In this work, we focus on the problem of image instance retrieval with deep descriptors extracted from pruned Convolutional Neural Networks (CNN). The objective is to heavily prune convolutional edges while maintaining retrieval performance. To this end, we introduce both data-independent and data-dependent heuristics to prune convolutional edges, and evaluate their performance across various compression rates with different deep descriptors over several benchmark datasets. Further, we present an end-to-end framework to fine-tune the pruned network, with a triplet loss function specially designed for the retrieval task. We show that the combination of heuristic pruning and fine-tuning offers 5x compression rate without considerable loss in retrieval performance.
The stochastic gradient descent (SGD) method is most widely used for deep neural network (DNN) training. However, the method does not always converge to a flat minimum of the loss surface that can demonstrate high generalization capability. Weight noise injection has been extensively studied for finding flat minima using the SGD method. We devise a new weight-noise injection-based SGD method that adds symmetrical noises to the DNN weights. The training with symmetrical noise evaluates the loss surface at two adjacent points, by which convergence to sharp minima can be avoided. Fixed-magnitude symmetric noises are added to minimize training instability. The proposed method is compared with the conventional SGD method and previous weight-noise injection algorithms using convolutional neural networks for image classification. Particularly, performance improvements in large batch training are demonstrated. This method shows superior performance compared with conventional SGD and weight-noise injection methods regardless of the batch-size and learning rate scheduling algorithms.
Prostate cancer is the most prevalent cancer among men in Western countries, with 1.1 million new diagnoses every year. The gold standard for the diagnosis of prostate cancer is a pathologists' evaluation of prostate tissue. To potentially assist pathologists deep-learning-based cancer detection systems have been developed. Many of the state-of-the-art models are patch-based convolutional neural networks, as the use of entire scanned slides is hampered by memory limitations on accelerator cards. Patch-based systems typically require detailed, pixel-level annotations for effective training. However, such annotations are seldom readily available, in contrast to the clinical reports of pathologists, which contain slide-level labels. As such, developing algorithms which do not require manual pixel-wise annotations, but can learn using only the clinical report would be a significant advancement for the field. In this paper, we propose to use a streaming implementation of convolutional layers, to train a modern CNN (ResNet-34) with 21 million parameters end-to-end on 4712 prostate biopsies. The method enables the use of entire biopsy images at high-resolution directly by reducing the GPU memory requirements by 2.4 TB. We show that modern CNNs, trained using our streaming approach, can extract meaningful features from high-resolution images without additional heuristics, reaching similar performance as state-of-the-art patch-based and multiple-instance learning methods. By circumventing the need for manual annotations, this approach can function as a blueprint for other tasks in histopathological diagnosis. The source code to reproduce the streaming models is available at https://github.com/DIAGNijmegen/pathology-streaming-pipeline .
Convolutional Neural Networks (CNNs) have gained a remarkable success on many real-world problems in recent years. However, the performance of CNNs is highly relied on their architectures. For some state-of-the-art CNNs, their architectures are hand-crafted with expertise in both CNNs and the investigated problems. To this end, it is difficult for researchers, who have no extended expertise in CNNs, to explore CNNs for their own problems of interest. In this paper, we propose an automatic architecture design method for CNNs by using genetic algorithms, which is capable of discovering a promising architecture of a CNN on handling image classification tasks. The proposed algorithm does not need any pre-processing before it works, nor any post-processing on the discovered CNN, which means it is completely automatic. The proposed algorithm is validated on widely used benchmark datasets, by comparing to the state-of-the-art peer competitors covering eight manually designed CNNs, four semi-automatically designed CNNs and additional four automatically designed CNNs. The experimental results indicate that the proposed algorithm achieves the best classification accuracy consistently among manually and automatically designed CNNs. Furthermore, the proposed algorithm also shows the competitive classification accuracy to the semi-automatic peer competitors, while reducing 10 times of the parameters. In addition, on the average the proposed algorithm takes only one percentage of computational resource compared to that of all the other architecture discovering algorithms.
Imaging phantoms are test patterns used to measure image quality in computer tomography (CT) systems. A new phantom platform (Mercury Phantom, Gammex) provides test patterns for estimating the task transfer function (TTF) or noise power spectrum (NPF) and simulates different patient sizes. Determining which image slices are suitable for analysis currently requires manual annotation of these patterns by an expert, as subtle defects may make an image unsuitable for measurement. We propose a method of automatically classifying these test patterns in a series of phantom images using deep learning techniques. By adapting a convolutional neural network based on the VGG19 architecture with weights trained on ImageNet, we use transfer learning to produce a classifier for this domain. The classifier is trained and evaluated with over 3,500 phantom images acquired at a university medical center. Input channels for color images are successfully adapted to convey contextual information for phantom images. A series of ablation studies are employed to verify design aspects of the classifier and evaluate its performance under varying training conditions. Our solution makes extensive use of image augmentation to produce a classifier that accurately classifies typical phantom images with 98% accuracy, while maintaining as much as 86% accuracy when the phantom is improperly imaged.
