Recently, it has been shown that individual classification of subjects across scans is possible using correlation matrices derived from their resting state data. In the current study, a cohort of patient with major depression and healthy controls each had five resting state fMRI scans over the course of a double blind randomized placebo controlled cross-over ketamine infusion study We apply this classification scheme to our data to see if we can link individual functional connectivity profiles to behavioral improvements after ketamine.
Major Depressive Disorder (MDD) is associated with a heavy burden of disability and can also lead to suicidal thoughts, attempts and deaths. The heterogeneity of symptoms associated with MDD has been an obstacle to identifying specific neural correlates of the presence of depression as well as response to treatment. Isolating specific depressive symptom clusters may be useful in identifying associations with biologic markers that could facilitate understanding the neurobiological underpinnings of depression.1-3 Recently, it has been shown that individual classification of subjects across scans is possible using correlation matrices derived from their resting state data4. In the current study, a cohort of MDD patients and healthy controls (HC) each had five resting state fMRI scans over the course of a double blind randomized placebo controlled cross-over ketamine infusion study. Here, we apply this classification scheme to our data to see if we can link individual functional connectivity profiles to behavioral improvements after ketamine.
29 MDD (ages 20-65, 17 female) and 25 healthy (ages 24-56, 12 female) subjects are included in this analysis. Resting state fMRI scans were 8 minutes long with the subject’s eyes closed (fMRI parameters: 3T, TR:2.5s, TE: 25 ms, FA: 90, res: 3.75x3.75x3.5 mm, matrix 64x64) along with a high resolution MPRAGE anatomical scan (1mm isotropic). Cardiac and respiration data were also recorded. The data was preprocessed using AFNI5, motion and physiological noise corrected, blurred to 6 mm, filtered: bandpassed between 0.01 and 0.1 Hz, and aligned to the MNI 152 standard template. Connectivity matrices were calculated from the 33 Raichle (2011)6 seeds and 238 region Power (2012)7 parcellation (that survived within a group mask). Individual identification, edge strength and prediction of behavioral variable prediction were calculated following Finn (2017)6 Only subjects with scans in the ’database’ and the ‘target’ set of scans were kept for each comparison.