Recent popularity of deep learning methods inspires to find new applications for them. One of promising areas is medical diagnosis support, especially analysis of medical images. In this paper we explore the possibility of using Deep Convolutional Neural Networks (DCNN) for detection of stenoses in angiographic images. One of the biggest difficulties is a need for large amounts of labelled data required to properly train deep model. We demonstrate how to overcome this difficulty by using generative model producing artificial data. Test results shows that DCNN trained on artificial data and fine-tuned using real samples can achieve up to 90% accuracy, exceeding results obtained by both traditional, feed-forward networks and networks trained using real data only.

The paper discusses regularization properties of artificial data for deep learning. Artificial datasets allow to train neural networks in the case of a real data shortage. It is demonstrated that the artificial data generation process, described as injecting noise to high-level features, bears several similarities to existing regularization methods for deep neural networks. One can treat this property of artificial data as a kind of “deep” regularization. It is thus possible to regularize hidden layers of the network by generating the training data in a certain way.

Abstract. The development of the Internet technology has caused telemedicine diagnosis systems to be commonly available. The most crucial aspect of the patient's safety in such systems is a problem of safe diagnosis. There are two factors at the stake here. The first one is a human factor in the form of user; it is especially severe when the user is not a physician. The second one is the accuracy of the diagnosis process. The best way to handle possible diagnosis errors is to apply measures of sensitivity, specificity and ROC (Receiver Operating Characteristic) curve. We examine these measures on a sample diagnosis system against a real-life data. It turns out that values of these measures are strictly associated with another measure: an indication threshold. Therefore, the accuracy of diagnosis may be to a large extent determined by the chosen threshold. We propose several methods for minimizing the impact of this factor.

Classification methods have multiple applications, with medical diagnosis being one of the most common. A powerful way to improve classification quality is to combine single classifiers into an ensemble. One of the approaches for creating such ensembles is to combine class rankings from base classifiers. In this paper, two rank-based ensemble methods are studied: Highest Rank and Borda Count. Furthermore, the effect of applying class rank threshold to these methods is analyzed. We performed tests using real-life medical data. It turns out that specificity of data domain can affect classification quality depending on classifier type.

No abstract