Classification of hippocampal cells using machine learning algorithms

Automatic identification and classification of neurons in microslides of nervous tissue is important for analyzing the effects of ionizing radiation on the brain. Manually assessing the condition of CNS cells by a specialist is a time-consuming and subjective process, while machine learning algorithms have shown potential in automating this task. In this work, 81 photographic images of mouse hippocampal samples were used, on which cells without visible damage or critical damage and degenerative cells were divided. The following parameters were calculated for each cell: Area, Roundness and Structural complexity of the nucleus. These parameters were used to train the RandomForestClassifier classifier using the scikit-learn library. The classification accuracy was 68%, with the most significant feature being the structural complexity of the nucleus. The proposed classifier can serve as the basis for an automatic system for analyzing neurons in microslides of the brain.