Macular holes are a common eye condition which result in visual impairment. We look at the application of deep convolutional neural networks to the problem of macular hole segmentation. We use the 3D U-Net architecture as a basis and experiment with a number of design variants. Manually annotating and measuring macular holes is time consuming and error prone. Previous automated approaches to macular hole segmentation take minutes to segment a single 3D scan. Our proposed model generates significantly more accurate segmentations in less than a second. We found that an approach of architectural simplification, by greatly simplifying the network capacity and depth, exceeds both expert performance and state-of-the-art models such as residual 3D U-Nets.
With the growing significance of graphs as an effective representation of data in numerous applications, efficient graph analysis using modern machine learning is receiving a growing level of attention. Deep learning approaches often operate over the entire adjacency matrix -- as the input and intermediate network layers are all designed in proportion to the size of the adjacency matrix -- leading to intensive computation and large memory requirements as the graph size increases. It is therefore desirable to identify efficient measures to reduce both run-time and memory requirements allowing for the analysis of the largest graphs possible. The use of reduced precision operations within the forward and backward passes of a deep neural network along with novel specialised hardware in modern GPUs can offer promising avenues towards efficiency. In this paper, we provide an in-depth exploration of the use of reduced-precision operations, easily integrable into the highly popular PyTorch framework, and an analysis of the effects of Tensor Cores on graph convolutional neural networks. We perform an extensive experimental evaluation of three GPU architectures and two widely-used graph analysis tasks (vertex classification and link prediction) using well-known benchmark and synthetically generated datasets. Thus allowing us to make important observations on the effects of reduced-precision operations and Tensor Cores on computational and memory usage of graph convolutional neural networks -- often neglected in the literature.
We show that correlations between the camera used to acquire an image and the class label of that image can be exploited by convolutional neural networks (CNN), resulting in a model that "cheats" at an image classification task by recognizing which camera took the image and inferring the class label from the camera. We show that models trained on a dataset with camera / label correlations do not generalize well to images in which those correlations are absent, nor to images from unencountered cameras. Furthermore, we investigate which visual features they are exploiting for camera recognition. Our experiments present evidence against the importance of global color statistics, lens deformation and chromatic aberration, and in favor of high frequency features, which may be introduced by image processing algorithms built into the cameras.
This paper investigates the application of deep convolutional neural networks with prohibitively small datasets to the problem of macular edema segmentation. In particular, we investigate several different heavily regularized architectures. We find that, contrary to popular belief, neural architectures within this application setting are able to achieve close to human-level performance on unseen test images without requiring large numbers of training examples. Annotating these 3D datasets is difficult, with multiple criteria required. It takes an experienced clinician two days to annotate a single 3D image, whereas our trained model achieves similar performance in less than a second. We found that an approach which uses targeted dataset augmentation, alongside architectural simplification with an emphasis on residual design, has acceptable generalization performance - despite relying on fewer than 15 training examples.
Handling large corpuses of documents is of significant importance in many fields, no more so than in the areas of crime investigation and defence, where an organisation may be presented with a large volume of scanned documents which need to be processed in a finite time. However, this problem is exacerbated both by the volume, in terms of scanned documents and the complexity of the pages, which need to be processed. Often containing many different elements, which each need to be processed and understood. Text recognition, which is a primary task of this process, is usually dependent upon the type of text, being either handwritten or machine-printed. Accordingly, the recognition involves prior classification of the text category, before deciding on the recognition method to be applied. This poses a more challenging task if a document contains both handwritten and machine-printed text. In this work, we present a generic process flow for text recognition in scanned documents containing mixed handwritten and machine-printed text without the need to classify text in advance. We realize the proposed process flow using several open-source image processing and text recognition packages1. The evaluation is performed using a specially developed variant, presented in this work, of the IAM handwriting database, where we achieve an average transcription accuracy of nearly 80% for pages containing both printed and handwritten text.
