Abstract—Mitosis count serves as a critical biomarker in breast cancer research, aiding in the prediction of aggressiveness, prognosis, and grade of the disease. However, accurately identifying mitotic cells amidst shape and stain variations, while distinguishing them from similar objects like lymphocytes and cells with dense nuclei, presents a significant challenge. Traditional machine learning methods have struggled with this task, particularly in detecting small mitotic cells, leading to high inter-rater variability among pathologists. In recent years, the rise in deep learning has reduced the subjectivity of mitosis detection. However, Deep Learning models face challenges with segmenting and classifying mitosis due to its intricate morphological variations, cellular heterogeneity, and overlapping structures. In response to these challenges, this study presents an Intelligent Mitosis Segmentation and Detection in Breast Cancer Histopathological Images Using Deep Learning (IMSD-BCHIDL) Model. The purpose of the IMSD-BCHIDL technique is to segment and classify mitosis in the histopathological images. To accomplish this, the IMSD-BCHIDL technique mainly employs YOLO-v5 model, which proficiently segments and classifies the mitosis cells. In addition, InceptionV3 is applied as a backbone network for the YOLO-v5 model, which helps in capturing extensive contextual details from the input image and results in improved detection tasks. For demonstrating the greater solution of the IMSD-BCHIDL method of the IMSD-BCHIDL technique, a wide range of experimental analyses is made. The simulation values portrayed the improved solution of the IMSD-BCHIDL system with other recent DL models.
 
Keywords—breast cancer, histopathological images, segmentation, YOLO-v5, mitosis cells, deep learning


Copyright © 2024 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.