ENSEMBLE OF CLASSIFIERS BASED ON DEEP LEARNING FOR MEDICAL IMAGE RECOGNITION

Abstract

The paper presents special forms of an ensemble of classi ers for analysis of medical images based on application of deep learning. The study analyzes di erent structures of convolutional neural networks applied in the recognition of two types of medical images: dermoscopic images for melanoma and mammograms for breast cancer. Two approaches to ensemble creation are proposed. In the rst approach, the images are processed by a convolutional neural network and the attened vector of image descriptors is subjected to feature selection by applying di erent selection methods. As a result, di erent sets of a limited number of diagnostic features are generated. In the next stage, these sets of features represent input attributes for the classical classi ers: support vector machine, a random forest of decision trees, and softmax. By combining di erent selection methods with these classi ers an ensemble classi cation system is created and integrated by majority voting. In the second approach, di erent structures of convolutional neural networks are directly applied as the members of the ensemble. The e ciency of the proposed classi cation systems is investigated and compared to medical data representing dermoscopic images of melanoma and breast cancer mammogram images. Thanks to fusion of the results of many classi ers forming an ensemble, accuracy and all other quality measures have been signi cantly increased for both types of medical images.