Target Audience

This talk is directed toward scientist interested in time-varying connectivity applied to clinical studies.

Objective

Audience will understand important aspects to consider for the application of dynamic connectivity to differentiate various clinical groups.

Purpose

The study of functional connectivity between well differentiated brain areas has provided much light in explaining the mechanism of action of several neuropsychiatric diseases (Greicius, 2008). Functional connectivity can be broadly categorized in two distinct flavors: static (where connectivity is assumed to be static over time) and dynamic (where connectivity is a time dependent assessment). Evidence suggests that dynamic connectivity provides features not characterized in the static-stationary assumption (Hutchison, et al., 2013). This work explores some of the clinical uses of the extra information that can be obtained from dynamic connectivity.

Methods

All studies followed similar fMRI data processing to estimate the dynamic functional network connectivity (dFNC). In summary, the initial steps include slice-timing correction, realignment, co-registration, and spatial normalization, transformation to the Montreal Neurological Institute standard space, despiking, realignment parameters regression, smoothing, group independent component analysis, sliding window correlation and a clustering analysis. Clustering is utilized to identify whole-brain dFNC states as time lapses with similar functional connectivity patterns. The similarity between patterns is determined by the distance metric of the clustering algorithm. The number of states detected ranges from 4 to 6 depending on the cohort characteristics. A statistical analysis is performed separate for each dynamic state to determine the relationship of psychometric assessments with the dFNC within a state. Another assessment (not possible in the static connectivity studies) is the analysis related to the temporal characteristics of each state: time spent on each state, and how long the brain holds a specific state. The inclusion of machine learning algorithm for the identification of patients is another dimension explored with promising clinical applications.

Results

Schizophrenia results indicated a pattern of hyperconnectivity between thalamus and sensorial brain networks as well as a hipoconnectivity within the sensorial network (Damaraju, et al., 2014). The dFNC results showed stronger effect sizes than those observed in static connectivity. Results in Huntington’s disease showed a different panorama since many results obtained in static connectivity (Espinoza, et al., 2018) were absent in the dFNC study of the same data (Espinoza, et al., 2017). In mild traumatic brain injury (mTBI), the use of dFNC for the identification of subjects with mTBI and those who were healthy with a classification accuracy close to the 90% (Vergara, et al., 2018a). The classification power was in this case concentrated in one of the dynamic states providing evidence that dFNC analysis based on sliding window correlation and clustering might aid in separating useful classification features from those not related to the contrast of interest. A dFNC analysis on a large polysubstance cohort revealed that substances of addiction might have different effects on each dFNC state. In this study (Vergara, et al., 2018b), the effects of nicotine, alcohol and cannabis and their comorbid abuse on showed dFNC patterns of different direction in each state that could not be observed in single static connectivity. However, static connectivity results (Vergara, et al., 2017) were more abundant than dFNC ones (Vergara, et al., 2018b) coinciding with what is observed in Huntington’s disease (Espinoza, et al., 2017; Espinoza, et al., 2018).

Discussion

Dynamic brain connectivity analysis have been described by a quasi-stable assumption where the whole brain exhibit stable connectivity patterns for relatively short lapses of time (Allen, et al., 2014). Results show that short lived patterns might be more efficient in revealing features of interest for the study of some mental illnesses. The advantages of dFNC range from strong effect sizes (Damaraju, et al., 2014) to high accuracy classification features (Vergara, et al., 2018a). We can state that in some instances, features of interest live for short periods of time and the dFNC method is able to separate the specific states containing these features from the rest. In other instances such as the Huntington’s disease case (Espinoza, et al., 2017), the differences in each short lived dFNC state might not achieve a strong effect size. Static connectivity in this case serves as a data fusion method where effects distributed in time aggregate to produce a stronger result.

Conclusion

Functional connectivity show promising application to clinical studies. One of them is the identification of mental illnesses by means of assessing brain connectivity. Static and dynamic connectivity exhibit complementary characteristics that must be carefully considered when clinical applications are designed.

Acknowledgements

This work was supported by grants from the National Institutes of Health grant numbers 2R01EB005846, R01REB020407, and P20GM103472; and the National Science Foundation (NSF) grants 1539067/1631819 to VDC.

References

Allen, E.A., Damaraju, E., Plis, S.M., Erhardt, E.B., Eichele, T., Calhoun, V.D. (2014) Tracking whole-brain connectivity dynamics in the resting state. Cereb Cortex, 24:663-76. Damaraju, E., Allen, E.A., Belger, A., Ford, J.M., McEwen, S., Mathalon, D.H., Mueller, B.A., Pearlson, G.D., Potkin, S.G., Preda, A., Turner, J.A., Vaidya, J.G., van Erp, T.G., Calhoun, V.D. (2014) Dynamic functional connectivity analysis reveals transient states of dysconnectivity in schizophrenia. Neuroimage Clin, 5:298-308. Espinoza, F., Vergara, V.M., Miller, R., Caprihan, A., Liu, J., Turner, J., Misiura, M., Ciarochi, J., Johnson, H., Long, J., Bockholt, J., Paulsen, J., Calhoun, V. (2017) Resting-state fMRI dynamic connectivity analysis in prodromal Huntington’s disease. 23rd Annual Meeting of the Organization for Human Brain Mapping. Vancouver, Canada. Espinoza, F.A., Turner, J.A., Vergara, V.M., Miller, R.L., Mennigen, E., Liu, J., Misiura, M.B., Ciarochi, J., Johnson, H.J., Long, J.D., Bockholt, H.J., Magnota, V.A., Paulsen, J.S., Calhoun, V.D. (2018) Whole-brain connectivity in a large study of Huntington's disease gene mutation carriers and healthy controls. Brain Connect. Greicius, M. (2008) Resting-state functional connectivity in neuropsychiatric disorders. Curr Opin Neurol, 21:424-30. Hutchison, R.M., Womelsdorf, T., Allen, E.A., Bandettini, P.A., Calhoun, V.D., Corbetta, M., Della Penna, S., Duyn, J.H., Glover, G.H., Gonzalez-Castillo, J. (2013) Dynamic functional connectivity: promise, issues, and interpretations. Neuroimage, 80:360-378. Vergara, V.M., Liu, J., Claus, E.D., Hutchison, K., Calhoun, V. (2017) Alterations of resting state functional network connectivity in the brain of nicotine and alcohol users. Neuroimage, 151:45-54. Vergara, V.M., Mayer, A., Kiehl, K.A., Calhoun, V.D. (2018a) Dynamic functional network connectivity discriminates mild traumatic brain injury through machine learning. NeuroImage: Clinical. Vergara, V.M., Weiland, B.J., Hutchison, K.E., Calhoun, V.D. (2018b) The Impact of Combinations of Alcohol, Nicotine, and Cannabis on Dynamic Brain Connectivity. Neuropsychopharmacology, 43:877-890.