Limits...
Validation of Shared and Specific Independent Component Analysis (SSICA) for Between-Group Comparisons in fMRI

View Article: PubMed Central - PubMed

ABSTRACT

Independent component analysis (ICA) has been widely used to study functional magnetic resonance imaging (fMRI) connectivity. However, the application of ICA in multi-group designs is not straightforward. We have recently developed a new method named “shared and specific independent component analysis” (SSICA) to perform between-group comparisons in the ICA framework. SSICA is sensitive to extract those components which represent a significant difference in functional connectivity between groups or conditions, i.e., components that could be considered “specific” for a group or condition. Here, we investigated the performance of SSICA on realistic simulations, and task fMRI data and compared the results with one of the state-of-the-art group ICA approaches to infer between-group differences. We examined SSICA robustness with respect to the number of allowable extracted specific components and between-group orthogonality assumptions. Furthermore, we proposed a modified formulation of the back-reconstruction method to generate group-level t-statistics maps based on SSICA results. We also evaluated the consistency and specificity of the extracted specific components by SSICA. The results on realistic simulated and real fMRI data showed that SSICA outperforms the regular group ICA approach in terms of reconstruction and classification performance. We demonstrated that SSICA is a powerful data-driven approach to detect patterns of differences in functional connectivity across groups/conditions, particularly in model-free designs such as resting-state fMRI. Our findings in task fMRI show that SSICA confirms results of the general linear model (GLM) analysis and when combined with clustering analysis, it complements GLM findings by providing additional information regarding the reliability and specificity of networks.

No MeSH data available.


Related in: MedlinePlus

Splitting factor probability histograms at different within-component variability levels. This figure shows the performance of SSICA when more specific components than the actual simulated components are extracted (M1 > Kg1 and M2 > Kg2). Splitting factor is binned at three levels: low (0–0.2) (noise condition), middle (0.2–0.8; repetition condition), and high (0.8–1; split condition). Histograms are generated based a total of 480 pairs of specific components (20 permutations × 4 anatomical noise levels × 6 SNR values) at each within-component variability level.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5037228&req=5

Figure 6: Splitting factor probability histograms at different within-component variability levels. This figure shows the performance of SSICA when more specific components than the actual simulated components are extracted (M1 > Kg1 and M2 > Kg2). Splitting factor is binned at three levels: low (0–0.2) (noise condition), middle (0.2–0.8; repetition condition), and high (0.8–1; split condition). Histograms are generated based a total of 480 pairs of specific components (20 permutations × 4 anatomical noise levels × 6 SNR values) at each within-component variability level.

Mentions: Figure 6 reports the performance of SSICA when more specific components than embedded specific patches are extracted (M1 > Kg1 and M2 > Kg2). The splitting factor probability histograms are shown at different within-component variability levels (Equation 7). 1920 (20 × 4 × 6 × 4) hybrid fMRI datasets were generated by various combinations of one specific patch embedded per group and 3 shared patches (20 random permutations), at different anatomical noise levels (n = 0, 1, 2, 3 voxels), with different SNR values (0.5–1 in 0.1 steps), and using various within-component variability levels (k = 0–1.5 in 0.5 steps). Whereas, only one specific component per group was simulated in each configuration, two specific components per group were allowed to be extracted in the SSICA. Again, in order to make the interpretations easier, the splitting factor was binned at three levels: low (0–0.2; noise condition where the strong, but not the weak, extracted specific component is correlated with the simulated specific patch), middle (0.2–0.8; repetition condition where some blobs of the simulated specific patch is repeated in both the weak and strong extracted specific components), and high (0.8–1; split condition where the simulated specific patch is split into the two extracted specific components). As shown in Figure 6, at no or low within-component variability levels (k < 1), SSICA extracts all the blobs of the specific patch as one specific component, while the second component represents only extraction noise (cf. splitting factor around zero). At intermediate within-component variability levels (k = 1), SSICA still extracts one specific component which includes the whole patch's blobs, while the second component duplicates some of the blobs (cf. splitting factor around 0.5). At high within-component variability levels (k > 1), SSICA splits the specific patch into the two extracted specific components for each group (cf. splitting factor around one). This simulation demonstrates that depending on the degree of within component variability, SSICA extracts the specific patch exclusively, repeats it in several components, or split it across different components.


