Read Duke’s Cerebellar Volume Proposal

Cerebellar volume based structural alterations in PTSD

I. Title:

Cerebellar volume based structural alterations in PTSD

II. Investigative Team:

1. Ashley Huggins
2. Melvin Briggs
3. Delin Sun
4. Sarah Laskowitz
5. C. Lexi Baird
6. Courtney Haswell
7. ENIGMA-PTSD Workgroup
8. PGC-PTSD Neuroimaging Workgroup
9. Rebecca Kerestes
10. Ian Harding
11. Rajendra Morey

III. Background:

In the past decade imaging studies have shown rather illuminating findings in Post-Traumatic Stress Disorder (PTSD). Although many structural imaging studies have been consistent in reporting decreased hippocampal volume [1] , grey matter reductions in vmPFC, MTL, and insula [2-4], a number of studies among PTSD patients have reported alterations in the cerebellum; including decreased grey matter volume in the right cerebellar crus [5], decreased left cerebellar hemisphere, and vermal volumes [6, 7] and decreased bilateral cerebellar volumes among adolescents with PTSD[8].

Such findings have typically been minimized since historically the cerebellum is believed to be involved exclusively in motor function and coordination [9]. Nonetheless there has been a growing body of evidence supporting cerebellar involvement in decision making, impulse inhibition, behavioral and emotional regulation[10-12].

For example, a PET study investigating cerebral blood flow in healthy volunteers has reported increased CBF in the right anterior vermis during subjective experiences of sadness and anxiety [13]. Besides sadness and anxiety, fear prediction has been correlated with increased CBF [14] and activity [15] in anterior vermis and lateral cerebellum. These circumstantial cerebellar activations may explain common PTSD symptoms such as hyperarousal and increased anxiety, especially since multiple fMRI studies in PTSD have reported increased activation in both cerebellar hemispheres [16, 17] and increased CBF in lateral cerebellum [18, 19].

Neuroanatomical cerebellar connections with non-motor cerebral regions are now well recognized [20]. Not only connections with limbic system that are involved in emotion and affect [21], but also closed loop connections with the PFC are found, which are primarily responsible for emotion regulation[22]. These cerebellar-PFC connections have been implicated in multiple psychiatric conditions such as schizophrenia and autism [23, 24]. This evidence of cortico-cerebellar connections suggests that the cerebellum may play an important role in PTSD symptom manifestation. However, further evidence is necessitated for cerebellar volume reduction involvement in PTSD pathophysiology and more exploration is required to better understand the extent of the cerebellar involvement and the role it plays in the clinical symptoms of PTSD.

Ample data in over 3,000 subjects from the ENIGMA PGC- PTSD neuroimaging workgroup will allow robust investigation of cerebellar volume alteration in PTSD.  We propose using cerebellar segmentation to test our hypothesis that volume reductions in the cerebellum/cerebellar substructures are associated with PTSD symptoms. Although there have been inconsistencies in reporting function of each cerebellar substructure, in this consensus paper [25] it is suggested that vermal and paravermal cerebellar activation is correlated with primary emotion( happiness, anger, disgust, fear and sadness), however activation of lateral hemispheres is observed in (anger, fear, disgust, autonomic function and pain). Our hypotheses do not specify cerebellar substructures since there is great overlap in the evidence of each substructure function so far. We are following an exploratory approach to better understand the substructures associated with PTSD symptom clusters and individual symptoms.

For cerebral segmentation we will be using automatic cerebellum anatomical parcellation using U-Net with locally constrained optimization (ACAPULCO) [26]. This is a convolutional neural network (CNNs) based cerebellar segmentation method. Unlike atlas-based segmentation, CNN bases segmentation uses manually delineated images as training data for parameter optimization that is then applied to the new data.  ACAPULCO uses a cascade of two CNNs. First CNN modified from the pre-activation ResNet [27] ,detects the cerebellum, which reduces the spatial size of the input. Second CNN was modified from a 3D U-NET [28] with residual connections as in the ResNet[27], it uses voxel wise classification to parcellate the cerebellum into 28 anatomically meaningful regions, including central structures (Corporus Medullare, Vermis VI, Vermis VII, Vermis VIII, Vermis IX, Vermis X), lateral structures (L/R Lobules I-III, L/R Lobule IV, L/R Lobule V, L/R Lobule VI, L/R Lobule VIIAF aka Crus I, L/R Lobule VIIAT aka. Crus II, L/R Lobule VIIB, L/R VIIIA, L/R VIIIB) see figure from [26]. ACAPULCO then produces an output of cerebellar volume for each segment as well as for the whole cerebellum.

This segmentation method was compared to the CERES2 (best atlas based method) [29], and was also compared to a version of CGCUTS [30] using 5 different data sets to show broad applicability, only to find that ACAPULCO outperformed both with a better mean Dice score. ACAPULCO reproducibility was also assessed using the Kirby data set [31]. ACAPULCO was trained from the Adult cohort [29], based on the one-way random model,[32] the interclass correlation (ICC) coefficients were all above 0.9, except left lobule V and right lobule X whose smallest value is 0.8894. ICC values greater than 0.75 are considered excellent. 

