Biological insights and novel biomarker discovery through deep learning approaches in breast cancer histopathology

Abstract

Breast cancer remains a highly prevalent disease with considerable inter- and intra-tumoral heterogeneity complicating prognostication and treatment decisions. The utilization and depth of genomic, transcriptomic and proteomic data for cancer has exploded over recent times and the addition of spatial context to this information, by understanding the correlating morphologic and spatial patterns of cells in tissue samples, has created an exciting frontier of research, histo-genomics. At the same time, deep learning (DL), a class of machine learning algorithms employing artificial neural networks, has rapidly progressed in the last decade with a confluence of technical developments - including the advent of modern graphic processing units (GPU), allowing efficient implementation of increasingly complex architectures at scale; advances in the theoretical and practical design of network architectures; and access to larger datasets for training - all leading to sweeping advances in image classification and object detection. In this review, we examine recent developments in the application of DL in breast cancer histology with particular emphasis of those producing biologic insights or novel biomarkers, spanning the extraction of genomic information to the use of stroma to predict cancer recurrence, with the aim of suggesting avenues for further advancing this exciting field.

Fig. 1. Example neural network architectures.

a Basic neural network demonstrating input nodes with dimensionality of 8, two hidden layers that involve dimension expansion and subsequent reduction with a final output node. Each input node is connected to each of the hidden nodes (a fully connected network), with lines connecting nodes representing weights applied to a source to reach a destination. The above representation can also be visualized as a weight matrix, translating from the dimensionality of the input nodes to the dimensionality of the output nodes. b A pictorial the training process – inputs are fed to a DL network, predictions are made and compared to ground truth labels and parameters are updated in a loop. c A convolutional network architecture demonstrating the typical backbone of pooling layers followed by convolutional layers. Note that convolutions increase the number of filters while reducing dimensionality in the x/y dimensions (https://alexlenail.me/NN-SVG/).

Fig. 2. Image tasks in computer vision.

Examples of (a) image classification, in which the task is one of classifying an image as one of 4 different types of animals (b) object detection, in which not only are objects classified but also identified in the image with boxes around their respective location and (c) image segmentation, where every pixel in the image is translated into some label, here the vessels, airways and contours of a lung. Sources: (a) the author’s photographs of dog (DM) and Jonesy the cat (AA). b https://commons.wikimedia.org/wiki/File:Detected-with-YOLO--Schreibtisch-mit-Objekten.jpg (c) https://commons.wikimedia.org/wiki/File:3D_CT_of_thorax.jpg.

Fig. 3. Deep learning features from WSI.

Categories of features extracted from deep learning systems in breast cancer are illustrated, including invasive carcinoma architecture, cellular makeup of the microenvironment, nuclei features (including segmentation, orientation and nucleoli prominence), and stroma characteristics including collagen fiber orientation. On the right, references of studies exploring the detection of the different categories of features.