Abstract
The diagnosis of biopsy tissue with hematoxylin and eosin (H&E) stained images has been widely used by pathologists to detect the lesions and assess the malignancy. Nevertheless, the diagnostic result relies on the visual observation of pathologists which may vary from person to person under different circumstances. With the advantage of automatically and adaptively learning features at multiple levels of abstraction, Convolutional Neural Networks (CNNs) have rapidly become promising alternatives for pathological image analysis. Therefore, in this paper, we propose an effective method for tumor classification called Hard Example Guided CNN. Our contribution is twofold: firstly, to optimize image representation, we design the CNN architecture as dual-branch, used for extracting global features and local features simultaneously. Secondly, we propose a re-weight training algorithm, which improves learning accuracy and accelerates the convergence by increasing the weight of hard examples. Extensive experiments on multiple datasets demonstrate the superiority of our proposed classification method.
Highlights
Cancer is a serious threat to people’s life and health
During the actual diagnosis process, pathologists analyze the overall tissue commonly stained with hematoxylin and eosin (H&E), along with nuclei organization, density, and variability, which requires intense tedious workloads
We propose the re-weight training algorithm based on hard examples to improve the generalization and learning accuracy of the Convolutional Neural Networks (CNNs) model
Summary
Cancer is a serious threat to people’s life and health. According to the newest global cancer statistics [1], an estimated 4.3 million new cancer cases and 2.9 million new cancer deaths occurred in China in 2018. Wang et al [9] proposes a method for identifying metastatic breast cancer based on an ensemble of two CNN architectures with hard-negative mining tactics These researches have not taken advantage of weighting samples differently. According to WHO classification of tumours of the digestive system, common tumor types include adenoma, polyp, adenocarcinoma, gastrointestinal stromal tumor, and neuroendocrine tumor based on pathological features and associated molecular alterations In such a context, the Hard Example Guided CNN is proposed in our work. We demonstrate experimentally that the prediction accuracy of our proposed algorithm reaches 97.36% on colorectal tumor dataset collected by ourselves and 87.32% on breast cancer dataset of ICIAR 2018 and 98.08% on DigestPath 2019, which, to the best of our knowledge, outperforms the state of art methods on this pathological image classification task
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
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