IV. Data harmonization and statistical analysis:

Our plan is to apply ComBat-GAM harmonization method using python script (https://github.com/rpomponio/neuroHarmonize) to the cerebellar volumes. Age, sex and PTSD diagnosis will be accounted for as biological variable while age will be marked as nonlinear term in the model. We will implement a mega-analysis using mixed effect linear regression model for all inferential analyses to account for random variance attributed to each independent cohort. For this, We plan to use lmerTest package in R [33]. Scanner location and manufacturer will be modeled as nested variable within each scanning site to account for imaging variance. As for fixed effects, we plan to adjust for: age, age², childhood traumatic events, lifetime trauma, psychiatric comorbidities (depression, alcohol use disorders), medications, gender, and intracranial volume (ICV).  We will assess model fit for series of random effect models consisting of (ICV) and demographic variables only. Optimal model fit will be determined from the Akaike information criterion and log likelihood ratios. Analyses with covariates that are not available at all sites will be run on subsamples using cohorts that have covariate data available.

Our primary outcome measure is the average volume of the whole cerebellum as well as volumes of each of the 28 cerebellar segments among patients with PTSD and healthy controls; main predictor will PTSD diagnosis, outcome variable is total cerebellar volume and substructural cerebellar volumes. We plan to report standardized beta’s (β) as a scaled parameter estimate of effect size rather than Cohen’s d, as methods for computing Cohen’s d are not uniquely defined in mixed models, and alternatives vary widely across studies and often do not capture variance attributable to covariates and random effects [34].We will use false discovery rate (FDR) procedure [35] to correct for multiple comparisons across each of the 28 subfields for each omnibus analysis. FDR corrected p-values (q-values) less than 0.1, but greater than 0.05, are referred to as statistical trends. Analyses will be conducted using two-tailed tests. 

Secondary goal will be to investigate whether the is cerebellar substructural alteration associated with certain PTSD symptom clusters (Dissociation, hyperarousal, self-destructive behavior). However, this will be contingent on availability of item-level or subscores across cohorts.

V. Authorship:

We will follow the authorship policy of the PGC-PTSD which can be found at https://pgc-ptsd.com/wp-content/uploads/2017/06/Authorship-Guidelines-PGC-PTSD.pdf

• are you following the authorship policies of the groups involved? YES see https://pgc-ptsd.com/wp-content/uploads/2017/06/Authorship-Guidelines-PGC-PTSD.pdf
• will there be a writing group and if so, who will be included? The writing group will be comprised of the investigative team (#1 – #6 and #8 – #10) listed above.
• what groups or individuals will be listed as authors? Authors will include the writing group plus individual and group contributors of data and analysis from each site (generally 2-3 co-authors from each site).
• will PGC members not listed as named authors be listed at the end of the manuscript? All individuals who meet the criteria established in the PGC-PTSD authorship policy will be co-authors. Other PGC members will not be listed at the end of the manuscript.
• will PGC members or groups be listed as “collaborators” on the PubMed abstract page? All individuals and groups who meet the authorship criteria of the PGC-PTSD authorship policy will be listed as collaborators on the PubMed abstract page. No other individuals or groups will be listed.
• how will funding sources be handled or acknowledged? All funding sources that supported data collection and analysis will be listed in the manuscript.

V. References:

1. Karl, A., et al., A meta-analysis of structural brain abnormalities in PTSD. Neuroscience & Biobehavioral Reviews, 2006. 30(7): p. 1004-1031.

2. Liberzon, I. and C.S. Sripada, The functional neuroanatomy of PTSD: a critical review. Prog Brain Res, 2008. 167: p. 151-69.

3. Meng, L., et al., Trauma-specific Grey Matter Alterations in PTSD. Scientific Reports, 2016. 6(1): p. 33748.

4. Akiki, T.J., C.L. Averill, and C.G. Abdallah, A Network-Based Neurobiological Model of PTSD: Evidence From Structural and Functional Neuroimaging Studies. Curr Psychiatry Rep, 2017. 19(11): p. 81.

5. Holmes, S.E., et al., Cerebellar and prefrontal cortical alterations in PTSD: structural and functional evidence. Chronic Stress (Thousand Oaks), 2018. 2.

6. Baldaçara, L., et al., Reduced cerebellar left hemisphere and vermal volume in adults with PTSD from a community sample. Journal of Psychiatric Research, 2011. 45(12): p. 1627-1633.

7. Baldaçara, L., et al., Relationship between structural abnormalities in the cerebellum and dementia, posttraumatic stress disorder and bipolar disorder. Dement Neuropsychol, 2012. 6(4): p. 203-211.

8. De Bellis, M.D. and M. Kuchibhatla, Cerebellar volumes in pediatric maltreatment-related posttraumatic stress disorder. Biol Psychiatry, 2006. 60(7): p. 697-703.

