Intelligent Gastric Histopathology Image Classification Using Hierarchical Conditional Random Field based Attention Mechanism

In this paper, an Intelligent Hierarchical Conditional Random Field based Attention Mechanism (HCRF-AM) model is proposed, which can be applied to the Gastric Histopathology Image Classification (GHIC) tasks to assist pathologists in medical diagnosis. However, there exists redundant information in a weakly supervised learning mission. Thus, designing the network that can extract effective distinguishing features has become the focus of research. The HCRF-AM model consists of attention mechanism (AM) module and image classification (IC) module. First, in the AM module, an HCRF model is built to extract attention areas. Then, a convolutional neural network (CNN) model is trained with the attention region selected. Thirdly, an algorithm called classification probability based Ensemble Learning (EL) is used to obtain the image-level result from patch-level output of the CNN. In the experiment, a classification specificity of 96.67% is achieved on a gastric histopathological dataset with 700 images.