Harmonizing resting-state functional MRI data to increase sample size and to classify tinnitus brain connectivity

Tinnitus, or ringing in the ears, is a hearing disorder that affects approximately 15% of human adults globally. The neuroscientific understanding of this subjective disorder is limited at present, partly due to a lack of sufficient neuroimaging data to compensate for the heterogeneity in the disorder; neural data acquisition is often impeded by low patient counts and doubled costs against scanning ample controls. However, with the increase in data sharing practices, data scarcity can be overcome with harmonization methods if scans from independent sites and studies are pooled successfully. To achieve data harmonization for limited data from a tinnitus neuroimaging study, we merged a modest fMRI dataset acquired in our lab (n=35) with age- and gender-matched samples from another large dataset, the Lifespan Human Connectome Project Aging (n=377). We measured functional connectivity (FC) between brain regions based on resting-state fMRI session scans, and used deep learning architecture to systematically combine individual FCs across the datasets. We assembled and trained neural network modules to purge dataset-identifying information from FC maps, and retained patterns that can characterize tinnitus-related FC. Hallmark artifactual information of the datasets was significantly reduced while reconstructing resting state FC data. Relevant relationships in brain connectivity that discern the condition of tinnitus from controls were well-preserved in the reconstructed resting state FC data. In agreement with existing literature on brain networks of tinnitus, connectivity related to default mode, ventral attention and frontoparietal control networks were identified to be crucial for distinguishing the condition, as masking these networks connections severely impacted downstream classification prediction on reconstructed FC. The data harmonization model was successful in merging FC matrices from independently acquired fMRI datasets, without losing out on meaningful functional connectivity patterns of interest for the neuropsychological disorder of tinnitus. Our findings demonstrate the potential of data harmonization by integrating independently acquired datasets on healthy controls to overcome neuroimaging data scarcity while preserving disorder-relevant neural signatures, advancing research on understudied neurological disorders.

Authors

Institutions

Publication Details

Journal
bioRxiv (Cold Spring Harbor Laboratory)
Published
2026-11-20
DOI
https://doi.org/10.64898/2026.11.19.767894
Primary Topic
Functional Brain Connectivity Studies
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Harmonizing resting-state functional MRI data to increase sample size and to classify tinnitus brain connectivity

Shagun Ajmera, Fatima Tazeena Husain
1 links
bioRxiv (Cold Spring Harbor Laboratory)
Functional Brain Connectivity Studies
preprint

Harmonizing resting-state functional MRI data to increase sample size and to classify tinnitus brain connectivity

Shagun Ajmera, Fatima Tazeena Husain
preprint en

Abstract

Tinnitus, or ringing in the ears, is a hearing disorder that affects approximately 15% of human adults globally. The neuroscientific understanding of this subjective disorder is limited at present, partly due to a lack of sufficient neuroimaging data to compensate for the heterogeneity in the disorder; neural data acquisition is often impeded by low patient counts and doubled costs against scanning ample controls. However, with the increase in data sharing practices, data scarcity can be overcome with harmonization methods if scans from independent sites and studies are pooled successfully. To achieve data harmonization for limited data from a tinnitus neuroimaging study, we merged a modest fMRI dataset acquired in our lab (n=35) with age- and gender-matched samples from another large dataset, the Lifespan Human Connectome Project Aging (n=377). We measured functional connectivity (FC) between brain regions based on resting-state fMRI session scans, and used deep learning architecture to systematically combine individual FCs across the datasets. We assembled and trained neural network modules to purge dataset-identifying information from FC maps, and retained patterns that can characterize tinnitus-related FC. Hallmark artifactual information of the datasets was significantly reduced while reconstructing resting state FC data. Relevant relationships in brain connectivity that discern the condition of tinnitus from controls were well-preserved in the reconstructed resting state FC data. In agreement with existing literature on brain networks of tinnitus, connectivity related to default mode, ventral attention and frontoparietal control networks were identified to be crucial for distinguishing the condition, as masking these networks connections severely impacted downstream classification prediction on reconstructed FC. The data harmonization model was successful in merging FC matrices from independently acquired fMRI datasets, without losing out on meaningful functional connectivity patterns of interest for the neuropsychological disorder of tinnitus. Our findings demonstrate the potential of data harmonization by integrating independently acquired datasets on healthy controls to overcome neuroimaging data scarcity while preserving disorder-relevant neural signatures, advancing research on understudied neurological disorders.

bioRxiv (Cold Spring Harbor Laboratory)
University of Illinois Urbana-Champaign (US)
Defense Health Agency, Discovery Partners Institute
Good health and well-being
Functional Brain Connectivity Studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon