Scholarlink

<p>Predicting behavioral and personality traits from neuroimaging data requires effective modeling of complex and high-dimensional brain dynamics. Although recent advances in neurocomputations have paved the way for trait prediction models, the attempts remain at a nascent stage, as reflected in the moderate prediction accuracies. To this end, in this work, we propose a novel unified framework that leverages both spatial and temporal information derived from resting-state functional MRI (rs-fMRI) for trait prediction. Task-activation maps are first estimated directly from rs-fMRI, providing spatial representations of task-evoked brain activity without requiring explicit task paradigms. To capture temporal dynamics, MultiRocket, an efficient time-series feature-extraction method, is employed, encoding trait-specific patterns in rs-fMRI signals. The spatial and temporal features are then fused to form a unique unified spatio-temporal representation that is subsequently provided to an ensemble framework to predict cognitive traits, viz., reading ability, fluid intelligence, and processing speed, and personality traits, viz., openness to experience and extraversion. Experimental validation of the proposed framework on the HCP dataset demonstrates superior predictive performance, achieving state-of-the-art correlations of up to 0.5284, thereby highlighting the importance of the proposed spatio-temporal modeling for understanding brain-behaviour relationships.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Cognitive and personality trait prediction using activation maps and temporal dynamics derived from resting-state fMRI

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月29日

查看全文
  • Sasideep Pasumarthi,
  • Harshith Jangam,
  • Nitya Tiwari,
  • Himanshu Padole

摘要

Predicting behavioral and personality traits from neuroimaging data requires effective modeling of complex and high-dimensional brain dynamics. Although recent advances in neurocomputations have paved the way for trait prediction models, the attempts remain at a nascent stage, as reflected in the moderate prediction accuracies. To this end, in this work, we propose a novel unified framework that leverages both spatial and temporal information derived from resting-state functional MRI (rs-fMRI) for trait prediction. Task-activation maps are first estimated directly from rs-fMRI, providing spatial representations of task-evoked brain activity without requiring explicit task paradigms. To capture temporal dynamics, MultiRocket, an efficient time-series feature-extraction method, is employed, encoding trait-specific patterns in rs-fMRI signals. The spatial and temporal features are then fused to form a unique unified spatio-temporal representation that is subsequently provided to an ensemble framework to predict cognitive traits, viz., reading ability, fluid intelligence, and processing speed, and personality traits, viz., openness to experience and extraversion. Experimental validation of the proposed framework on the HCP dataset demonstrates superior predictive performance, achieving state-of-the-art correlations of up to 0.5284, thereby highlighting the importance of the proposed spatio-temporal modeling for understanding brain-behaviour relationships.