Feature selection framework for functional connectome fingerprinting

Contrastive learning for neural fingerprinting from limited neuroimaging data

Introduction

Neural fingerprinting is a technique used to identify individuals based on their unique brain activity patterns. While deep learning techniques have been demonstrated to outperform traditional correlation-based methods, they often require retraining to accommodate new subjects. Furthermore, the limited availability of samples in neuroscience research can impede the quick adoption of deep learning methods, presenting a challenge for their broader application in neural fingerprinting.

Methods

This study addresses these challenges by using contrastive learning to eliminate the need for retraining with new subjects and developing a data augmentation methodology to enhance model robustness in limited sample size conditions. We utilized the LEMON dataset, comprising 3 Tesla MRI and resting-state fMRI scans from 138 subjects, to compute functional connectivity as a baseline for fingerprinting performance based on correlation metrics. We adapted a recent deep learning model by incorporating data augmentation with short random temporal segments for training and reformulated the fingerprinting task as a contrastive problem, comparing the efficacy of contrastive triplet loss against conventional cross-entropy loss.

Results

The results of this study confirm that deep learning methods can significantly improve fingerprinting performance over correlation-based methods, achieving an accuracy of about 98% in identifying a single subject out of 138 subjects utilizing 39 different functional connectivity profiles.

Discussion

The contrastive method showed added value in the "leave subject out" scenario, demonstrating flexibility comparable to correlation-based methods and robustness across different data sizes. These findings suggest that contrastive learning and data augmentation offer a scalable solution for neural fingerprinting, even with limited sample sizes.

Connectome-based fingerprinting: reproducibility, precision, and behavioral prediction

Functional magnetic resonance imaging-based functional connectivity enables the non-invasive mapping of individual differences in brain functional organization to individual differences in a vast array of behavioral phenotypes. This flexibility has renewed the search for neuroimaging-based biomarkers that exhibit reproducibility, prediction, and precision. Functional connectivity-based measures that share these three characteristics are key to achieving this goal. Here, we review the functional connectome fingerprinting approach and discuss its value, not only as a simple and intuitive conceptualization of the "functional connectome" that provides new insights into how the connectome is altered in association with psychiatric symptoms, but also as a straightforward and interpretable method for indexing the reproducibility of functional connectivity-based measures. We discuss how these advantages provide new avenues for strengthening reproducibility, precision, and behavioral prediction for functional connectomics and we consider new directions toward discovering better biomarkers for neuropsychiatric conditions.

Brain fingerprinting and cognitive behavior predicting using functional connectome of high inter-subject variability

The functional connectivity (FC) graph of the brain has been widely recognized as a ``fingerprint'' that can be used to identify individuals from a group of subjects. Research has indicated that individual identification accuracy can be improved by eliminating the impact of shared information among individuals. However, current research extracts not only shared information of inter-subject but also individual-specific information from FC graphs, resulting in incomplete separation of shared information and fingerprint information among individuals, leading to lower individual identification accuracy across all functional magnetic resonance imaging (fMRI) states session pairs and poor cognitive behavior prediction performance. In this paper, we propose a method to enhance inter-subject variability combining conditional variational autoencoder (CVAE) network and sparse dictionary learning (SDL) module. By embedding fMRI state information in the encoding and decoding processes, the CVAE network can better capture and represent the common features among individuals and enhance inter-subject variability by residual. Our experimental results on Human Connectome Project (HCP) data show that the refined connectomes obtained by using CVAE with SDL can accurately distinguish an individual from the remaining participants. The success accuracies reached 99.7 % and 99.6 % in the session pair rest1-rest2 and reverse rest2-rest1, respectively. In the identification experiment involving task-task combinations carried out on the same day, the identification accuracies ranged from 94.2 % to 98.8 %. Furthermore, we showed the Frontoparietal and Default networks make the most significant contributions to individual identification and the edges that significantly contribute to individual identification are found within and between the Frontoparietal and Default networks. Additionally, high-level cognitive behaviors can also be better predicted with the obtained refined connectomes, suggesting that higher fingerprinting can be useful for resulting in higher behavioral associations. In summary, our proposed framework provides a promising approach to use functional connectivity networks for studying cognition and behavior, promoting a deeper understanding of brain functions.

