Predicting the degree of trait emotional empathy from cortical features using surface-based morphometry.

The link between neuroanatomy and empathy was a primary aim of many studies in the past. However, these studies are often limited by partial volume effects inherent to Voxel-Based Morphometry (VBM), which hinder the accurate assessment of cortical features. To overcome these limitations, we used Surface-Based Morphometry (SBM) to investigate the relationship between cortical features such as sulcal depth and trait emotional empathy for the first time. We also tested several hypotheses based on previous research. Our study sample consisted of 62 adults. Trait empathy was measured by the Toronto Empathy Questionnaire. We used methods from machine learning (e.g., Leave-one-out cross-validation – LOO-CV – in conjunction with Sparse Partially Least Squares regression - SPLSR) to evaluate whether cortical features (i.e., gyrification, sulcal depth, and cortical thickness) could accurately predict the degree of trait emotional empathy. We tested the hypotheses of the present study using Linear Mixed Effects models. The Regions of Interest (ROI) tested within these hypotheses included bilateral insula and dorsal Anterior Cingulate Cortex (dACC). A significant negative association was found between the cortical thickness of the left insula and trait emotional empathy score (B = -4.72; 95% CI [-8.52, -0.92]; p = 0.019). In addition, there was a negative association between cortical thickness in the left dACC and trait emotional empathy (B = -5.47; 95% CI [-9.48, -1.46]; p = 0.008). The SPLSR models showed moderate training fit ([Formula: see text] ≈ 60% in gyrification) but failed to generalize to unseen data, with negative cross-validated [Formula: see text] values across all 3 cortical features. While ROI analyses revealed significant negative associations between cortical thickness and trait emotional empathy, machine learning models failed to achieve reliable out-of-sample prediction, likely due to the high-dimensional feature space relative to sample size. The predictive findings should therefore be considered exploratory. Future studies should use larger samples and explore how differences in white matter structure can explain individual differences in empathy.