Diabetic Retinopathy is the leading cause of blindness in the working-age population of the world. The main aim of this paper is to improve the accuracy of Diabetic Retinopathy detection by implementing a shadow removal and color correction step as a preprocessing stage from eye fundus images. For this, we rely on recent findings indicating that application of image dehazing on the inverted intensity domain amounts to illumination compensation. Inspired by this work, we propose a Shadow Removal Layer that allows us to learn the pre-processing function for a particular task. We show that learning the pre-processing function improves the performance of the network on the Diabetic Retinopathy detection task.
Despite the recent advances in automatically describing image contents, their applications have been mostly limited to image caption datasets containing natural images (e.g., Flickr 30k, MSCOCO). In this paper, we present a deep learning model to efficiently detect a disease from an image and annotate its contexts (e.g., location, severity and the affected organs). We employ a publicly available radiology dataset of chest x-rays and their reports, and use its image annotations to mine disease names to train convolutional neural networks (CNNs). In doing so, we adopt various regularization techniques to circumvent the large normal-vs-diseased cases bias. Recurrent neural networks (RNNs) are then trained to describe the contexts of a detected disease, based on the deep CNN features. Moreover, we introduce a novel approach to use the weights of the already trained pair of CNN/RNN on the domain-specific image/text dataset, to infer the joint image/text contexts for composite image labeling. Significantly improved image annotation results are demonstrated using the recurrent neural cascade model by taking the joint image/text contexts into account.
Content-Based Image Retrieval (CBIR) locates, retrieves and displays images alike to one given as a query, using a set of features. It demands accessible data in medical archives and from medical equipment, to infer meaning after some processing. A problem similar in some sense to the target image can aid clinicians. CBIR complements text-based retrieval and improves evidence-based diagnosis, administration, teaching, and research in healthcare. It facilitates visual/automatic diagnosis and decision-making in real-time remote consultation/screening, store-and-forward tests, home care assistance and overall patient surveillance. Metrics help comparing visual data and improve diagnostic. Specially designed architectures can benefit from the application scenario. CBIR use calls for file storage standardization, querying procedures, efficient image transmission, realistic databases, global availability, access simplicity, and Internet-based structures. This chapter recommends important and complex aspects required to handle visual content in healthcare.
Traditional image tagging and retrieval algorithms have limited value as a result of being trained with heavily curated datasets. These limitations are most evident when arbitrary search words are used that do not intersect with training set labels. Weak labels from user generated content (UGC) found in the wild (e.g., Google Photos, FlickR, etc.) have an almost unlimited number of unique words in the metadata tags. Prior work on word embeddings successfully leveraged unstructured text with large vocabularies, and our proposed method seeks to apply similar cost functions to open source imagery. Specifically, we train a deep learning image tagging and retrieval system on large scale, user generated content (UGC) using sampling methods and joint optimization of word embeddings. By using the Yahoo! FlickR Creative Commons (YFCC100M) dataset, such an approach builds robustness to common unstructured data issues that include but are not limited to irrelevant tags, misspellings, multiple languages, polysemy, and tag imbalance. As a result, the final proposed algorithm will not only yield comparable results to state of the art in conventional image tagging, but will enable new capability to train algorithms on large, scale unstructured text in the YFCC100M dataset and outperform cited work in zero-shot capability.
Predictive Physics has been historically based upon the development of mathematical models that describe the evolution of a system under certain external stimuli and constraints. The structure of such mathematical models relies on a set of hysical hypotheses that are assumed to be fulfilled by the system within a certain range of environmental conditions. A new perspective is now raising that uses physical knowledge to inform the data prediction capability of artificial neural networks. A particular extension of this data-driven approach is Physically-Guided Neural Networks with Internal Variables (PGNNIV): universal physical laws are used as constraints in the neural network, in such a way that some neuron values can be interpreted as internal state variables of the system. This endows the network with unraveling capacity, as well as better predictive properties such as faster convergence, fewer data needs and additional noise filtering. Besides, only observable data are used to train the network, and the internal state equations may be extracted as a result of the training processes, so there is no need to make explicit the particular structure of the internal state model. We extend this new methodology to continuum physical problems, showing again its predictive and explanatory capacities when only using measurable values in the training set. We show that the mathematical operators developed for image analysis in deep learning approaches can be used and extended to consider standard functional operators in continuum Physics, thus establishing a common framework for both. The methodology presented demonstrates its ability to discover the internal constitutive state equation for some problems, including heterogeneous and nonlinear features, while maintaining its predictive ability for the whole dataset coverage, with the cost of a single evaluation.