Automatic Identification of Neurons in Complex Cultures Using Image Processing Algorithms
-
Gerard Villarroya-Pique [1] ; Víctor M. Gonzáles [1] ; Esther Serrano-Pertierra [1] ; Antonello Novelli [1] ; M. Teresa Fernández-Sánchez [1] ; Angel Rio-Alvarez [1]
-
[1] Universidad de Oviedo
Universidad de Oviedo
Oviedo, España
-
- Localización: CASEIB 2024. Libro de Actas del XLII Congreso Anual de la Sociedad Española de Ingeniería Biomédica, 2024, ISBN 978-84-09-67332-2, págs. 9-12
- Idioma: inglés
- Enlaces
-
Resumen
In the study of neurodegenerative diseases, and more generally in neuroscience, cell culture viability tests are a very common technique. These tests are based on chemical staining methods designed to target only the living cells, but at the expense of killing the neurons in the mid-term. Researchers then have to manually count the number of living neurons identified in the microscope images obtained using fluorescence techniques. This manual task is very tedious and prone to errors. Computer vision may help to solve these two issues. Many deep learning (DL)–based algorithms have been developed to identify neurons in a culture. When the cultures are particularly complex due to their variability, as well as the presence of other types of elements, the use of supervised DL models is required. This kind of technique requires a large dataset of images manually labeled by experts for the models to perform accurately. The manual labeling of neurons can introduce researcher bias, optimizing the models for a single type of culture. This work addresses the automatic identification of neurons in fluorescence images of neuronal cultures using classical algorithms, thus avoiding the need for manual labeling of images. The method employs a peak extraction algorithm to locate the centroids of the neurons in the image. Results indicate that the method can count neurons with reliability similar to that achieved by experts.
