Fractal Neural Network: A new ensemble of fractal geometry and convolutional neural networks for the classification of histology images

Abstract

Classification of histology images is a task that has been widely explored on recent computer vision researches. The most studied approach for this task has been the application of deep learning through a convolutional neural network (CNN) model. However, the use of CNNs in the context of histological images classification has yet some limitations such as the need of large datasets, the slow training time and the difficult to implement a generalized model able to classify different types of histology tissues. In this paper, we propose an ensemble model based on handcrafted fractal features and deep learning that consists of combining the classification of two CNNs by applying the sum rule. We apply feature extraction to obtain 300 fractal features from different histological datasets. These features are reshaped into a 10×10×3 matrix to compose an artificial image that is given as input to the first CNN. The second CNN model receives as input the correspondent original image. After combining the results of both CNNs, accuracies that range from 89.66% up to 99.62% were obtained from five different datasets. Moreover, our model was able to classify images from datasets with imbalanced classes, without the need for images having the same resolution, and in relative fast training time. We also verified that the obtained results are compatible with the most recent and relevant studies recently published in the context of histology image classification.