A New and Efficient Approach to Cell Segmentation and Tumor Detection in Histopathological Images

Abstract

Detecting and segmenting cells or tumors in histopathological images is a very challenging task. This study presents a novel technique for detecting tumors and segmenting cells in histopathological images. The EBHI-SEG dataset contains 5170 images of six types of tumor differentiation stages and the corresponding ground truth images. Normalization was applied using a specific target image. Then, the Gaussian blur technique was applied to reduce the noise in the original image. The Otsu's thresholding method was applied to obtain binary images, followed by morphological operations. The results obtained were evaluated using the Jaccard index, Intersection Over Union (IOU), Precision, Recall, F1-score, and Structural Similarity Index (SSIM), showing satisfactory results. A comparison with six other well-known methods showed that the proposed approach provides promising and sufficient results.