Innovative statistical approaches: the use of neural networks reduces the sample size in the splenectomy-MCAO mouse model

Scientific paper - Original scientific paper

• Author: Dominik Romić
• Author: Monika Berecki
• Author: Sanja Srakočić
• Author: Paula Josić Dominović
• Author: Helena Justić
• Author: Dominik Hamer
• Author: Daniela Petrinec
• Author: Marina Radmilović
• Author: Branimir Hackenberger
• Author: Srećko Gajović
• Author: Anton Glasnović
• Author's institution: University of Zagreb; School of Medicine; (Croatian Institiute for Brain Research)
• Author's institution: University of Zagreb; School of Medicine; (Department of Histology and Embryology)
• Scientific / art field, discipline and subdiscipline: BIOMEDICINE AND HEALTHCARE; Basic Medical Sciences; Cytology, Histology and Embryology
• Scientific / art field, discipline and subdiscipline: BIOMEDICINE AND HEALTHCARE; Basic Medical Sciences; Neuroscience

Abstract (english)

Aim: To compare the effectiveness of artificial neural network (ANN) and traditional statistical analysis on identical data sets within the splenectomy-middle carotid artery occlusion (MCAO) mouse model.

Methods: Mice were divided into the splenectomized (SPLX) and sham-operated (SPLX-sham) group. A splenectomy was conducted 14 days before middle carotid artery occlusion (MCAO). Magnetic resonance imaging (MRI), bioluminescent imaging, neurological scoring (NS), and histological analysis, were conducted at two, four, seven, and 28 days after MCAO. Frequentist statistical analyses and ANN analysis employing a multi-layer perceptron architecture were performed to assess the probability of discriminating between SPLX and SPLX-sham mice.

Results: Repeated measures ANOVA showed no significant differences in body weight (F (5, 45)=0.696, P=0.629), NS (F (2.024, 18.218)=1.032, P=0.377) and brain infarct size on MRI between the SPLX and SPLX-sham groups post-MCAO (F (2, 24)=0.267, P=0.768). ANN analysis was employed to predict SPLX and SPL-sham classes. The highest accuracy in predicting SPLX class was observed when the model was trained on a data set containing all variables (0.7736±0.0234). For SPL-sham class, the highest accuracy was achieved when it was trained on a data set excluding the variable combination MR contralateral/animal mass/NS (0.9284±0.0366).

Conclusion: This study validated the neuroprotective impact of splenectomy in an MCAO model using ANN for data analysis with a reduced animal sample size, demonstrating the potential for leveraging advanced statistical methods to minimize sample sizes in experimental biomedical research.

• Keywords (english): Animals; Disease Models, Animal; Infarction, Middle Cerebral Artery / diagnostic imaging; Infarction, Middle Cerebral Artery / surgery; Magnetic Resonance Imaging; Male; Mice; Neural Networks, Computer; Sample Size; Splenectomy / methods
• Language: english
• Publication type: Scientific paper - Original scientific paper
• Publication status: Published
• Peer review: Peer review
• Publication version: Published version
• Journal title: Croatian Medical Journal
• Numbering: vol. 65, no. 2, pp. 122-137
• p-ISSN: 0353-9504
• e-ISSN: 1332-8166
• DOI: https://doi.org/10.3325/cmj.2024.65.122
• URN:NBN: urn:nbn:hr:105:242122
• Publication: 2024-04-30
• Document URL: http://www.cmj.hr/2024/65/2/38706238.htm
• Type of resource: Text
• Access conditions: Open access
• Created on: 2024-06-20 11:04:53