Classifying First-Episode Schizophrenia Using Functional Connectivity Patterns: A Comparative Study.
Poster No:
1231
Submission Type:
Abstract Submission
Authors:
David Tomecek1, Zbynek Pitra2, Dominik Klepl2, Jaroslav Tintera1, Filip Spaniel1, Jaroslav Hlinka2
Institutions:
1National Institute of Mental Health, Klecany, Czech Republic, 2Institute of Computer Science, The Czech Academy of Sciences, Prague, Czech Republic
First Author:
Co-Author(s):
Introduction:
The abnormal developmental processes of the brain likely appear long before clinical symptoms of schizophrenia (Stachowiak et al., 2013). Several studies have been conducted describing impaired connectivity between multiple brain networks in schizophrenia (see Dong et al. (2018) or Ruiz-Torras et al. (2023)). However, the reported findings are highly inconsistent, possibly due to a priori choice of selected brain regions of interest, heterogenic clinical samples (Li et al., 2016), or underpowered studies (Dong et al., 2018). In this study, we utilized resting-state fMRI data to build a model for the classification of healthy controls and patients with a first episode of schizophrenia.
Methods:
We obtained resting-state fMRI data from 190 subjects in total - 90 healthy controls (average age/SD: 27.69/6.82) and 100 patients (average age/SD: 28.75/6.83). A default preprocessing pipeline for volume-based analyses (direct normalization to MNI-space) was applied to fMRI data in the CONN toolbox for Matlab 2016b. In the next step, we used two popular data extraction techniques - atlas-based - using Automated Anatomical Labeling (AAL) atlas with 90 regions (Tzourio-Mazoyer et al., 2002) and the Craddock atlas with 200 regions (Craddock et al., 2012) and - data-driven - decomposition of the functional data using spatial independent component analysis (ICA) in the GIFT toolbox for Matlab 2016b. In the ICA approach, we created a full set of 27 components (ICA27) and a selected subset of nine components (ICA9) that matched those from Arbabshirani et al. (2013). All extracted time series were band-pass filtered with a window of 0.017−0.15 Hz and linearly detrended. Eventually, we computed functional connectivity (FC) between 90 regions of the AAL atlas (AAL90) and 200 regions of the Craddock atlas (Craddock200) and functional network connectivity (FNC) between the independent components separately for the ICA27 and ICA9 data sets. In addition to the data sets with the original number of features, we also created their PCA-reduced version into 36 principal components. Ten randomly selected subjects were removed from the patients' group to balance the number of subjects in both classes. We used a set of multiple classifiers implemented directly in Matlab 2016b, including linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), support vector machine (SVM) with linear, quadratic, polynomial, and RBF kernels, k-nearest neighbor (KNN), Naive Bayes (NB), random forest (RF) in default, with bagging or boosting, and decision tree (DT) in default. Hyper-parameter values were selected by a grid search performed on the training set and evaluated using a five-fold cross-validation. All classifiers were evaluated using a leave-one-out cross-validation scheme.
Results:
In terms of the highest achieved accuracy, see Fig. 1, we reached 86.11% with the AAL90 data set (linear SVM+PCA), which outperformed both ICA9 (71.11%, NB) and ICA27 (72.78%, NB+PCA). The Craddock200 features performed slightly better than all ICA-based models but still worse than the AAL90 features. Similarly, the dimensionality reduction did not substantially decrease the performance of the classifiers in general. In fact, in many cases, it has led to considerable improvement.
Conclusions:
On average, features from atlas-based functional connectivity performed better compared to ICA-based functional network connectivity features. We found the best classification accuracy with features from the AAL90 data set that achieved 86.11% using the linear SVM classifier. On average, linear classifiers performed slightly better on all data sets compared to other algorithms. Reduction of the data dimension using PCA before running the classifier might be warranted, as it generally improved the performance across other design choices.
Acknowledgement: The publication was supported by ERDF-Project Brain dynamics No. CZ.02.01.01/00/22_008/0004643, GACR project No. 23-07074S, AZV project No. NU21-08-00432.
Acknowledgement: The publication was supported by ERDF-Project Brain dynamics No. CZ.02.01.01/00/22_008/0004643, GACR project No. 23-07074S, AZV project No. NU21-08-00432.
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Connectivity (eg. functional, effective, structural) 1
fMRI Connectivity and Network Modeling
Task-Independent and Resting-State Analysis
Novel Imaging Acquisition Methods:
BOLD fMRI
Keywords:
FUNCTIONAL MRI
Machine Learning
Psychiatric Disorders
Schizophrenia
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.
Craddock, R. C., James, G. A., Holtzheimer, P. E., Hu, X. P., & Mayberg, H. S. (2012). A whole brain fMRI atlas generated via spatially constrained spectral clustering. Human Brain Mapping, 33(8), 1914–1928. https://doi.org/10.1002/hbm.21333
Dong, D., Wang, Y., Chang, X., Luo, C., & Yao, D. (2018). Dysfunction of Large-Scale Brain Networks in Schizophrenia: A Meta-analysis of Resting-State Functional Connectivity. Schizophrenia Bulletin, 44(1), 168–181. https://doi.org/10.1093/schbul/sbx034
Li, T., Wang, Q., Zhang, J., Rolls, E. T., Yang, W., Palaniyappan, L., Zhang, L., Cheng, W., Yao, Y., Liu, Z., Gong, X., Luo, Q., Tang, Y., Crow, T. J., Broome, M. R., Xu, K., Li, C., Wang, J., Liu, Z., … Feng, J. (2016). Brain-Wide Analysis of Functional Connectivity in First-Episode and Chronic Stages of Schizophrenia. Schizophrenia Bulletin, sbw099. https://doi.org/10.1093/schbul/sbw099
Ruiz-Torras, S., Gudayol-Ferré, E., Fernández-Vazquez, O., Cañete-Massé, C., Peró-Cebollero, M., & Guàrdia-Olmos, J. (2023). Hypoconnectivity networks in schizophrenia patients: A voxel-wise meta-analysis of Rs-fMRI. International Journal of Clinical and Health Psychology, 23(4), 100395. https://doi.org/10.1016/j.ijchp.2023.100395
Stachowiak, M. K., Kucinski, A., Curl, R., Syposs, C., Yang, Y., Narla, S., Terranova, C., Prokop, D., Klejbor, I., Bencherif, M., Birkaya, B., Corso, T., Parikh, A., Tzanakakis, E. S., Wersinger, S., & Stachowiak, E. K. (2013). Schizophrenia: A neurodevelopmental disorder — Integrative genomic hypothesis and therapeutic implications from a transgenic mouse model. Schizophrenia Research, 143(2–3), 367–376. https://doi.org/10.1016/j.schres.2012.11.004
Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix, N., Mazoyer, B., & Joliot, M. (2002). Automated Anatomical Labeling of Activations in SPM Using a Macroscopic Anatomical Parcellation of the MNI MRI Single-Subject Brain. NeuroImage, 15(1), 273–289. https://doi.org/10.1006/nimg.2001.0978