基于多目标优化约束独立成分分析方法的fMRI数据分析研究

Study on fMRI Data Analysis Based on Multi-Objective Optimization CICA

约束独立成分分析(CICA)通过加入先验信息,可极大地提高独立成分分析(ICA)的盲源信号分析性能,但还存在先验信息难以获取、先验信息约束条件阈值参数难以选择以及先验信息难以被有效利用等问题,需要进一步研究和解决。在多目标优化框架的基础上,建立一种同时融合时空先验信息的CICA模型,可有效规避CICA中阈值参数选择的问题。此外,提出一种从多被试fMRI数据中提取本真先验信息来指导fMRI组分析的自适应挖掘算法,从而为CICA获取先验信息提供一种新途径。最后,利用10例模拟数据、5例任务态和23例静息态fMRI数据,验证所提方法的有效性。结果表明:基于多目标优化的CICA(MOPCICA)获得的时空源信号总体上优于ICA、包含时间信息的CICA(CICA-tR)和包含空间信息的CICA(CICA-sR)(P<0.05)(如在模拟数据中,对应的空间AUC和时间相关系数分别0.75±0.05、0.62±0.02、0.72±0.03、0.71±0.06和0.81±0.13、0.67±0.04、0.74±0.09、0.77±0.13),而空间独立性则优于CICA-tR和CICA-sR (P<0.05)(如在任务态数据中,对应的峭度和负熵分别为69.20±23.36、17.60±13.22、36.71±13.43和0.031 2±0.007 7、0.003 7±0.002 1、0.018 4±0.004 5),从而说明它具有更好的源信号恢复性能。同时,在静息态数据中利用fMRI本真先验信息,MOPCICA获得的组成分与每个被试相应成分之间的相关系数平均高于ICA、基于牛顿迭代法的CICA(CICA-nR)和基于不动点迭代法的CICA(CICAfR)(P<0.05)(分别为0.46±0.08、0.44±0.08、0.45±0.08和0.44±0.08),从而更能代表组中被试的共性。研究表明,所提出的方法对fMRI脑功能连通性检测具有重要意义。

Constrained independent component analysis(CICA)greatly improves the performance of blind source signal analysis of independent component analysis(ICA)by incorporating priori information,nevertheless,the current CICA method has some problems,such as the difficulty in obtaining prior information,selecting threshold parameters of prior information constraints,and using priori information effectively,which need to be improved. Targeting to these problems,this paper established a CICA model that simultaneously integrated temporal and spatial priori information on the basis of multi-objective optimization framework,and solved the problem of selecting threshold parameters in CICA through multi-objective optimization strategy.Furthermore,an adaptive mining algorithm was proposed to extract intrinsic a priori information from the fMRI data of multiple subjects to guide the analysis of fMRI group data,thus providing a new way for CICA to obtain priori information. Finally,10 simulated data,5 task-state and 23 resting-state fMRI data were used to verify the effectiveness of the proposed method. The results showed that the spatio-temporal source signals obtained by multi-objective optimization based CICA(MOPCICA)were generally superior to those obtained by ICA,CICA with temporal reference(CICA-tR)and CICA with spatial reference(CICA-sR)(P<0. 05)(in the simulation data,the corresponding spatial AUC and temporal correlation coefficients were 0. 75 ± 0. 05,0. 62 ± 0. 02,0. 72±0. 03,0. 71±0. 06 and 0. 81 ± 0. 13,0. 67 ± 0. 04,0. 74 ± 0. 09,0. 77 ± 0. 13,respectively);while the spatial independence was superior to CICA-tR and CICA-sR(P < 0. 05)(in the task-related data,the corresponding kurtosis and negentropy were 69. 20 ± 23. 36,17. 60 ± 13. 22,36. 71 ± 13. 43 and 0. 031 2 ±0. 007 7,0. 003 7±0. 002 1,0. 018 4±0. 004 5,respectively),which indicated that it had a better performance for the blind source signal recovery. Meanwhile,the correlation coefficient between the group component obtained by MOPCICA through using the fMRI intrinsic priori information in the resting state data and the corresponding component of each subject in the group was on average higher than that of ICA,CICA-nR and CICA-fR(P<0. 05),which were 0. 46±0. 08,0. 44±0. 08,0. 45±0. 08 and 0. 44±0. 08 separately,thus can better represented the commonality of the subjects in the group. Therefore,it has a great significance for the fMRI brain functional connectivity detection.