Data space dimensionality reduction in the problem of diagnosing a thyroid disease

Analysis of a set of data space dimensionality reduction methods in image recognition problems is carried out. A problem of diagnosis of thyroid diseases with the use of images of cytological preparation is investigated. It is shown that the morphological image analysis combined with the method of diffusion maps makes it possible to obtain a higher recognition accuracy in the problem of diagnosis of a thyroid disease than with the previously proposed method based on the Fourier correction spectrum and selection of principal components.