CytoSPM: Open-Vocabulary Cytopathology Detection with Structured Prompt Bank

2026-09-25

AI Summary

This study addresses the open-vocabulary recognition challenges in cytopathological detection arising from fine-grained categories, long-tailed distributions, and continuously evolving classes. To this end, it proposes the CytoSPM detector and constructs PentaCyto, a multi-domain benchmark. Methodologically, this work pioneers a structured prompt library based on morphological attributes and designs a two-stage decoupled architecture that separates class-agnostic visual representation extraction from class-aware structured prompt matching, thereby enabling efficient recognition. Experimental results demonstrate that the proposed method significantly outperforms existing approaches in novel category discovery and open-vocabulary detection tasks while maintaining high inference efficiency. Ultimately, this work provides a scalable detection paradigm for dynamically evolving pathological scenarios.

Abstract

Cytopathology detection requires open-vocabulary recognition because cellular categories are fine-grained, long-tailed, and continuously evolving across different organ systems. However, existing cytology detectors are mostly single-domain and closed-set, and there is still no unified benchmark for evaluating open-vocabulary cytopathology detection. We present PentaCyto, a multi-domain benchmark covering cervical, urinary, respiratory, serous fluid, and thyroid cytology, with 24 base categories and 9 held-out novel categories. Each category is associated with structured cytomorphology prompts that describe diagnostic morphological attributes and provide clinically grounded textual knowledge. We further propose CytoSPM, an efficient detector based on a decoupled two-stage design. It first extracts reusable class-agnostic visual representations, and then performs class-aware structural prompt matching with class names and cytomorphology prompts. On PentaCyto, CytoSPM outperforms existing methods in novel-category detection and open-vocabulary detection while maintaining efficient inference.

Problem

• Open-vocabulary detection
• Cytopathology
• Fine-grained recognition
• Long-tailed distribution
• Benchmark

Innovation

• Open-Vocabulary Detection
• Cytopathology
• Structured Prompt Bank
• Decoupled Two-Stage Design
• Multi-Domain Benchmark