Abstract
Automatic recognition of histopathological image plays an important role in building computer-aid diagnosis system. Traditionally hand-craft features are widely used for representing histopathological images when building recognition models. Currently with the development of deep learning algorithms, deep features obtained directly from pre-trained networks at less costs show that they perform better than traditional ones. However, most recent work adopts common pre-trained networks for feature extraction and train a classifier with domain knowledge which generates a gap between the common extracted features and the application domain. To fill the gap and improve the performance of the recognition model, in this paper we propose a deep model for histopathological image feature representation in a supervised-learning manner. The proposed model is constructed based on some pre-trained convolutional neural networks. After supervised learning, the feature learning network captures most domain knowledge. The proposed model is evaluated on two histopathological image datasets and the results show that the proposed model is superior to current state-of-the-art models.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call