SVM Classification of Endurance Athletes and Controls: A Model with Practical Implications for Athletic Performance

The gray matter plasticity in athletes is crucial for understanding the neural mechanisms of sports. However, traditional MRI data processing methods face limitations due to the large volume of data from voxel - based analyses. Support vector machine (SVM) may offer advantages and could serve as a potential solution. MRI scans were performed to obtain gray matter volume data for each voxel from 20 endurance athletes and 20 control subjects. We applied a SVM classification algorithm with leave-one-out cross-validation to construct the classification model and calculate accuracy. Brain regions with significant contribution weights during classification were identified as regions of interest (ROI). We then extracted their gray matter volume signals and conducted a correlation analysis with endurance capacity (maximum oxygen uptake) to evaluate the relationship between plasticity of these brain regions and athletic performance. The SVM classification algorithm using gray matter volume (GMV) as a feature achieved excellent performance in distinguishing between endurance athletes and control subjects, with a total accuracy and balanced accuracy of percent of 82.5. The differing brain regions were located in extensive cortical and subcortical areas. Specifically, the left precuneus, the left superior frontal orbital gyrus, the left thalamus, as well as the right thalamus exhibited a positive correlation with maximal oxygen uptake ($V O_{2}$ max), while the right hippocampus and right parahippocampal showed a negative correlation. This study shows that SVM classification can distinguish endurance athletes from controls using GMV as feature. Key brain regions correlate with maximum oxygen uptake, suggesting that GMV plasticity may correlate with endurance capacity. These findings provide insights for future research and training strategies.