Consistency of functional connectivity across different movies

NeuroImage

Xuyu Cao¹  Mengting Ye²  Tianhui Shen³  Lixia Tian⁴  Chen Chen⁴ 

[1]Beijing Key Laboratory of Traffic Data Analysis and Mining, Beijing Jiaotong University, Beijing 100044, China

[2]Corresponding author at: School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044, China.

[3]Beijing Key Laboratory of Traffic Data Analysis and Mining, Beijing Jiaotong University, Beijing 100044, China

[4]School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044, China

关键词: Intra-class correlation (ICC);    Movie fMRI;    Individual identification;    Prediction;    Frontoparietal network;    Default mode network;   

DOI  :  

来源: DOAJ

【 摘 要 】

Movie fMRI has emerged as a powerful tool for investigating human brain function, and functional connectivity (FC) plays a predominant role in fMRI-based studies. Accordingly, movie-watching FC may have great potential for future studies on human brain function. Before wide application of movie-watching FC, however, it is essential to evaluate how much it is influenced by differences in movies. The main aim of this study was to investigate the consistency of movie-watching FC across different movies. For this purpose, we performed three sets of analyses on the four movie fMRI runs (with different movie stimuli) included in the HCP dataset. The first set was performed to evaluate the agreement of movie-watching FC in exact values using intra-class correlation (ICC), and the ICC of movie-watching FC across different movies (0.37 on average) was found to be comparable to that of resting-state FC across repeated scans. The second set was performed to evaluate the agreement of movie-watching FC in connectivity patterns, and the results indicate that individuals could be identified with relatively high accuracies (94%-99%) across different movies based on their FC matrices. The final set was performed to test the generalizability of predictive models based on movie-watching FC, as this generalizability is highly dependent on the consistency of the FC. The results indicate that predictive models trained based on FC extracted from one movie fMRI run can make good predictions on FC extracted from runs with different movie stimuli. Taken together, our findings indicate that movie-watching FC is highly consistent across different movies, and conclusions drawn based on movie-watching FC are generalizable.

【 授权许可】

Unknown