RA-UNet: Achieving robust blood cell segmentation in Complex microscopic environments
BackgroundTracking, classifying, and segmenting cell images are critical clinical tasks essential in diagnosing various diseases such as leukemia and in performing blood counts. While deep neural architectures are preferred for these tasks, the presence of noise in images, overlapping cells, and indistinct cell boundaries make it a challenging endeavor.ObjectiveIn this article, an attention mechanism based on U-Net has been developed for cell segmentation. This attention mechanism enhances detail capture while suppressing irrelevant information.MethodsThe proposed model has been tested on four publicly available cell datasets and evaluated using seven different performance metrics. The performance of the proposed model was benchmarked against state-of-the-art methods.ResultsThe proposed model demonstrated the best overall performance on the dataset named Dataset-III, while it was most successful on Dataset-I. The experimental results were evaluated both quantitatively and qualitatively.ConclusionThe outcomes have substantiated the efficacy and potential of the proposed model to provide a competitive approach.
Authors
- Buket Toptaş (ORCID: https://orcid.org/0000-0003-2556-8199)
- Davut Hanbay (ORCID: https://orcid.org/0000-0003-2271-7865)
- Murat Toptaş (ORCID: https://orcid.org/0000-0001-6978-5289)
Institutions
- Inonu University (TR)
- Bandırma Onyedi Eylül University (TR)
Publication Details
- Journal
- Technology and Health Care
- Published
- 2026-11-18
- DOI
- https://doi.org/10.1177/09287329261509252
- Primary Topic
- Digital Imaging for Blood Diseases
- Type
- article
- Field-Weighted Citation Impact
- 0.00