Poster No:
2513
Submission Type:
Abstract Submission
Authors:
Scott Holmes1, Nicholas Reyes2, Jaxon Huang3, Anat Galor2, Pradip Pattany4, Elizabeth Felix4, Eric Moulton1
Institutions:
1Boston Children's Hospital, Boston, MA, 2Miami Veterans Administration Medical Center, Miami, FL, 3Miami Veterans Affairs Administration Medical Center, Miami, FL, 4University of Miami, Miami, FL
First Author:
Co-Author(s):
Jaxon Huang
Miami Veterans Affairs Administration Medical Center
Miami, FL
Introduction:
The factors that mediate the expression of ocular pain and the mechanisms that promote chronic ocular pain symptoms are poorly understood. Central nervous system involvement is suspected based on observations of pain out of proportion to nociceptive stimuli in some patients. We focused on understanding functional connectivity between brain regions implicated in pain in persons reporting ocular pain symptoms.
Methods:
Using a 3T Siemens MAGNETOM Vida scanner, we performed a resting state fMRI investigation using a region of interest (ROI) analysis that focused on subcortical brain structures including the trigeminal nucleus and performed a battery of ophthalmological examinations. 53 persons were divided into two cohorts: no ocular pain, and chronic ocular pain. A GE-EPI scan (TE/TR=30/2000ms, resolution: 1.94x1.94x1.50mm, duration: 10:06, volumes: 303) was collected with the participants instructed to keep their eyes open and to blink normally. Our field of view focused on brainstem and sub-cortical structures. ROIs included the thalamus, caudate, putamen, pallidum, amygdala, accumbens, brainstem, periaqueductal gray, and trigeminal nucleus. We also included bilateral precentral and postcentral gyrus based on prior work in persons with chronic pain (1).
fMRI data was processed using the CONN toolbox (2). Standard pipeline settings were used, including realignment and unwarping, slice-timing correction, outlier detection, segmentation, normalization, spatial smoothing with an 8 mm filter. Resulting images were processed using a band-pass filter (0.008-0.09 Hz). Statistical significance for functional connectivity analysis was considered significant for p<0.05 (uncorrected for multiple comparisons).
K-means clustering analysis was performed using Sklearn (3) on our chronic ocular pain cohort (n = 37) to separate these participants. We use silhouette coefficients to determine the number of clusters to use in our pain sub-cohorts. All clinical data were included as well as the output functional connectivity matrices from all ROIs. We performed a principal component analysis on the matrices from each participant and extracted the top 20 components that represented 80% of the variance. We chose two clusters based on the limited sample size of the starting population. We also determined the feature performance in differentiating the two pain cohorts using a K-means clustering analysis using the SelectKBest package and then back projected cluster data into original clinical metrics for clinical interpretation.
Results:
The ocular pain cohort reported higher levels of pain symptoms relating to neuropathic pain and ocular surface disease, as well as more abnormal tear metrics (stability and tear production) than the non-pain cohort. Functional connectivity analysis between groups evinced multiple connections exemplifying both increases and decreases in connectivity including regions such as the trigeminal nucleus, amygdala, and sub-regions of the thalamus.
K-based clustering analysis of the ocular pain cohort highlighted two new constituent pain sub-cohorts (n=17, n=20) presenting with unique phenotypes. Clinically, these groups represent individuals with lower (Group 1) and higher (Group 2) levels of ocular pain with neuropathic features. The top five distinguishing features were "Tear production in the left eye", "Tear breakup time in the right eye", "Resting State Network - Principal Component 13," and "Ocular pain-before and after anesthesia," and "Worst pain scores from the prior week." This combination of features reflects fMRI and clinical metrics and suggests that even when considering pain and ocular surface metrics, functional brain connectivity remains an important differentiator between pain cohorts.
Conclusions:
Our findings support centralized involvement in those patients reporting chronic ocular pain and allude to mechanisms through which pain treatment services may be directed in future research.
Modeling and Analysis Methods:
fMRI Connectivity and Network Modeling 2
Novel Imaging Acquisition Methods:
BOLD fMRI
Perception, Attention and Motor Behavior:
Perception: Pain and Visceral 1
Keywords:
ADULTS
Basal Ganglia
Brainstem
Cerebellum
Cortex
Data analysis
FUNCTIONAL MRI
Pain
1|2Indicates the priority used for review
Provide references using author date format
1 Holmes, S.A., Barakat, N., Bhasin, M., Lopez, N.I., Lebel A., Zurakowski D., et al. (2020), ‘Biological and behavioral markers of pain following nerve injury in humans’ Neurobiol Pain, vol. 7, p. 100038. doi: 10.1016/j.ynpai.2019.100038
2 Nieto-Castanon, A., Whitfield-Gabrieli, S. (2012), ‘Conn: a functional connectivity toolbox for correlated and anticorrelated brain networks’ Brain Connect. vol. 2, pp. 125–141. doi: 10.1089/brain.2012.0073
3 Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., et al. (2011) ‘Scikit=−learn: machine learning in Python’ J Mach Learn Res. vol. 12, pp. 2825–2830.