A pancreatic tumor subregion segmentation method based on hierarchical prior learning and constrained clustering

By employing hierarchical prior learning and constrained clustering methods, combined with multidimensional feature vectors and expert annotations, fine segmentation of pancreatic tumor subregions was achieved. This solved the interpretability and quantification issues of tumor image subregion segmentation under non-invasive conditions, improving the accuracy of tumor assessment and the reliability of treatment.

CN122289699APending Publication Date: 2026-06-26TONGJI HOSPITAL ATTACHED TO TONGJI MEDICAL COLLEGE HUAZHONG SCI TECH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
TONGJI HOSPITAL ATTACHED TO TONGJI MEDICAL COLLEGE HUAZHONG SCI TECH
Filing Date
2026-05-26
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Current technologies cannot perform subregion segmentation of pancreatic tumor images with high biological interpretability and clinical relevance under non-invasive conditions, and cannot effectively quantify intratumoral spatiotemporal heterogeneity.

Method used

A method based on hierarchical prior learning and constrained clustering is adopted. By acquiring dual-phase enhanced CT images, multi-dimensional feature vectors are extracted, and hierarchical prior learning is performed by combining sparse point labels with expert experience annotation. The constrained clustering framework is then incorporated to force the division into three sub-regions, followed by post-processing and quality verification.

Benefits of technology

It enables refined, interpretable characterization and quantification of tumor internal heterogeneity on non-invasive imaging, improving the accuracy of preoperative assessment and providing reliable imaging evidence for individualized treatment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122289699A_ABST
    Figure CN122289699A_ABST
Patent Text Reader

Abstract

This invention relates to the field of medical image segmentation technology, specifically to a pancreatic tumor subregion segmentation method based on hierarchical prior learning and constrained clustering. By introducing sparse prior knowledge annotated by experts to guide the clustering process, a two-stage algorithm framework of hierarchical prior learning and constrained clustering is formed. This combines the advantages of data-driven and knowledge-driven approaches, enabling the algorithm to learn the distribution patterns of expert annotations and generalize to the entire tumor for robust segmentation even when they are sparse. Furthermore, the constraints avoid generating meaningless small fragments or inconsistent class numbers, ensuring the stability and consistency of the segmentation results on large-scale datasets. This is a key prerequisite for subsequent quantitative calculation of tumor heterogeneity index and construction of accurate preoperative grading models, ensuring from the source that the extracted heterogeneity features truly reflect the biological nature of the tumor, thereby significantly improving the performance of downstream tasks.
Need to check novelty before this filing date? Find Prior Art