Nuclei Detection Using Deep Learning
Eliezer Farrant Braz, Roberto de Alencar Lotufo

DOI: 10.14209/sbrt.2017.48
Evento: XXXV Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2017)
Keywords: Nucleus detection Convolutional Neural Networks
Abstract
An important objective in the analysis of Pap smear images is automatic detection of a cell’s nucleus. Various methods that automatically detect the nuclei of cervical cells have been proposed to improve the analysis of screening test images. In this paper, we propose a Convolutional Neural Networksbased method that automatically detects the nuclei of cervical cells. Following training using a public dataset provided by the Overlapping Cervical Cytology Image Segmentation Challenge - ISBI 2014, the network’s fully connected layers are converted to convolutional layers to enable processing of images of any size. Our results were then compared with those achieved by other participants who successfully submitted their work to ISBI 2014 and other studies that used the same dataset. Our experimental results indicate that the methodology provides fast nuclei detection with precision and recall that are comparable with the state-of-the-art methods used to detect the nuclei of cervical cells.

Download