A shape context fully convolutional neural network for segmentation and classification of cervical nuclei in Pap smear images

作者：

Highlights：

• The concurrent application of instance segmentation and classification on whole slide Pap smear images has been done for the first time.

• Instance Segmentation and classification has been accomplished using a fully convolutional neural network (FCN) model.

• The proposed model is built upon standard Unet architecture by the addition of residual blocks, densely connected blocks and a bottleneck layer.

• Additionally, a shape representation model has been integrated with the model which acts as a regularizer, making the whole framework robust.

摘要

•The concurrent application of instance segmentation and classification on whole slide Pap smear images has been done for the first time.•Instance Segmentation and classification has been accomplished using a fully convolutional neural network (FCN) model.•The proposed model is built upon standard Unet architecture by the addition of residual blocks, densely connected blocks and a bottleneck layer.•Additionally, a shape representation model has been integrated with the model which acts as a regularizer, making the whole framework robust.

论文关键词：Liquid-based cytology,Pap smear,Fully convolutional neural network,Segmentation,Classification

论文评审过程：Received 18 October 2019, Revised 21 May 2020, Accepted 29 May 2020, Available online 2 June 2020, Version of Record 10 June 2020.

论文官网地址：https://doi.org/10.1016/j.artmed.2020.101897