Validation of Shared and Specific Independent Component Analysis (SSICA) for Between-Group Comparisons in fMRI
Splitting factor probability histograms at different within-component variability levels. This figure shows the performance of SSICA when more specific components than the actual simulated components are extracted (M1 > Kg1 and M2 > Kg2). Splitting factor is binned at three levels: low (0–0.2) (noise condition), middle (0.2–0.8; repetition condition), and high (0.8–1; split condition). Histograms are generated based a total of 480 pairs of specific components (20 permutations × 4 anatomical noise levels × 6 SNR values) at each within-component variability level.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5037228&req=5

Figure 6: Splitting factor probability histograms at different within-component variability levels. This figure shows the performance of SSICA when more specific components than the actual simulated components are extracted (M1 > Kg1 and M2 > Kg2). Splitting factor is binned at three levels: low (0–0.2) (noise condition), middle (0.2–0.8; repetition condition), and high (0.8–1; split condition). Histograms are generated based a total of 480 pairs of specific components (20 permutations × 4 anatomical noise levels × 6 SNR values) at each within-component variability level.
Mentions: Figure 6 reports the performance of SSICA when more specific components than embedded specific patches are extracted (M1 > Kg1 and M2 > Kg2). The splitting factor probability histograms are shown at different within-component variability levels (Equation 7). 1920 (20 × 4 × 6 × 4) hybrid fMRI datasets were generated by various combinations of one specific patch embedded per group and 3 shared patches (20 random permutations), at different anatomical noise levels (n = 0, 1, 2, 3 voxels), with different SNR values (0.5–1 in 0.1 steps), and using various within-component variability levels (k = 0–1.5 in 0.5 steps). Whereas, only one specific component per group was simulated in each configuration, two specific components per group were allowed to be extracted in the SSICA. Again, in order to make the interpretations easier, the splitting factor was binned at three levels: low (0–0.2; noise condition where the strong, but not the weak, extracted specific component is correlated with the simulated specific patch), middle (0.2–0.8; repetition condition where some blobs of the simulated specific patch is repeated in both the weak and strong extracted specific components), and high (0.8–1; split condition where the simulated specific patch is split into the two extracted specific components). As shown in Figure 6, at no or low within-component variability levels (k < 1), SSICA extracts all the blobs of the specific patch as one specific component, while the second component represents only extraction noise (cf. splitting factor around zero). At intermediate within-component variability levels (k = 1), SSICA still extracts one specific component which includes the whole patch's blobs, while the second component duplicates some of the blobs (cf. splitting factor around 0.5). At high within-component variability levels (k > 1), SSICA splits the specific patch into the two extracted specific components for each group (cf. splitting factor around one). This simulation demonstrates that depending on the degree of within component variability, SSICA extracts the specific patch exclusively, repeats it in several components, or split it across different components.

View Article: PubMed Central - PubMed

ABSTRACT

Independent component analysis (ICA) has been widely used to study functional magnetic resonance imaging (fMRI) connectivity. However, the application of ICA in multi-group designs is not straightforward. We have recently developed a new method named &ldquo;shared and specific independent component analysis&rdquo; (SSICA) to perform between-group comparisons in the ICA framework. SSICA is sensitive to extract those components which represent a significant difference in functional connectivity between groups or conditions, i.e., components that could be considered &ldquo;specific&rdquo; for a group or condition. Here, we investigated the performance of SSICA on realistic simulations, and task fMRI data and compared the results with one of the state-of-the-art group ICA approaches to infer between-group differences. We examined SSICA robustness with respect to the number of allowable extracted specific components and between-group orthogonality assumptions. Furthermore, we proposed a modified formulation of the back-reconstruction method to generate group-level t-statistics maps based on SSICA results. We also evaluated the consistency and specificity of the extracted specific components by SSICA. The results on realistic simulated and real fMRI data showed that SSICA outperforms the regular group ICA approach in terms of reconstruction and classification performance. We demonstrated that SSICA is a powerful data-driven approach to detect patterns of differences in functional connectivity across groups/conditions, particularly in model-free designs such as resting-state fMRI. Our findings in task fMRI show that SSICA confirms results of the general linear model (GLM) analysis and when combined with clustering analysis, it complements GLM findings by providing additional information regarding the reliability and specificity of networks.

No MeSH data available.


Related in: MedlinePlus