High-Resolution Breast Cancer Screening with Multi-View Deep Convolutional Neural Networks

Abstract

Advances in deep learning for natural images have prompted a surge of interest in applying similar techniques to medical images. The majority of the initial attempts focused on replacing the input of a deep convolutional neural network with a medical image, which does not take into consideration the fundamental differences between these two types of images. Specifically, fine details are necessary for detection in medical images, unlike in natural images where coarse structures matter most. This difference makes it inadequate to use the existing network architectures developed for natural images, because they work on heavily downscaled images to reduce the memory requirements. This hides details necessary to make accurate predictions. Additionally, a single exam in medical imaging often comes with a set of views which must be fused in order to reach a correct conclusion. In our work, we propose to use a multi-view deep convolutional neural network that handles a set of high-resolution medical images. We evaluate it on largescale mammography-based breast cancer screening (BI-RADS prediction) using 886,000 images. We focus on investigating the impact of the training set size and image size on the prediction accuracy. Our results highlight that performance increases with the size of training set, and that the best performance can only be achieved using the original resolution. In the reader study, performed on a random subset of the test set, we confirmed the efficacy of our model, which achieved performance comparable to a committee of radiologists when presented with the same data.

Beyond Fine-tuning: Classifying High Resolution Mammograms using Function-Preserving Transformations

“…At the algorithmic level, breast diagnosis can be casted as a classification task, in which several developments using deep ConvNets have been reported e.g. [10], [11], [12], [13]. However despite the rapid development of DL based techniques, mammography classification remains an unsolved problem.…”

Knowledge-Guided Multiview Deep Curriculum Learning for Elbow Fracture Classification

“…Another group of multiview learning algorithms explore Multiple Kernel Learning (MKL), which was originally proposed to restrict the search space of kernels [4,6]. Recent work on multiview learning based modeling shows promising effects for medical fields such as bone fracture and breast cancer detection [13,17,8].…”

Improving Breast Cancer Detection using Symmetry Information with Deep Learning

“…Breast cancer screening has shown a reduction in mortality rate of between 40% and 45% for women who were undergoing mammogram screening regularly [4]. However, mammogram screening has drawbacks due to False Positive (FP) recalls, such as FP biopsy and cost associated with the unnecessary follow up [5]. Therefore, it is necessary to increase sensitivity for early stage detection and increase specificity to reduce FP detection.…”

“…DL has been explored for Digital Mammogram (DM) image analysis. Some of them work directly on the whole image [5], [3], and others focused on patch based [8]. [5] proposed a multi-view single stage CNN breast mammogram classification that works at original resolution.…”

“…Some of them work directly on the whole image [5], [3], and others focused on patch based [8]. [5] proposed a multi-view single stage CNN breast mammogram classification that works at original resolution. To address memory issue, aggressive convolution and pooling layers with stride greater than one were proposed.…”