Hierarchical Pretrained Backbone Vision Transformer for Image Classification in Histopathology

Histopathology plays a crucial role in clinical diagnosis, treatment planning, and research by enabling the examination of diseases in tissues and organs. However, the manual analysis of histopathological images is time-consuming and labor-intensive, requiring expert pathologists. To address this issue, this work proposes a novel architecture called Hierarchical Pretrained Backbone Vision Transformer for automated histopathological image classification, a critical tool in clinical diagnosis, treatment planning, and research. Current deep learning-based methods for image classification require a large amount of labeled data and significant computational resources to be trained effectively. By leveraging pretrained Visual Transformer backbones, our approach can classify histopathology images, achieve state-of-the-art performance, and take advantage of the pretrained backbones’ weights. We evaluated it on the Chaoyang histopathology dataset, comparing it with other state-of-the-art Visual Transformers. The experimental results demonstrate that the proposed architecture outperforms the others, indicating its potential to be an effective tool for histopathology image classification.