Mitotic Instance Detection in Stain Normalized Histopathological Images using Faster R-CNN
H Anand, Anoop K Rajan, G Santhosh, Lekha S. Nair · 2022 7th International Conference on Communication and Electronics Systems (ICCES) · 2022
Our method aims to detect and identify the possibility of malignancy based on abnormal mitotic cell division using Artificial Intelligence. In tumour grading, the mitotic activity index is an important prognostic factor. The standard approach entails expert pathologists manually examining H&E stained histopathological sections on glass slides under a microscope, which is time-consuming and error-prone, thereby a need for automation. Due to the lack of pixel-level annotations, the presence of mitotic nuclei in numerous morphological configurations, their sparse representation, and their strong similarity to other cellular and nuclear bodies, automated mitotic nuclei recognition faces many obstacles. Hence, an efficient and accurate automated approach is necessary for early prognosis. This early detection helps the pathologists to find the mitotic cells considerably faster. This research work has used various backbone architectures of the Faster R-CNN model to classify the mitotic nuclei and compare their performance.