Individual cognitive abilities vary widely, yet the structural brain features underlying these differences remain poorly understood. Conventional brain‑behavior analyses typically rely on predefined metrics such as volume or cortical thickness and assume linear relationships with behavior, limiting sensitivity in small datasets. We therefore introduce a pairwise convolutional ranking framework that learns directly from whole‑brain T1‑weighted MRI by making comparative rather than absolute predictions, which remains effective with limited samples (N=108). The method was evaluated on four spatial‑ability tasks; regression analysis achieved high‑accuracy performance across all tasks and outperformed standard linear regression and pointwise deep‑learning baselines. An iterative cortical‑region removal (backward elimination) procedure revealed brain regions common to all four tasks as well as task‑specific regions. These findings demonstrate that pairwise learning can extract meaningful behavioral information from structural MRI, offering both predictive power and interpretable insight into the neuroanatomy of individual differences.
Review