Predicting Sensation-Seeking from Resting-State fMRI: The Need for Age-Specific Models

Sensation-seeking, as a sub-dimension of impulsivity, reflects an individual’s tendency for novel and stimulating experiences. High sensation-seeking often involves novelty-seeking and risk-taking, which may lead to risky behaviors such as reckless driving, addiction, and substance use, significantly impacting individuals’ social and personal functioning. Recent studies have utilized functional magnetic resonance imaging (fMRI) to study the neural mechanisms of sensation-seeking. However, the influence of demographic factors like age on the neural patterns associated with sensation-seeking remains unexplored. In this paper, we predicted sensation-seeking scores from resting-state fMRI data in a large-scale study involving 131 male participants aged 20 to 79. By developing separate predictive models for different age groups (age-specific model), we achieved an R2 of 0.38 between the actual and predicted sensation-seeking scores. The proposed age-specific model significantly outperformed the baseline that fitted a single model for the entire dataset, indicating the importance of demographic factors in understanding the neural correlates of sensation-seeking. We identified key brain regions associated with sensation-seeking, including the prefrontal areas, cerebellum, subcortical regions, parietal lobe, and cerebral cortex areas. Notably, the brain connectivity patterns linked to sensation-seeking varied across age groups, further demonstrating the age-related variation in neural correlates of sensation-seeking. Our proposed age-specific modeling of sensation-seeking acknowledges the diversity in neural patterns across different aging stages and potentially offers more accurate insights into the neural correlates of sensation-seeking.

Recommended citation: Zishen Li, Bishal Lamichhane, Ankit Patel, Ramiro Salas, Nidal Moukaddam, and Ashutosh S (2024), Predicting Sensation-Seeking from Resting-State fMRI: The Need for Age-Specific Models, IEEE BHI 2024.
Download Paper