9. Koziol, L.F., et al., Consensus paper: the cerebellum’s role in movement and cognition. Cerebellum (London, England), 2014. 13(1): p. 151-177.

10.           Phillips, J.R., et al., The cerebellum and psychiatric disorders. Frontiers in public health, 2015. 3: p. 66-66.

11.           Schmahmann, J.D., J.B. Weilburg, and J.C. Sherman, The neuropsychiatry of the cerebellum – insights from the clinic. Cerebellum, 2007. 6(3): p. 254-67.

12.           Bugalho, P., B. Correa, and M. Viana-Baptista, [Role of the cerebellum in cognitive and behavioural control: scientific basis and investigation models]. Acta Med Port, 2006. 19(3): p. 257-67.

13.           Liotti, M., et al., Differential limbic–cortical correlates of sadness and anxiety in healthy subjects: implications for affective disorders. Biological Psychiatry, 2000. 48(1): p. 30-42.

14.           Frings, M., et al., Involvement of the human cerebellum in fear-conditioned potentiation of the acoustic startle response: a PET study. Neuroreport, 2002. 13(10): p. 1275-8.

15.           Ernst, T.M., et al., The cerebellum is involved in processing of predictions and prediction errors in a fear conditioning paradigm. Elife, 2019. 8.

16.           Rabellino, D., et al., The cerebellum after trauma: Resting-state functional connectivity of the cerebellum in posttraumatic stress disorder and its dissociative subtype. Hum Brain Mapp, 2018. 39(8): p. 3354-3374.

17.           Wang, T., et al., Altered resting-state functional activity in posttraumatic stress disorder: A quantitative meta-analysis. Sci Rep, 2016. 6: p. 27131.

18.           Bremner, J.D., et al., Neural correlates of exposure to traumatic pictures and sound in Vietnam combat veterans with and without posttraumatic stress disorder: a positron emission tomography study. Biol Psychiatry, 1999. 45(7): p. 806-16.

19.           Bremner, J.D., et al., Neural correlates of memories of childhood sexual abuse in women with and without posttraumatic stress disorder. Am J Psychiatry, 1999. 156(11): p. 1787-95.

20.           Schutter, D.J.L.G. and J. Van Honk, The cerebellum on the rise in human emotion. The Cerebellum, 2005. 4(4): p. 290-294.

21.           Blatt, G.J., A.L. Oblak, and J.D. Schmahmann, Cerebellar Connections with Limbic Circuits: Anatomy and Functional Implications, in Handbook of the Cerebellum and Cerebellar Disorders, M. Manto, et al., Editors. 2013, Springer Netherlands: Dordrecht. p. 479-496.

22.           Watson, T.C., et al., Back to front: cerebellar connections and interactions with the prefrontal cortex. Front Syst Neurosci, 2014. 8: p. 4.

23.           Andreasen, N.C. and R. Pierson, The role of the cerebellum in schizophrenia. Biol Psychiatry, 2008. 64(2): p. 81-8.

24.           Fatemi, S.H., et al., Consensus paper: pathological role of the cerebellum in autism. Cerebellum, 2012. 11(3): p. 777-807.

25.           Adamaszek, M., et al., Consensus Paper: Cerebellum and Emotion. The Cerebellum, 2017. 16(2): p. 552-576.

26.           Han, S., et al., Automatic cerebellum anatomical parcellation using U-Net with locally constrained optimization. NeuroImage, 2020. 218: p. 116819.

27.           He, K., et al. Identity Mappings in Deep Residual Networks. 2016. Cham: Springer International Publishing.

28.           Baris Kayalibay, G.J., Patrick van der Smagt, CNN-based Segmentation of Medical Imaging Data. arXiv.org, 2017.

29.           Carass, A., et al., Comparing fully automated state-of-the-art cerebellum parcellation from magnetic resonance images. NeuroImage, 2018. 183: p. 150-172.

30.           Yang, Z., et al., Automated cerebellar lobule segmentation with application to cerebellar structural analysis in cerebellar disease. NeuroImage, 2016. 127: p. 435-444.

31.           Landman, B.A., et al., Multi-parametric neuroimaging reproducibility: A 3-T resource study. NeuroImage, 2011. 54(4): p. 2854-2866.

32.           Bartko, J.J., The Intraclass Correlation Coefficient as a Measure of Reliability. Psychological Reports, 1966. 19(1): p. 3-11.

33.           Kuznetsova, A., P.B. Brockhoff, and R.H.B. Christensen, lmerTest Package: Tests in Linear Mixed Effects Models. 2017, 2017. 82(13): p. 26.

34.           Judd, C.M., J. Westfall, and D.A. Kenny, Experiments with More Than One Random Factor: Designs, Analytic Models, and Statistical Power. Annual Review of Psychology, 2017. 68(1): p. 601-625.

35.           Benjamini, Y. and Y. Hochberg, Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society: Series B (Methodological), 1995. 57(1): p. 289-300.