The validity of different multiple comparison correction methods in the analysis of brain function image data
10.3760/cma.j.issn.1674-6554.2019.10.015
- VernacularTitle: 不同多重比较校正方法在脑功能影像数据分析中的有效性
- Author:
Yingchao SONG
1
;
Li HU
2
,
3
;
Meng LIANG
1
Author Information
1. School of Medical Imaging, Tianjin Medical University, Tianjin 300203, China
2. CAS Key Laboratory of Mental Health, Institute of Psychology, Chinese Academy of Sciences, Beijing 100101, China
3. Department of Psychology, University of Chinese Academy of Sciences, Beijing 100049, China
- Publication Type:Journal Article
- Keywords:
fMRI;
Brain activation;
Multiple comparisons correction;
False positive rate;
Detection rate
- From:
Chinese Journal of Behavioral Medicine and Brain Science
2019;28(10):941-946
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To explore the effectiveness of different multiple comparisons correction methods by comparing the detection rate and false positive rate of brain activation analysis using functional magnetic resonance imaging (fMRI) data.
Methods:On the basis of task-based fMRI dataset (including low-intensity and high-intensity stimuli condition, n=20) and resting-state fMRI dataset(n=32), brain activation results were corrected by multiple comparsion correction methods in SPM and SnPM13 software, and the activation detection rate and false positive rate were compared with different correction methods.
Results:Voxel-or peak-based correction methods had relatively low false positive rate.When P<0.05 after correction, the proportion of the subjects with false-positive were 0.19 and 0.16, and the number of false-positive voxels were 404 and 2 448, respectively.But the two methods had low detection rate, which were more suitable for detecting strong activation.While cluster-based correction methods had relative high detection rate and high false positive rate.When P<0.05 after correction, the proportion of the subjects with false-positive were 0.34 and 0.38, and the number of false-positive voxels were 7 870 and 8 320, respectively.And thus they were more suitable for detecting weak activation. Group-level analysis could effectively reduce false positive rate.
Conclusion:In practice, researchers should choose a suitable correction method based on their specific research objectives and data to achieve a balance between the detection rate and false positive rate.