Nuclei Classification Using Dual View CNNs with Multi-crop Module in Histology Images

Histopathology image diagnostic technique is a quite common requirement; however, cell nuclei classification is still one of key challenge due to complex tissue structure and diversity of nuclear morphology. Cell nuclei categories are often defined by contextual information, including central nucleus and surrounding background. In this paper, we propose a Dual-View Convolutional Neural Networks (DV-CNNs) that captures contextual contents from different views. The DV-CNNs are composed of two independent pathways, one for global region and another for center local region. Noted that each pathway with “multi-crop module” can extract five different feature regions. Common networks do not fully utilize the local information, but the designed cropping module catches information for more complete features. In experiments, two pipelines are complementary to each other in score fusion. To verify the performance in proposed framework, it is evaluated on a colorectal adenocarcinoma image database with more than 20,000 nuclei. Compared with existing methods, our proposed DV-CNNs with multi-crop module demonstrate better performance.

[1]  Hernandes F. Carvalho,et al.  Colon Adenocarcinoma Diagnosis in Human Samples by Multicontrast Nonlinear Optical Microscopy of Hematoxylin and Eosin Stained Histological Sections , 2014 .

[2]  Max A. Viergever,et al.  Breast Cancer Histopathology Image Analysis: A Review , 2014, IEEE Transactions on Biomedical Engineering.

[3]  Jürgen Schmidhuber,et al.  Deep learning in neural networks: An overview , 2014, Neural Networks.

[4]  George Lee,et al.  Image analysis and machine learning in digital pathology: Challenges and opportunities , 2016, Medical Image Anal..

[5]  Nasir M. Rajpoot,et al.  Locality Sensitive Deep Learning for Detection and Classification of Nuclei in Routine Colon Cancer Histology Images , 2016, IEEE Trans. Medical Imaging.

[6]  Fabio A. González,et al.  Cascaded ensemble of convolutional neural networks and handcrafted features for mitosis detection , 2014, Medical Imaging.

[7]  Sergey Ioffe,et al.  Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift , 2015, ICML.

[8]  Hongfei Lin,et al.  Deep Transfer Learning for Modality Classification of Medical Images , 2017, Inf..

[9]  R. Kumar,et al.  Detection and Classification of Cancer from Microscopic Biopsy Images Using Clinically Significant and Biologically Interpretable Features , 2015, Journal of medical engineering.

[10]  H. Irshad,et al.  Methods for Nuclei Detection, Segmentation, and Classification in Digital Histopathology: A Review—Current Status and Future Potential , 2014, IEEE Reviews in Biomedical Engineering.

[11]  Eric Cosatto,et al.  Classification of mitotic figures with convolutional neural networks and seeded blob features , 2013, Journal of pathology informatics.

[12]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[13]  David J. Hand,et al.  A Simple Generalisation of the Area Under the ROC Curve for Multiple Class Classification Problems , 2001, Machine Learning.

[14]  Janne Heikkilä,et al.  Transfer Learning for Cell Nuclei Classification in Histopathology Images , 2016, ECCV Workshops.

[15]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[16]  Jörg Bredno,et al.  Using contextual information to classify nuclei in histology images , 2015, 2015 IEEE 12th International Symposium on Biomedical Imaging (ISBI).

[17]  Joel H. Saltz,et al.  Center-Focusing Multi-task CNN with Injected Features for Classification of Glioma Nuclear Images , 2016, 2017 IEEE Winter Conference on Applications of Computer Vision (WACV).