Aspect improves accuracy of pathology image reasoning and cell counting
See, Measure, and Reason: Learning Visually Grounded Reasoning in Pathology
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Pathologists need to spot tiny details in microscope images to diagnose diseases correctly. The authors found that current AI models often guess cell details poorly even if their final answers seem right. They created a new system called ASPECT that helps AI better see and count cells by learning from expert feedback and special training steps. When tested on a new benchmark with expert-reviewed questions, ASPECT made fewer counting mistakes and gave more reliable explanations for its answers.
What this means in practice
- •For pathology ai developers: Build AI systems that improve accuracy in detecting and counting cells in histology images to support pathology workflows.
- •For medical imaging software teams: Integrate improved cell measurement modules to enhance reliability and interpretability of pathology question answering tools.
Authors
Chengyang Zhang, Wenchuan Zhang, Bo Li, Mengran Li, Xinyu Liu, Jiaming Yang, Jie Chen, Zhang Zhang, Yuhao Yi, Hong Bu, Jiancheng Lv
Abstract
Pathological assessment relies on recognizing fine-grained visual details in histological images. Vision-language models (VLMs) increasingly support pathology interpretation, yet their ability to perceive these details remains inadequate. This weakness leads to inaccurate cellular observations that can persist even when final answers are correct. In this paper, we propose ASPECT to improve visually grounded reasoning through explicit supervision of cellular appearance and abundance. ASPECT trains intermediate visual tokens through pathology feature reconstruction, cell feature alignment, and count supervision. Three-stage supervised fine-tuning teaches the model to perceive, generate visual tokens, and reason, followed by reinforcement learning that rewards answer correctness and consistency with reported measurements. We also introduce PathoVernier, a benchmark of 759 expert-reviewed questions from five pathology datasets covering four cellular composition tasks. It evaluates both final answers and intermediate measurements to expose errors hidden by answer accuracy. On PathoVernier, ASPECT achieves relative accuracy gains of approximately 19.2% over the strongest baseline, Gemini-3.1-Pro, and 99.3% over its Qwen3-VL-8B backbone, while reducing RAWR, which measures counting errors within correct responses, by 28.1% and 42.7%, respectively. ASPECT also improves over its backbone on three external pathology benchmarks covering classification and question answering beyond cellular composition tasks.