Evaluation of Functional Connectivity Pipeline Performance in Autism
Poster No:
1249
Submission Type:
Abstract Submission
Authors:
Iva Ilioska1,2, Marianne Oldehinkel3, Seyed Kia4, Christian Beckmann5, Andre Marquand6, Jan Buitelaar2, Alex Fornito7
Institutions:
1Cambridge University, Cambridge, United Kingdom, 2Donders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center, Nijmegen, Netherlands, 3Donders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center, Nijemgen, Netherlands, 4Tilburg University, Tilburg, North Brabant, 5Donders Institute for Brain, Cognition, and Behaviour, Nijmegen, Netherlands, 6Donders Institute for Brain, Cognition and Behaviour, Nijmegen, Gelderland, 7Monash University, Clayton, Victoria
First Author:
Iva Ilioska
Cambridge University|Donders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center
Cambridge, United Kingdom|Nijmegen, Netherlands
Cambridge University|Donders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center
Cambridge, United Kingdom|Nijmegen, Netherlands
Co-Author(s):
Marianne Oldehinkel
Donders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center
Nijemgen, Netherlands
Donders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center
Nijemgen, Netherlands
Jan Buitelaar
Donders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center
Nijmegen, Netherlands
Donders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center
Nijmegen, Netherlands
Introduction:
Normative modelling offers a powerful framework for benchmarking individual scores on brain phenotypes against population expectations (1). There are many ways of processing fMRI data, and the effects of these variations on functional connectivity (FC) estimates is well-documented. We understand less about how some of these choices affect the results of normative modelling studies.
This work evaluates the impact of three choices for FC estimation on normative models: 1) s-score normalization of FC correlations; 2) the use of global signal regression (GSR); and 3) the use of mixture modeling (MM) normalization (Figure 1). Each of these three steps were applied in standard FC analyses and used to estimate FC deviations from a normative model, yielding a total of six different pipelines. We examined a large dataset of 1,824 participants (796 autistic; ages 5-58) from LEAP (2) and ABIDE (3), to evaluate how these pipelines impact motion contamination, sensitivity for detecting group differences, and behavioral prediction accuracy.
This work evaluates the impact of three choices for FC estimation on normative models: 1) s-score normalization of FC correlations; 2) the use of global signal regression (GSR); and 3) the use of mixture modeling (MM) normalization (Figure 1). Each of these three steps were applied in standard FC analyses and used to estimate FC deviations from a normative model, yielding a total of six different pipelines. We examined a large dataset of 1,824 participants (796 autistic; ages 5-58) from LEAP (2) and ABIDE (3), to evaluate how these pipelines impact motion contamination, sensitivity for detecting group differences, and behavioral prediction accuracy.
Methods:
All pipelines relied on a basic pre-processing pipeline that also used ICA-AROMA (4). FC was calculated as Pearson's correlation. For the normative modeling pipelines, we used gaussian process regression to estimate edge-wise FC normative models accounting for age, sex, and framewise displacement (1). Motion artifact assessment was conducted through QC-FC distance dependence analysis (5), while group differences were evaluated using t-tests corrected for age, sex and false discovery rate. Group difference consistency was measured through t-statistic cosine similarity and Dice Similarity Coefficient (DSC) for spatial consistency of positive and negative significant connections. We used support vector regression with 5-fold cross-validation to develop multivariate predictive models of clinical and cognitive measures using FC values and deviation scores. We used ADOS social affect, ADOS restricted and repetitive behavior (RRB), ADI RRB, ADI communication, ADI social interaction (6), Social Responsiveness Scale-2 (SRS) (7), full scale IQ from the Wechsler Abbreviated Scales of Intelligence-Second Edition (8) and the Short Sensory Profile scale (9) and Autism Quotient (10).
Results:
Normative modeling pipelines showed minimal QC-FC distance dependence (Dm: -0.01 to 0.01), outperforming the FC pipelines and especially the Fisher's Z pipeline, which showed the highest distance dependence (Dm: 0.02 to 0.09) (Figure 2A). All pipelines consistently identified patterns of somatosensory hypoconnectivity and transmodal networks hyperconnectivity in autism, though their sensitivity to these patterns varied considerably. The Fisher's Z pipeline demonstrated the highest sensitivity to hypoconnectivity, while the MM pipeline revealed balanced hypo- and hyperconnectivity patterns (Figure 2B,C). T-statistic maps were highly similar across pipelines with similarity ranging from 0.74 to 0.98 (Figure 2D). The similarity of positive values was lowest between the MM and GSR pipelines with DSC=0.38, and highest between FC and NM pairs with the highest being the MM and MM+NM DSC =0.87. Negative values were least similar between Fisher's Z and GSR pipelines DSC = 0.4 and most similar between the MM and MM+NM pair DSC=0.87 (Figure 2E). All pipelines showed significant clinical and cognitive variable predictive power for full scale IQ and SRS (Figure 2F).
Conclusions:
While showing minimal differences in motion contamination and maintaining similar clinical prediction capabilities, the pipelines exhibited notable variations in their sensitivity to group differences. The differences in sensitivity to connectivity patterns suggest that methodological choices in preprocessing and analysis can substantially influence research findings. This variation in pipeline sensitivity may help explain some of the inconsistencies observed in the autism connectivity literature and underscores the importance of carefully considering analytical approaches in future studies.
Disorders of the Nervous System:
Neurodevelopmental/ Early Life (eg. ADHD, autism) 2
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural) 1
fMRI Connectivity and Network Modeling
Motion Correction and Preprocessing
Multivariate Approaches
Keywords:
Autism
Data analysis
FUNCTIONAL MRI
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
2. Loth E, Charman T, Mason L, Tillmann J, Jones EJH, Wooldridge C, et al. The EU-AIMS Longitudinal European Autism Project (LEAP): design and methodologies to identify and validate stratification biomarkers for autism spectrum disorders. Mol Autism. 2017;8:24.
3. Di Martino A, Yan CG, Li Q, Denio E, Castellanos FX, Alaerts K, et al. The autism brain imaging data exchange: towards a large-scale evaluation of the intrinsic brain architecture in autism. Mol Psychiatry. 2014;19(6):659-67.
4. Pruim RHR, Mennes M, van Rooij D, Llera A, Buitelaar JK, Beckmann CF. ICA-AROMA: A robust ICA-based strategy for removing motion artifacts from fMRI data. Neuroimage. 2015;112:267-77.
5. Parkes L, Fulcher B, Yucel M, Fornito A. An evaluation of the efficacy, reliability, and sensitivity of motion correction strategies for resting-state functional MRI. Neuroimage. 2018;171:415-36.
6. Rutter M, Le Couteur A, Lord C. Autism diagnostic interview-revised. Los Angeles, CA: Western Psychological Services. 2003;29(2003):30.
7. Constantino JN, Gruber CP. Social responsiveness scale: SRS-2: Western Psychological Services Torrance, CA; 2012.
8. Wechsler D. Wechsler abbreviated scale of intelligence. 1999.
9. McIntosh D, Miller L, Shyu V, Dunn W. Development and validation of the short sensory profile. Sensory profile manual. 1999:59-73.
10. Baron-Cohen S, Wheelwright S, Skinner R, Martin J, Clubley E. The autism-spectrum quotient (AQ): Evidence from asperger syndrome/high-functioning autism, malesand females, scientists and mathematicians. Journal of autism and developmental disorders. 2001;31:5-17.