Automated Nuclei Detection on Pleural Effusion Cytopathology Images using YOLOv3


Kilic B., Baykal E. , Ekinci M., Dogan H., Ercin M. E. , Ersoz S.

4th International Conference on Computer Science and Engineering, UBMK 2019, Samsun, Turkey, 11 - 15 September 2019, pp.418-422 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • Doi Number: 10.1109/ubmk.2019.8907125
  • City: Samsun
  • Country: Turkey
  • Page Numbers: pp.418-422
  • Keywords: nuclei detection, cytopathology, pleural effusion, convolutional neural networks, YOLOv3

Abstract

Nuclei detection is a critical step in cytopathology for pleural cancer diagnosis. Because nuclei provide quantitative information about malignancy of cancer. It is known that Convolutional Neural Networks (CNNs) provide considerable performance on object detection. We proposed convolutional object detector, YOLOv3, to detect nuclei on pleural effusion (PE) cytopathology images. The experiments were conducted on 80 PE cytopathology images containing 11157 nuclei. The proposed method achieved precision of %94.10, recall of %98.98, and F-measure of %96.48. The most important contribution of the YOLOv3 is that it provides 10x speed up versus the some state-of-the-art published methods, which is very important for real time Computer-Aided Diagnosis (CAD) applications in digital cytopathology.