Prostate Cancer Recognition Using Incremental Transformer Pipeline
Year: 2026
Type: conference-paper
Abstract: Prostate cancer (PCa) is one of the leading causes of cancer-related deaths in men, making it so that precise and accurate analysis of tissue structures is essential for detecting PCa and improving treatment outcomes. Traditionally, deep learning approaches for prostate tissue recognition usually rely on large-scale annotated datasets, the preparation of said datasets is both costly and time-intensive. To address this limitation, we propose an incremental transformer pipeline that progressively adapts segmentation-based transformer models to distinguish between distinct prostate tissue instances in whole-slide images (WSIs). The extraction of these tissue instances reduces the amount of annotation needed while providing a robust foundation for PCa identification in accordance with clinical guidelines. In this work, experimental evaluation on two public datasets, Louisville and PANDA, showed that the proposed approach outperformed state-of-the-art methods by$\mathbf{1 8. 6 8 \%}$and$\mathbf{1 7. 4 7 \%}$, respectively, in terms of mean intersection-over-union$(\mu \mathbf{I o U})$. These results display potential of integration for the proposed pipeline into clinical workflows, supporting pathologists and allowing for easier diagnostic workload.
Institution: Abu Dhabi University
Language: English
Topic: AI in cancer detection
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204254247