A Pyramid Architecture-Based Deep Learning Framework for Breast Cancer Detection

Abstract

Breast cancer diagnosis is a critical step in clinical decision making, and this is achieved by making a pathological slide and gives a decision by the doctors, which is the method of final decision making for cancer diagnosis. Traditionally, the doctors usually check the pathological images by visual inspection under the microscope. Whole-slide images (WSIs) have supported the state-of-the-art diagnosis results and have been admitted as the gold standard clinically. However, this task is time-consuming and labour-intensive, and all of these limitations make low efficiency in decision making. Medical image processing protocols have been used for this task during the last decades and have obtained satisfactory results under some conditions; especially in the deep learning era, it has exhibited the advantages than those in the shallow learning period. In this paper, we proposed a novel breast cancer region mining framework based on deep pyramid architecture from multilevel and multiscale breast pathological WSIs. We incorporate the tissue- and cell-level information together and integrate these into a LSTM model for the final sequence modelling, which successfully keeps the WSIs' integration and is not mentioned by the prevalence frameworks. The experiment results demonstrated that our proposed framework greatly improved the detection accuracy than that only using tissue-level information.

Figure 1

Data set in this paper. (a) Is the Camelyon 2017 whole-slide images [7]; (b) is the TMI 2015 data set from Xu et al. [31].

Figure 2

GAN model for pathological image generation.

Figure 3

The scheme of pyramid deconvolution network framework for breast cancer detection.

Figure 4

The backbone network architecture of the DeconvNet. We make a slight change on kernel size and stride for different input image scales.

Figure 5

Long short-term memory enhanced channeled fully convolutional network pipeline with a dimension shuffle layer.

Figure 6

Pathological images and their detection results. The subimages (a), (c), (e), and (g) are the original pathological images; (b) and (f) are the heat maps generated from our framework.

Figure 7

Comparisons of the segmentation results among different methods. (a–d) Are different views of the WSIs; from 1 to 4 are the segmentation results of original image, FCN, U-net, and our proposed method.

Algorithm 1

Tissue-level pathological RoI extraction.