×

You are using an outdated browser Internet Explorer. It does not support some functions of the site.

Recommend that you install one of the following browsers: Firefox, Opera or Chrome.

Contacts:

+7 961 270-60-01
ivdon3@bk.ru

  • Evaluation of the entropy of fragments of X-ray images of the lungs

    The emergence of digital X-ray machines and the development of cloud storage technology has led to the accumulation of a huge number of medical X-rays, in particular, chest x-rays - fluorography. The accumulated image bases after high-quality preprocessing can be used to train the deep convolutional neural networks that have received the most development in recent years, the trained network performs preliminary binary classification of the incoming flow of images and can be used as a radiologist assistant. For this purpose, it is necessary to adequately train the neural net-work to minimize errors of the first and second kind. A possible approach to improving the efficiency of neural networks, reducing the computational complexity and quality of image classification by the criteria is the use of auxiliary approaches of image preprocessing and preliminary entropy calculation. The article presents an algorithm for the X-ray image preprocessing, its division into fragments and the calculation of the entropy of individual fragments. During the preprocessing, the region of interest with lungs and the spine is selected from the entire image, constituting about 30-40% of the entire image, then the image is divided into a fragment matrix and the entropy of individual fragments is calculated using the Shannon formula, by analyzing individual pixels. By determining the frequency of each of the 255 colors, the total entropy is calculated. The use of entropy for detecting pathologies is based on the assumption of dif-ferences in its values for individual fragments and the overall picture of its distribution between images with the norm and pathologies. Statistical indicators are analyzed: standard deviation of error, variance.

    Keywords: image entropy, fragments, deep convolutional neural network, machine learning, x-rays images, computational experiment, matrix of elements, image preprocessing, statistical analysis, binary classification