Cervical Nuclei Segmentation in Whole Slide Histopathology Images Using Convolution Neural Network

Pathologists generally diagnose whether or not cervical cancer cells have the potential to spread to other organs and assess the malignancy of cancer through whole slide histopathology images using virtual microscopy. In this process, the morphology of nuclei is one of the significant diagnostic indices, including the size, the orientation and arrangement of the nuclei. Therefore, accurate segmentation of nuclei is a crucial step in clinical diagnosis. However, several challenges exist, namely a single whole slide image (WSI) often occupies a large amount of memory, making it difficult to manipulate. More than that, due to the extremely high density and variant shapes, sizes and overlapping nuclei, as well as low contrast, weakly defined boundaries, different staining methods and image acquisition techniques, it is difficult to achieve accurate segmentation. A method is proposed, comprised of two main parts to achieve lesion localization and automatic segmentation of nuclei. Initially, a U-Net model was used to localize and segment lesions. Then, a multi-task cascade network was proposed to combine nuclei foreground and edge information to obtain instance segmentation results. Evaluation of the proposed method for lesion localization and nuclei segmentation using a dataset comprised of cervical tissue sections collected by experienced pathologists along with comparative experiments, demonstrates the outstanding performance of this method.

[1]  Yi Li,et al.  R-FCN: Object Detection via Region-based Fully Convolutional Networks , 2016, NIPS.

[2]  Yi Li,et al.  Fully Convolutional Instance-Aware Semantic Segmentation , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[3]  Thomas Brox,et al.  U-Net: Convolutional Networks for Biomedical Image Segmentation , 2015, MICCAI.

[4]  Yoshua Bengio,et al.  Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation , 2014, EMNLP.

[5]  Kaiming He,et al.  Mask R-CNN , 2017, 2017 IEEE International Conference on Computer Vision (ICCV).

[6]  Vibhav Vineet,et al.  Conditional Random Fields as Recurrent Neural Networks , 2015, 2015 IEEE International Conference on Computer Vision (ICCV).

[7]  Pascal Vincent,et al.  Stacked Denoising Autoencoders: Learning Useful Representations in a Deep Network with a Local Denoising Criterion , 2010, J. Mach. Learn. Res..

[8]  Trevor Darrell,et al.  Fully Convolutional Networks for Semantic Segmentation , 2017, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[9]  Jerome H. Saltzer,et al.  End-to-end arguments in system design , 1984, TOCS.

[10]  S. McGuire.  World Cancer Report 2014. Geneva, Switzerland: World Health Organization, International Agency for Research on Cancer, WHO Press, 2015. , 2016, Advances in nutrition.

[11]  Roberto Cipolla,et al.  SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[12]  Iasonas Kokkinos,et al.  DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs , 2016, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[13]  Jian Sun,et al.  Instance-Aware Semantic Segmentation via Multi-task Network Cascades , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[14]  Hao Chen,et al.  DCAN: Deep Contour-Aware Networks for Accurate Gland Segmentation , 2016, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).