The Use of Kohonen's Neural Network in Clustering of Parasitological Micrographs

This article examines in detail a promising direction for the application of a neural network such as the Kohonen self-organizing map ( SOM ) in the field of applied parasitology. It considers theoretical justification of using this type of artificial neural networks for analyzing parasitological samples, with an emphasis on their ability to effectively cluster without preliminary data labeling. Various clustering quality metrics typical for self-organizing maps are presented and critically analyzed, including topological error, quantization error. As part of the experimental part of the work, a comprehensive study was conducted, including the creation and preparation of a specialized data set consisting of micrographs of the biohelminth Trichinella Spiralis . To implement the tasks set, a Kohonen network with an optimized architecture was developed and modeled in Python using modern machine learning libraries. Based on the results obtained, reasonable conclusions were formulated about the performance of the developed model, its accuracy and efficiency .