<![CDATA[Background: Accurate segmentation of cell nuclei in histopathological images plays a crucial role in computational pathology, as the results serve as a foundation for various clinical practices, including disease diagnosis, prediction, and prognosis. Deep learning methods like U-Net have greatly enhanced performance, but challenges such as tissue heterogeneity, cell nucleus overlap, and complex staining patterns still exist. Objective: This study aims to assess the effectiveness of the Attention Mechanism model within the U-Net architecture for cell nucleus segmentation in Hematoxylin and Eosin (H&E) stained histopathology images. By focusing on relevant spatial features, the Attention Mechanism is expected to improve the model’s ability to accurately distinguish and segment areas with overlapping cells. Specifically, this study also aims to examine whether the proposed model outperforms the conventional U-Net model. Methods: This study used a quantitative experimental approach, utilizing an H&E-stained histopathology image dataset from Saitama Medical University International Medical Center (SIMC). The Attention-Enhanced U-Net Model was trained and tested on pathologist-annotated cell nucleus data, then evaluated using performance metrics such as Dice Coefficient, Accuracy, Precision, Recall, F1-Score, AUROC Mean, and Intersection over Union (IoU). The experimental results showed that the model produced a Dice Coefficient of 0.927, Precision of 0.889, Recall of 0.861, F1-Score of 0.875, and IoU of 0.793. These findings indicate that the model can accurately capture the structure of a cell nucleus, even in challenging conditions such as different cell shapes and the presence of H&E staining. Results: Furthermore, integrating Attention Mechanisms allows the model to focus on extracting relevant features while reducing background noise. This improves its potential as a reliable segmentation solution in clinical pathology workflows. For future research, it is recommended to validate the model using a larger, more diverse dataset to improve its generalization and reliability in real-world clinical practice. Conclusion: The research concludes that the Attention-Enhanced U-Net model effectively achieves high-precision cell nucleus segmentation in H&E-stained histopathology images. It demonstrates strong performance across five metrics: Dice (0.927), Precision (0.889), Recall (0.861), F1-Score (0.875), and IoU (0.793). The model accurately detects nuclei, even in challenging conditions such as morphological variation, staining artifacts, and overlapping structures. Its attention mechanism improves feature extraction by focusing on relevant regions and reducing background noise, enhancing localization and delineation. The lightweight design supports clinical use with limited resources. Future studies should validate its generalizability on larger, more diverse datasets and clinical scenarios. Keywords: Cell Nuclei Segmentation, Attention Enhanced U-Net, H&E Staining; Deep Learning, Medical Image Analysis.]]>
