Institution profile

Silesian Technical University of Gliwice

Academic institutioneurope · pl
Official website
Research library17linked papers
Opportunities0open roles
Selected work

Representative Papers

Can Unsupervised Methods Outperform Supervised Deep Learning When Ground Truth Is Sparse? A Case Study of Bronchovascular Bundle Segmentation in Low-Dose CT

Aug 17, 2026

This study addresses the challenge of missed pulmonary nodule detection caused by difficult bronchovascular bundle segmentation in low-dose CT. We propose RONALD, a multi-stage pipeline incorporating lung lobe and mediastinum preprocessing alongside independent vessel and bronchus segmentation strategies. Notably, this approach demonstrates that unsupervised methods outperform supervised learning in ground-truth-sparse scenarios. Experimental results indicate that RONALD significantly enhances early lung cancer screening efficacy, achieving nodule retention rates of 100% and 99.92% on the DLCS and Pomeranian datasets, respectively. Consequently, this framework effectively resolves precise segmentation difficulties within complex anatomical structures, ensuring robust nodule preservation during automated analysis.

0 citationsRead paper

Decomposing Whole Slide Image Report Generation with Graph-Constrained Multiple Instance Learning Workflows

Aug 15, 2026

This study addresses the challenges of opaque reasoning and poor narrative coherence in whole-slide image report generation by proposing an organ-conditioned graph-based decomposable framework. The approach decouples the generation process into visual recognition, structured reasoning, and text synthesis, leveraging multi-instance learning and large language models to construct interpretable reasoning chains that facilitate modular debugging and error attribution. Experimental results demonstrate that the model achieves a chain-wise Jaccard score of 0.702 on the reg2026 test set. Furthermore, diagnostic consistency on external TCGA data improves significantly from 61.8% to 92.6%, effectively enabling stage-level error localization and enhancing overall report quality through transparent, structured inference.

0 citationsRead paper
Recent publications

Latest Papers

Can Unsupervised Methods Outperform Supervised Deep Learning When Ground Truth Is Sparse? A Case Study of Bronchovascular Bundle Segmentation in Low-Dose CT

Aug 17, 2026

This study addresses the challenge of missed pulmonary nodule detection caused by difficult bronchovascular bundle segmentation in low-dose CT. We propose RONALD, a multi-stage pipeline incorporating lung lobe and mediastinum preprocessing alongside independent vessel and bronchus segmentation strategies. Notably, this approach demonstrates that unsupervised methods outperform supervised learning in ground-truth-sparse scenarios. Experimental results indicate that RONALD significantly enhances early lung cancer screening efficacy, achieving nodule retention rates of 100% and 99.92% on the DLCS and Pomeranian datasets, respectively. Consequently, this framework effectively resolves precise segmentation difficulties within complex anatomical structures, ensuring robust nodule preservation during automated analysis.

0 citationsRead paper

Decomposing Whole Slide Image Report Generation with Graph-Constrained Multiple Instance Learning Workflows

Aug 15, 2026

This study addresses the challenges of opaque reasoning and poor narrative coherence in whole-slide image report generation by proposing an organ-conditioned graph-based decomposable framework. The approach decouples the generation process into visual recognition, structured reasoning, and text synthesis, leveraging multi-instance learning and large language models to construct interpretable reasoning chains that facilitate modular debugging and error attribution. Experimental results demonstrate that the model achieves a chain-wise Jaccard score of 0.702 on the reg2026 test set. Furthermore, diagnostic consistency on external TCGA data improves significantly from 61.8% to 92.6%, effectively enabling stage-level error localization and enhancing overall report quality through transparent, structured inference.

0 citationsRead paper