PREDICTions: Using High-Performance Computing to Visualize the HD Brain
By: Hans Johnson, PhD
For more than a decade, the PREDICT-HD study has followed approximately 1,400 individuals around the world, including both premanifest carriers of the mutant Huntingtin allele and normal control individuals, in an attempt to identify the earliest signs of HD. PREDICT investigators at 28 sites obtain yearly MRI scans and cognitive, behavioral, and physical assessments. This enormous dataset has spurred researchers at the University of Iowa to develop novel algorithms and apply advanced computing techniques to the analysis of this imaging data. Dr. Hans Johnson, Director of the Image Processing Lab at Iowa Mental Health Clinical Research Center, is the chief technical officer for the study. Here, he describes some of the advanced computing techniques currently utilized to analyze thousands of MRIs to identify and characterize HD-related changes in the brain.
The Scalable Informatics, Neuroimaging, Analysis, Processing and Software Engineering (SINAPSE) group focuses primarily on creating specialized software tools to efficiently process and analyze enormous quantities of neuroimaging data to accelerate brain research. SINAPSE has identified novel imaging markers of early HD using high-performance computing to analyze multi-modal brain scans.1 High performance computing refers to the use of supercomputers or computer clusters to solve advanced computation problems too complex for single workstations.
Between November 2012 and March 2013, the group analyzed approximately 5,000 brain images from both control individuals and individuals with premanifest HD. More than 350 compute years (one compute year equals one computer running full-time for one year) were applied, producing more than 40 terabytes of intermediate data. One analysis lead to the development of individualized brain atlases that depict brain structure and enable detailed analysis of changes over time (Figure 1). In another project, the group used high-performance computing resources with FreeSurfer, a software program that integrates hundreds of MRI images to create three-dimensional digital reconstructions of the brain, to separate white matter areas from the rest of the brain in over 200 MRI scans from individuals with premanifest HD.2 As large-scale, international observational studies continue to generate large amounts of imaging data, the ability to quickly and accurately process these data will become essential to timely analysis and reporting.
Data from PREDICT continue to be cataloged and shared with HD clinicians and researchers worldwide, and can be combined with data from other HD studies to explore relationships between brain imaging markers and clinical, psychiatric, movement, mood, and genetic aspects of HD. (For a list of publications, visit PREDICT-HD Publications and Presentations.) Multiple avenues of exploration of HD progression have emerged from the imaging data, including white matter changes, grey matter loss, white matter tractography, default mode compensatory networks, and functional connectivity.2-5 In the future, these analyses may contribute to a more precise, objective measure of the progression of HD that could inform the design of clinical trials of novel therapeutic agents.
HD Insights would like to thank Amanda Miller, LMSW, Jeffrey Long, PhD, and Dawei Liu, PhD, for their assistance in preparing this article.
Editor’s Desk: Randomized Controlled Trials in Populations At Risk for HD
At the HSG 2013 HD Clinical Research Symposium, Kevin Biglan, MD, MPH, Associate Professor of Neurology at the University of Rochester Medical Center, presented the results of the PREQUEL study, which evaluated the safety and tolerability of three different doses of Coenzyme Q10 (CoQ10) in individuals with premanifest HD (see HD Insights, Vol. 6). Retention of study participants was approximately 93% and CoQ10 was well tolerated in the study. Serum CoQ10 levels increased, but a biomarker of oxidative stress remained unchanged.6The study’s Principal Investigator, Chris Ross, MD, PhD, Professor of Psychiatry, Neurology and Neuroscience at Johns Hopkins Medicine, commented that the study will “greatly increase the feasibility of [future] trials” in this population and is “complementary to [the approach] of the PRECREST study.”
The PRECREST study, published online in Neurology earlier this month, investigated the safety and tolerability of high-dose creatine in 64 individuals with pre-manifest HD and 50% at-risk for HD. Ten of the 32 individuals randomized to creatine discontinued participation in the placebo-controlled phase. Neuroimaging results showed treatment-related slowing of cortical and striatal atrophy. Together, both studies lay the foundation for broad participation and evaluation of individuals who are at risk for HD in future, possibly preventive, clinical trials.7
1 Harrington D, Liu D, Smith M, et al. Neuroanatomical correlates of cognitive functioning in prodromal Huntington disease. Brain and Behavior. 2014 Jan; 4(1):29-40.
2 Nopoulos PC, Aylward EH, Ross CA, et al. PREDICT-HD Investigators Coordinators of Huntington Study Group (HSG) (2010). Cerebral cortex structure in prodromal Huntington disease. Neurobiol Dis. 2010 Dec;40(3):544-54. doi: 10.1016/j.nbd. 2010.07.014. Epub 2010 Aug 2.
3 Nopoulos P, Magnotta VA, Mikos A, et al. Morphology of the cerebral cortex in preclinical Huntington’s disease. Am J Psychiatry 2007;164(9):1428-34. doi:10.1176/appi.ajp. 2007.06081266 !
4 Paulsen, Jane S. Functional imaging in Huntington’s disease. Exp Neurol. 2009 Apr;216(2):272-7. doi: 10.1016/j.expneurol. 2008.12.015. Epub 2009 Jan 3.
5 Aylward E, Liu D, Nopoulos P, et al. Striatal volume contributes to the prediction of onset of Huntington disease in incidence cases. Biol Psychiatry. 2012 May 1;71(9):822-8. doi: 10.1016/j.biopsych.2011.07.030. Epub 2011 Sep 9.
6 Killoran A, Biglan KM, Beal F, et al. The PREQUEL Multi-Center Phase II Study of Coenzyme Q10 in Pre-Manifest Huntington Disease. [abstract] 2013 Annual Meeting of the American Neurological Association. 2013 Oct. 13-15.