Segmentation, Detection, and Classification of Cell Nuclei on Oral Cytology Samples Stained with Papanicolaou

Access the paper here.

Abstract

Although oral cancer is considered a global health issue with 350.000 people diagnosed every year, it can successfully be treated if diagnosed at an early stage. The Papanicolaou staining is an inexpensive and non-invasive method, generally applied to detect cervical cancer, but it can also be useful to detect cancer in oral cavities. The manual process of analyzing cells to detect abnormalities is a time-consuming process and is subject to variations in perceptions from different professionals. Aiming at an objective, quantitative and automated cytology method, this paper compares three different deep learning approaches: segmentation, object detection and image classification. Our results show that the binary object detection with Faster R-CNN is the best approach for nuclei detection and localization (0.76 IoU). Since ResNet34 had a good performance on abnormal nuclei classification (0.86 $F_1$ score) we conclude that these two models can be used in combination to perform a reliable localization and classification pipeline. This work indicates that the automated analysis of oral cytology to build a pipeline for nuclei classification and localization using deep learning can contribute to minimize the subjectivity of the human analysis and also support the detection of cancer at early stages.

Bibtex

@article{MatiasMultimethodPAP2021,
    doi = {10.1007/s42979-021-00676-8},
    url = {https://doi.org/10.1007/s42979-021-00676-8},
    year = {2021},
    month = may,
    publisher = {Springer Science and Business Media {LLC}},
    volume = {2},
    number = {4},
    author = {Andr{\'{e}} Vict{\'{o}}ria Matias and Allan Cerentini and Luiz Antonio Buschetto Macarini and Jo{\~{a}}o Gustavo Atkinson Amorim and Felipe Perozzo Dalto{\'{e}} and Aldo von Wangenheim},
    title = {Segmentation, Detection, and Classification of Cell Nuclei on Oral Cytology Samples Stained with Papanicolaou},
    journal = {SN Computer Science}
}