Hyperspectral images neural network analysis of unstained micropreparations

Abstract

The article presents the results of a study of hyperspectral imaging in microscopy to assess pathological changes in unstained medical micropreparations. Hyperspectral imaging was carried out using a system of synchronous shooting and movement of a movable table combined with a stepper motor. To improve the quality of the obtained images, software correction of the illumination of the spectral channels was used. The classification was carried out by a convolutional neural network. This method may be promising for assessing pathological changes in clinical practice. Experimental studies were carried out on histological preparations with different types of tissues without staining with contrasting medical dyes. To assess the reliability of the classification method, a comparison was made with the standard method using staining of the studied samples.