Machine Learning Exploration of Brain Morphological Features and Sensory Measures

Previous investigations have implicated the neuroanatomical basis of sensory systems; however, definitive neuroimaging biomarkers remain elusive. The present study employs machine learning techniques to probe the relationship between brain morphological features and sensory measures of audition, olfaction, taste, and visual contrast sensitivity using a large dataset from the publicly available Human Connectome Project (n = 874). Applying both 5-fold cross-validation and leave-one-out cross-validation methods, performance of several machine learning models was evaluated. Feature selection methods including Random Forest, SelectKBest, Recursive Feature Elimination, and SHapley Additive exPlanations (SHAP) were utilized to identify the most significant neuroanatomical features for sensory performance. Binary classification via machine learning was also conducted to distinguish individuals with high vs. low sensory test scores based on brain morphological features, achieving a satisfactory accuracy of 67% and 71% for olfaction and visual contrast sensitivity, respectively. By integrating machine learning with high-dimensional neuroimaging data, this preliminary study offers new insights into the neural correlates of sensory performance.