EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Morphology-Driven and Symmetry-Guided Geometric Segmentation of White Blood Cells With Self-Consistency Reliability Assessment

P. Umarani, P. Viswanathan

IEEE Access · 2026 · 14 · 73163-73179

DOI: 10.1109/access.2026.3690641 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von P. Umarani und 1 weiteren Autor:innen. Die Arbeit nennt 21 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: VIT University.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate segmentation of blood cells in microscopic images is a fundamental step in automated hematological analysis. However, it remains challenging because of staining variability, illumination inconsistencies, imaging noise, and severe cell overlap. This study presents a unified, fully automated morphological and symmetry-based framework for white blood cell region segmentation in blood smear images. The proposed method integrates Laplacian-of-Gaussian-based contrast enhancement and hybrid color-space analysis in RGB, $\mathrm{CIELab}^{*}$ , and HSV domains for candidate region extraction, hierarchical segmentation refinement, and marker-controlled watershed for instance-level separation of touching and overlapping cells. To support annotation-free internal assessment, a novel Self-Consistency Reliability Index (SCRI) is introduced by combining stage-wise structural stability, boundary-alignment fidelity, and geometric shape reliability. An experimental evaluation of 50 high-resolution microscopic blood smear images yielded 1,422 detected regions, corresponding to an average of 28.44 cells per image. The mean Dice coefficients were 0.8777 for Level 2 versus Level 3 and 0.9997 for Level 3 versus final segmentation, while the corresponding mean IoU for the final refinement stage was 0.9993. Boundary-alignment evaluation produced a precision proxy of 0.7890, recall proxy of 0.0571, and F1 proxy of 0.1045. In addition, consensus analysis between the refinement stages achieved precision, recall, and F1-score values of 1.0000, 0.9993, and 0.9997, respectively. The mean shape reliability was 0.6884, and the overall SCRI was $0.5975 \pm 0.0204$ . These findings demonstrate a strong internal consistency and stable geometric performance across the dataset. The framework provides an interpretable and data-efficient alternative for microscopy image segmentation, although external validation against annotated ground-truth datasets is necessary for definitive accuracy assessment.

Förderung

  • VIT University

Zitieren

APA

Umarani, P., & Viswanathan, P. (2026). Morphology-Driven and Symmetry-Guided Geometric Segmentation of White Blood Cells With Self-Consistency Reliability Assessment. IEEE Access, 14, 73163-73179. https://doi.org/10.1109/access.2026.3690641

Vancouver

Umarani P, Viswanathan P. Morphology-Driven and Symmetry-Guided Geometric Segmentation of White Blood Cells With Self-Consistency Reliability Assessment. IEEE Access. 2026;14:73163-73179. doi:10.1109/access.2026.3690641

BibTeX
@article{Umarani2026,
  author = {Umarani, P. and Viswanathan, P.},
  title = {Morphology-Driven and Symmetry-Guided Geometric Segmentation of White Blood Cells With Self-Consistency Reliability Assessment},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {73163--73179},
  doi = {10.1109/access.2026.3690641}
}
Kostenlos starten →