Novel methodology for detecting and localizing cancer area in histopathological images based on overlapping patches

Cancer disease is one of the most important pathologies in the world, as it causes the death of millions of people, and the cure of this disease is limited in most cases. Rapid spread is one of the most important features of this disease, so many efforts are focused on its early-stage detection and localization. Medicine has made numerous advances in the recent decades with the help of artificial intelligence (AI), reducing costs and saving time. In this paper, deep learning models (DL) are used to present a novel method for detecting and localizing cancerous zones in WSI images, using tissue patch overlay to improve performance results. A novel overlapping methodology is proposed and discussed, together with different alternatives to evaluate the labels of the patches overlapping in the same zone to improve detection performance. The goal is to strengthen the labeling of different areas of an image with multiple overlapping patch testing. The results show that the proposed method improves the traditional framework and provides a different approach to cancer detection. The proposed method, based on applying 3x3 step 2 average pooling filters on overlapping patch labels, provides a better result with a 12.9% correction percentage for misclassified patches on the HUP dataset and 15.8% on the CINIJ dataset. In addition, a filter is implemented to correct isolated patches that were also misclassified. Finally, a CNN decision threshold study is performed to analyze the impact of the threshold value on the accuracy of the model. The alteration of the threshold decision along with the filter for isolated patches and the proposed method for overlapping patches, corrects about 20% of the patches that are mislabeled in the traditional method. As a whole, the proposed method achieves an accuracy rate of 94.6%. The code is available at https://github.com/sergioortiz26/Cancer_overlapping_filter_WSI_images.

Self-attention random forest for breast cancer image classification

Introduction

Early screening and diagnosis of breast cancer can not only detect hidden diseases in time, but also effectively improve the survival rate of patients. Therefore, the accurate classification of breast cancer images becomes the key to auxiliary diagnosis.

Methods

In this paper, on the basis of extracting multi-scale fusion features of breast cancer images using pyramid gray level co-occurrence matrix, we present a Self-Attention Random Forest (SARF) model as a classifier to explain the importance of fusion features, and can perform adaptive refinement processing on features, thus, the classification accuracy can be improved. In addition, we use GridSearchCV technique to optimize the hyperparameters of the model, which greatly avoids the limitation of artificially selected parameters.

Results

To demonstrate the effectiveness of our method, we perform validation on the breast cancer histopathological image-BreaKHis. The proposed method achieves an average accuracy of 92.96% and a micro average AUC value of 0.9588 for eight-class classification, and an average accuracy of 97.16% and an AUC value of 0.9713 for binary classification on BreaKHis dataset.

Discussion

For the sake of verify the universality of the proposed model, we also conduct experiments on MIAS dataset. An excellent average classification accuracy is 98.79% on MIAS dataset. Compared to other state-of-the-art methods, the experimental results demonstrate that the performance of the proposed method is superior to that of others. Furthermore, we can analyze the influence of different types of features on the proposed model, and provide theoretical basis for further optimization of the model in the future.