A novel Nine-SequenceNet with attention for liver histopathological classification

Liver cancer is a prevalent cancer worldwide and is also the leading cause of cancer-related deaths. Histopathological image diagnosis is taken as the standard to identify liver cancer. However, manually histopathological examination of liver cancer WSIs (Whole Slide Image) is time consuming. To address this issue, researchers have applied the automatic classification methods to recognize cancerous tissues in histopathological images, which can effectively reduce the examination time and improving the recognition efficiency. In recent years, deep learning based methods have been widely used in histopathological image analysis because of their impressive performance level. But only a little research work based on transfer learning has been done on liver cancer histopathological images because the sample dataset is difficult to collected and the classification results lack of effectiveness and pertinence. One issue with fine-grained classification is that using a single patch for classification results in the loss of edge information, degrading classification performance. To solve the problem, this paper proposes a novel deep learning model based on sequence prediction with attention, which is used to identify and classify liver cancer histopathological images. The proposed method uses the surrounding 8 patches to supplement information to the current patch, resulting in better results than other methods. The experimental results show that this method achieves the best F1-score of 95.23% which illustrates that the proposed method is efficient in liver cancer histopathological image classification.