Fine-tuning improves cell nuclei detection in kidney pathology images
Comprehensive Evaluation and Fine-Tuning of Foundational Cell Nuclei Segmentation Models in Renal Pathology
Computer Vision and Pattern Recognition
Summary
Detecting cell nuclei accurately in kidney tissue images is important for medical analysis but is difficult because of complex shapes and background noise. The authors improved this by combining many automatic and expert-made labels that show different levels of difficulty. They tested several models and ways of fine-tuning, finding that training with medium-difficulty labels helped most models perform better. Different models favored different mixes of label difficulties for best results.
What this means in practice
- •For medical imaging teams: Improve cell nuclei segmentation in kidney pathology scans by using a mixed-difficulty annotation approach to fine-tune segmentation models.
- •For digital pathology software developers: Develop more accurate renal pathology image analysis tools by incorporating foundation models fine-tuned on curated annotation sets.$Commercial implications: Enables commercial pathology software products to provide better quantitative analysis of kidney tissue for healthcare providers.
Authors
Ruijie Wu, Junlin Guo, Ruining Deng, Yu Wang, Shilin Zhao, Haichun Yang, Yuankai Huo
Abstract
Accurate nuclei instance segmentation is essential for quantitative renal pathology, yet general-purpose models often struggle with low contrast, dense nuclei, complex morphology, and strong background staining. In this work, we extended a human-in-the-loop framework by combining 5,901 foundation-model-generated pseudo-labels from well-segmented cases (Easy), 860 newly expert-annotated unresolved challenging cases (Medium), and 198 expert-annotated consensus failure cases (Hard). These annotations, spanning different levels of segmentation difficulty, enabled the systematic evaluation of seven single-source and mixed-source fine-tuning strategies across nine cell segmentation model configurations. Fine-tuning improved all models, with Medium data included in seven of the nine best-performing strategies. LSP-DETR achieved the highest F1 score of 0.8725 with Hard-only fine-tuning, while StarDist showed the largest improvement, increasing from 0.7380 to 0.8332 with Medium-only fine-tuning. These findings show that annotations spanning multiple difficulty levels support effective model adaptation, although the optimal annotation composition remains model dependent.