<![CDATA[The manual interpretation of Hematoxylin and Eosin (H&E) histopathology images for breast cancer diagnosis is hindered by time constraints and observer bias. This research aims to develop an automated system using deep learning for cell detection and classification by evaluating two key approaches: Multi-class Segmentation (single-stage) and Segmentation followed by Classification (two-stage). The U-Net architecture was employed for segmentation, while MobileNetV2 and VGG16 were used for classification. The models were evaluated on the public IHC4BC dataset and primary data from Airlangga University Hospital (RSUA). The study also evaluated the impact of Resizing and Tiling data processing strategies. Experimental results showed that although the MobileNetV2 and VGG16 classification models achieved a high testing accuracy of 98.80%, the integrated two-stage system exhibited a high counting error, with a Mean Absolute Error (MAE) of 119.87 for positive cells, primarily due to under-segmentation of overlapping cells. In contrast, the Multi-class Segmentation approach utilizing the Tiling strategy demonstrated superior performance. This model effectively preserved spatial resolution while distinguishing cell types simultaneously, achieving the lowest MAE of 18.46 for positive cells and 1.66 for negative cells. This study concludes that Multi-class Segmentation with the Tiling strategy is the most effective and accurate approach for automated cell counting in histopathology images.]]>
Copyrights © 2026
