Search Abstracts | Symposia | Slide Sessions | Poster Sessions
The effect of language task type and participant factors on functional Near-Infrared Spectroscopy data quality from left and right hemisphere stroke survivors
Poster Session B, Friday, October 25, 10:00 - 11:30 am, Great Hall 3 and 4
Erin Meier1, Lisa Bunker2, Hana Kim3, Alexandra Zezinka Durfee4, Victoria Tilton-Bolowsky5, Argye Hillis6; 1Northeastern University, 2Midwestern University, 3University of South Florida, 4Towson University, 5Johns Hopkins University School of Medicine
Introduction: Few functional neuroimaging studies of early aphasia recovery exist, likely in part due to the logistical challenges of conducting fMRI in this population (e.g., lack of portability, contraindications). Functional Near-Infrared Spectroscopy (fNIRS) is less costly than fMRI and can be performed in different testing sites (e.g., at bedside) but comes with its own limitations, such as signal attenuation due to hair thickness/color and skin pigmentation, especially in people of color. Relatively little research to date (and thus gold standards) in stroke survivors with and without aphasia also limits the application of this tool to address research gaps. Thus, in this study, we determined if fNIRS data quality varied by (1) stroke laterality and chronicity, (2) fNIRS task type, and (3) demographic variables. Methods: Ninety-five individuals (mean age: 63.4 years, 65 men, 55 white/35 Black/3 Asian/2 other race), including 57 left and 38 right hemisphere stroke survivors, completed at least one fNIRS session, enrolling at either the acute (n=52), subacute (n=21), or chronic (n=22) stroke stage. Participants completed three fNIRS tasks that varied in language and speech motor demands, from low to high: (1) resting state, (2) discourse comprehension (passive listening to 60s stories), and (3) picture naming (overt naming of pictured objects and actions). Data were acquired with a 16x16 NIRx NIRSport2 and 46 long and 8 short channels positioned over bilateral language areas. To assess data quality, we extracted peak spectral power (PSP), scalp coupling indices (SCI), and number of bad channels (channels with below-threshold PSP and SCI for more than 30% of data acquisition windows) from the QT-NIRS Toolbox (Hernandez & Pollonini, 2020). Results: Left and right hemisphere stroke survivors did not differ in PSP (p>0.59), SCI (p>0.29), or number of bad channels (p> 0.40) for any task. Worse PSP (FDR-p<0.001) and more bad channels (FDR-p<0.001) were found for picture naming compared to resting state and discourse comprehension. Before correction for multiple comparisons, higher PSP was associated with greater stroke chronicity for resting state (r=0.21, p=0.048, FDR-p>0.05) and picture naming (r=0.29, p=0.022, FDR-p>0.05). Across tasks, SCI tended to be higher for older participants (range: r=0.188-0.290, FDR-p=0.222-0.027) and women compared to men (range: W=188-540, FDR-p=0.001-0.138). Black participants had lower SCI across all three tasks (FDR-p<0.022) and lower PSP for resting state (FDR-p=0.047) than white participants. Conclusions: This study is motivated by the need to better understand stroke, task, and demographic variables that influence fNIRS data quality in order to inform task design and fNIRS analysis for future studies. While language task data quality did not differ by stroke laterality, participants earlier in their recovery tended to have worse PSP, which has implications for longitudinal stroke recovery studies. As expected, data quality was worse for picture naming due to its higher speech motor demands than the other two tasks. While data quality varied by demographic factors in expected ways, these findings illustrate the need to control for such measures in analyses rather than exclude participants from fNIRS studies (Girolamo et al., 2022; Kwasa et al., 2023).
Topic Areas: Disorders: Acquired,