While deep learning has seen many recent applications to drug discovery, most have focused on predicting activity or toxicity directly from chemical structure. Phenotypic changes exhibited in cellular images are also indications of the mechanism of action (MoA) of chemical compounds. In this paper, we show how pre-trained convolutional image features can be used to assist scientists in discovering interesting chemical clusters for further investigation. Our method reduces the dimensionality of raw fluorescent stained images from a high throughput imaging (HTI) screen, producing an embedding space that groups together images with similar cellular phenotypes. Running standard unsupervised clustering on this embedding space yields a set of distinct phenotypic clusters. This allows scientists to further select and focus on interesting clusters for downstream analyses. We validate the consistency of our embedding space qualitatively with t-sne visualizations, and quantitatively by measuring embedding variance among images that are known to be similar. Results suggested the usefulness of our proposed workflow using deep learning and clustering and it can lead to robust HTI screening and compound triage.
The detection of vascular structures from noisy images is a fundamental process for extracting meaningful information in many applications. Most well-known vascular enhancing techniques often rely on Hessian-based filters. This paper investigates the feasibility and deficiencies of detecting curve-like structures using a Hessian matrix. The main contribution is a novel enhancement function, which overcomes the deficiencies of established methods. Our approach has been evaluated quantitatively and qualitatively using synthetic examples and a wide range of real 2D and 3D biomedical images. Compared with other existing approaches, the experimental results prove that our proposed approach achieves high-quality curvilinear structure enhancement.
Graphs are a commonly used construct for representing relationships between elements in complex high dimensional datasets. Many real-world phenomenon are dynamic in nature, meaning that any graph used to represent them is inherently temporal. However, many of the machine learning models designed to capture knowledge about the structure of these graphs ignore this rich temporal information when creating representations of the graph. This results in models which do not perform well when used to make predictions about the future state of the graph -- especially when the delta between time stamps is not small. In this work, we explore a novel training procedure and an associated unsupervised model which creates graph representations optimised to predict the future state of the graph. We make use of graph convolutional neural networks to encode the graph into a latent representation, which we then use to train our temporal offset reconstruction method, inspired by auto-encoders, to predict a later time point -- multiple time steps into the future. Using our method, we demonstrate superior performance for the task of future link prediction compared with none-temporal state-of-the-art baselines. We show our approach to be capable of outperforming non-temporal baselines by 38% on a real world dataset.
A wide range of biomedical applications requires enhancement, detection, quantification and modelling of curvilinear structures in 2D and 3D images. Curvilinear structure enhancement is a crucial step for further analysis, but many of the enhancement approaches still suffer from contrast variations and noise. This can be addressed using a multiscale approach that produces a better quality enhancement for low contrast and noisy images compared with a single-scale approach in a wide range of biomedical images. Here, we propose the Multiscale Top-Hat Tensor (MTHT) approach, which combines multiscale morphological filtering with a local tensor representation of curvilinear structures in 2D and 3D images. The proposed approach is validated on synthetic and real data and is also compared to the state-of-the-art approaches. Our results show that the proposed approach achieves high-quality curvilinear structure enhancement in synthetic examples and in a wide range of 2D and 3D images.
We introduce style augmentation, a new form of data augmentation based on random style transfer, for improving the robustness of convolutional neural networks (CNN) over both classification and regression based tasks. During training, our style augmentation randomizes texture, contrast and color, while preserving shape and semantic content. This is accomplished by adapting an arbitrary style transfer network to perform style randomization, by sampling input style embeddings from a multivariate normal distribution instead of inferring them from a style image. In addition to standard classification experiments, we investigate the effect of style augmentation (and data augmentation generally) on domain transfer tasks. We find that data augmentation significantly improves robustness to domain shift, and can be used as a simple, domain agnostic alternative to domain adaptation. Comparing style augmentation against a mix of seven traditional augmentation techniques, we find that it can be readily combined with them to improve network performance. We validate the efficacy of our technique with domain transfer experiments in classification and monocular depth estimation, illustrating consistent improvements in generalization.