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.


Patch extraction performance. Figure shows r2 (left panels) and RMSE (right panels) averaged over all extracted patches and permutations at different SNR and anatomical noise levels. At each SNR and anatomical noise level, the curves show the mean performance averaged over 2000 (100 permutations × 4 anatomical noise level × 5 patches) and 3000 (100 permutations × 6 SNR levels × 5 patches) extracted patches, respectively. The SSICA outperforms the regular approach, especially in cases where the power of the specific component is low or where the between subject variability in location of a component is high. Shaded area indicates ± standard error of the mean (sem).
© Copyright Policy
Related In: Results  -  Collection

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

Figure 3: Patch extraction performance. Figure shows r2 (left panels) and RMSE (right panels) averaged over all extracted patches and permutations at different SNR and anatomical noise levels. At each SNR and anatomical noise level, the curves show the mean performance averaged over 2000 (100 permutations × 4 anatomical noise level × 5 patches) and 3000 (100 permutations × 6 SNR levels × 5 patches) extracted patches, respectively. The SSICA outperforms the regular approach, especially in cases where the power of the specific component is low or where the between subject variability in location of a component is high. Shaded area indicates ± standard error of the mean (sem).

Mentions: To investigate patch reconstruction performance, RMSE and r2 measures were calculated and averaged over the results of the SSICA and the regular gICA approach on a large number of hybrid fMRI data sets. Data sets were generated using various combinations of patches as shared and specific components (100 random permutations of up to 2 specific patches per group) at different anatomical noise levels (n = 0, 1, 2, 3 voxels) and with different SNR values (0.5–1 in 0.1 steps). In total, 30 components were extracted by both algorithms. In SSICA, up to 3 components were allowed to be extracted as specific component per group. Among all extracted components, the five showing the highest correlation with the source patches were selected to estimate extraction performance using the metric defined in Equation (5). Figure 3 illustrates RMSE and r2 averaged over all extracted patches and permutations at different SNR and anatomical noise levels, for the SSICA in red and the regular gICA approach in blue. At each SNR and anatomical noise level, the curves show the mean performance averaged over 2000 (100 permutations × 4 anatomical noise level × 5 patches) and 3000 (100 permutations × 6 SNR levels × 5 patches) extracted patches, respectively. The SSICA outperforms the regular gICA approach, especially in cases where the strength of the desired component was low, or where the between subject variability in the location of the desired component (as expected when registration errors are still present) was high.


Validation of Shared and Specific Independent Component Analysis (SSICA) for Between-Group Comparisons in fMRI
Patch extraction performance. Figure shows r2 (left panels) and RMSE (right panels) averaged over all extracted patches and permutations at different SNR and anatomical noise levels. At each SNR and anatomical noise level, the curves show the mean performance averaged over 2000 (100 permutations × 4 anatomical noise level × 5 patches) and 3000 (100 permutations × 6 SNR levels × 5 patches) extracted patches, respectively. The SSICA outperforms the regular approach, especially in cases where the power of the specific component is low or where the between subject variability in location of a component is high. Shaded area indicates ± standard error of the mean (sem).
© Copyright Policy
Related In: Results  -  Collection

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

Figure 3: Patch extraction performance. Figure shows r2 (left panels) and RMSE (right panels) averaged over all extracted patches and permutations at different SNR and anatomical noise levels. At each SNR and anatomical noise level, the curves show the mean performance averaged over 2000 (100 permutations × 4 anatomical noise level × 5 patches) and 3000 (100 permutations × 6 SNR levels × 5 patches) extracted patches, respectively. The SSICA outperforms the regular approach, especially in cases where the power of the specific component is low or where the between subject variability in location of a component is high. Shaded area indicates ± standard error of the mean (sem).
Mentions: To investigate patch reconstruction performance, RMSE and r2 measures were calculated and averaged over the results of the SSICA and the regular gICA approach on a large number of hybrid fMRI data sets. Data sets were generated using various combinations of patches as shared and specific components (100 random permutations of up to 2 specific patches per group) at different anatomical noise levels (n = 0, 1, 2, 3 voxels) and with different SNR values (0.5–1 in 0.1 steps). In total, 30 components were extracted by both algorithms. In SSICA, up to 3 components were allowed to be extracted as specific component per group. Among all extracted components, the five showing the highest correlation with the source patches were selected to estimate extraction performance using the metric defined in Equation (5). Figure 3 illustrates RMSE and r2 averaged over all extracted patches and permutations at different SNR and anatomical noise levels, for the SSICA in red and the regular gICA approach in blue. At each SNR and anatomical noise level, the curves show the mean performance averaged over 2000 (100 permutations × 4 anatomical noise level × 5 patches) and 3000 (100 permutations × 6 SNR levels × 5 patches) extracted patches, respectively. The SSICA outperforms the regular gICA approach, especially in cases where the strength of the desired component was low, or where the between subject variability in the location of the desired component (as expected when registration errors are still present) was high.

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.