Introduction
During the fall semester of 2020, the role as an undergraduate research assistant in Dr. Christina Hugenschmidt’s lab at Wake Forest School of Medicine consisted of providing quality control to data collected from the pilot trial of the IMOVE study conducted in 2017. Briefly, the neurological research goal for the IMOVE study was to determine the statistically significant changes in the brain recorded after participants with mild cognitive impairment enrolled in a 12-week trial consisting of IMPROVment classes. These codified improvisational movement classes were hypothesized to improve neural connections within the somatomotor cortex as well as in the default mode network, areas of the brain associated with coordination, movement, and an improvement in the quality of life.
Although the MRI images of the participants were collected in 2017, further processing is essential to ensure accurate data analysis¹. In order to assist in this processing, the purpose of this research project was to manually edit the MRI images to ensure proper skull-stripping or the removal of non-brain areas such as cerebral spinal fluid and dura². In order for these neuroimages to be compared between participants, the scans need to be warped to a standard template in a process called normalization. However, in aging brains, normalization tends to be more difficult due to larger differences between patient images and the template². In order to do so more accurately, programs such as ANTs are used to strip away areas of the image not classified as brain³. However, in the process of skull stripping, the program may create a mask, or outline, that incorrectly attributes areas as brain, which can be heightened in images with abnormalities². The consequences of these inaccuracies create confounding variables that will affect the warping of the image¹. Because of this, masks were manually edited to provide as accurate of an image as possible prior to warping and normalization.
In addition to providing quality control to masked images, this project acted as a pedagogical model for educating undergraduate neuroscience students in neuroimaging. Increased education in neuroscience research methods within undergraduate schooling may benefit those pursuing advanced degrees in a rapidly developing field ⁴. The protocol for neuroimaging quality control created within this project may provide inspiration for future classes that mutually benefit undergraduate neuroscience students with neuroimaging exposure and research labs with additional assistance.
Methods
In order to clean up data collected from the 2017 pilot study, image editing was conducted in MRIcron, a software tool that can be used as a cross-platform NIfTI format image viewer. Each automated skullstripping mask was manually edited by analyzing T1 images within MRIcron. An overlay was added of the program-generated skullstripped mask without the cerebral spinal fluid. Once added, the draw function within MRIcron was utilized to exclude and include voxels that had been attributed inaccurately. This task created voxels of interest (VOIs) that would track inaccurately assigned voxels. Two masks were generated for each T1 image: one that would remove voxels inaccurately attributed as part of the brain (mask.nii) and one that would include voxels inaccurately attributed as CSF or skull (mask_add.nii). This task provided a more accurate skull-stripped mask for the warping procedure.
Results
Within the course of ten weeks, the skull-stripped masks of 20 fMRI images were manually edited to improve accuracy and quality. These neuroimages were acquired from 10 dyads, each consisting of a participant with dementia and their caregiver. Due to the abnormalities and changes found in the aged brain, more editing was required when correcting the program generated mask. Over the course of the ten-week period, a protocol was created in order to track and ensure equal advancements were being made within the group. In this protocol, each of the four undergraduate research members was assigned 5 of 20 images. Every week, the members would present a mask to the group on Citrix, a video conferencing tool. During this meeting, the fellow undergraduates, the PI, a neuroscience doctoral student, and a programmer from the Department of Radiology would review and constructively critique the edits.
During these critiques, neuroanatomy, neuroimaging procedures, and editing skills were reviewed. This allowed for more meticulous quality control as well as an opportunity to educate the students with hands-on instruction. After reviewing each MRI image, the students would revise their edits and submit them for completion, which would be approved by the programmer.
Conclusion
Through assisting with quality control of skull-stripped images of pilot data, undergraduate neuroscience students were able to advance their knowledge of neuroimaging, neuroanatomy, and working collaboratively to generate more accurate data. The protocol that was generated over the course of the 10 week period may provide a testament to the mutual benefit received by educating undergraduate students while assisting in medical school research. Following this research project, further processing will be completed in order to proceed towards utilizing the IMOVE pilot data for publication.
References
Kalavathi, P., & Prasath, V. B. S. (2016). Methods on Skull Stripping of MRI Head Scan Images—A Review. Journal of Digital Imaging, 29(3), 365–379.
2. Fischmeister, F. Ph. S., Höllinger, I., Klinger, N., Geissler, A., Wurnig, M. C., Matt, E., Rath, J., Robinson, S. D., Trattnig, S., & Beisteiner, R. (2013). The benefits of skull stripping in the normalization of clinical fMRI data. NeuroImage: Clinical, 3, 369–380.
3. Roy, S., Butman, J. A., & Pham, D. L. (2017). Robust Skull Stripping Using Multiple MR Image Contrasts Insensitive to Pathology. NeuroImage, 146, 132–147.
4. Mickley Steinmetz, K. R., & Atapattu, R. K. (2010). Meeting the Challenge of Preparing Undergraduates for Careers in Cognitive Neuroscience. Journal of Undergraduate Neuroscience Education, 9(1), A36–A42.
