMs either focus on small regions of interest within whole-slide images, provide only static slide-level outputs, or rely on data that is not publicly available, limiting reproducibility. Furthermore, training data containing WSIs paired with detailed clinical reports is scarce, restricting progress toward transparent and generalisable VLMs. We address these limitations with three main contributions. First, we introduce Polysome, a standardised tool for synthetic instruction generation. Second, we apply Polysome to the public HISTAI dataset, generating HISTAI-Instruct, a large whole-slide instruction tuning dataset spanning 24,259 slides and over 1.1 million instruction-response pairs. Finally, we use HISTAI-Instruct to train ANTONI-{\alpha}, a VLM capable of visual-question answering (VQA). We show that ANTONI-{\alpha} outperforms MedGemma on WSI-level VQA tasks of tissue identification, neoplasm detection, and differential diagnosis. We also compare the performance of multiple incarnations of ANTONI-{\alpha} trained with different amounts of data. All methods, data, and code are publicly available.