Individual variability in functional connectivity of human auditory cortex

Individual variability in functional connectivity underlies individual differences in cognition and behaviors, yet its association with functional specialization in the auditory cortex remains elusive. Using resting-state functional magnetic resonance imaging data from the Human Connectome Project, this study was designed to investigate the spatial distribution of auditory cortex individual variability in its whole-brain functional network architecture. An inherent hierarchical axis of the variability was discerned, which radiates from the medial to lateral orientation, with the left auditory cortex demonstrating more pronounced variations than the right. This variability exhibited a significant correlation with the variations in structural and functional metrics in the auditory cortex. Four auditory cortex subregions, which were identified from a clustering analysis based on this variability, exhibited unique connectional fingerprints and cognitive maps, with certain subregions showing specificity to speech perception functional activation. Moreover, the lateralization of the connectional fingerprint exhibited a U-shaped trajectory across the subregions. These findings emphasize the role of individual variability in functional connectivity in understanding cortical functional organization, as well as in revealing its association with functional specialization from the activation, connectome, and cognition perspectives.

ConnSearch: A framework for functional connectivity analysis designed for interpretability and effectiveness at limited sample sizes

Functional connectivity studies increasingly turn to machine learning methods, which typically involve fitting a connectome-wide classifier, then conducting post hoc interpretation analyses to identify the neural correlates that best predict a dependent variable. However, this traditional analytic paradigm suffers from two main limitations. First, even if classifiers are perfectly accurate, interpretation analyses may not identify all the patterns expressed by a dependent variable. Second, even if classifiers are generalizable, the patterns implicated via interpretation analyses may not replicate. In other words, this traditional approach can yield effective classifiers while falling short of most neuroscientists' goals: pinpointing the neural correlates of dependent variables. We propose a new framework for multivariate analysis, ConnSearch, which involves dividing the connectome into components (e.g., groups of highly connected regions) and fitting an independent model for each component (e.g., a support vector machine or a correlation-based model). Conclusions about the link between a dependent variable and the brain are based on which components yield predictive models rather than on interpretation analysis. We used working memory data from the Human Connectome Project (N = 50-250) to compare ConnSearch with four existing connectome-wide classification/interpretation methods. For each approach, the models attempted to classify examples as being from the high-load or low-load conditions (binary labels). Relative to traditional methods, ConnSearch identified neural correlates that were more comprehensive, had greater consistency with the WM literature, and better replicated across datasets. Hence, ConnSearch is well-positioned to be an effective tool for functional connectivity research.

Reliability and subject specificity of personalized whole-brain dynamical models

Dynamical whole-brain models were developed to link structural (SC) and functional connectivity (FC) together into one framework. Nowadays, they are used to investigate the dynamical regimes of the brain and how these relate to behavioral, clinical and demographic traits. However, there is no comprehensive investigation on how reliable and subject specific the modeling results are given the variability of the empirical FC. In this study, we show that the parameters of these models can be fitted with a "poor" to "good" reliability depending on the exact implementation of the modeling paradigm. We find, as a general rule of thumb, that enhanced model personalization leads to increasingly reliable model parameters. In addition, we observe no clear effect of the model complexity evaluated by separately sampling results for linear, phase oscillator and neural mass network models. In fact, the most complex neural mass model often yields modeling results with "poor" reliability comparable to the simple linear model, but demonstrates an enhanced subject specificity of the model similarity maps. Subsequently, we show that the FC simulated by these models can outperform the empirical FC in terms of both reliability and subject specificity. For the structure-function relationship, simulated FC of individual subjects may be identified from the correlations with the empirical SC with an accuracy up to 70%, but not vice versa for non-linear models. We sample all our findings for 8 distinct brain parcellations and 6 modeling conditions and show that the parcellation-induced effect is much more pronounced for the modeling results than for the empirical data. In sum, this study provides an exploratory account on the reliability and subject specificity of dynamical whole-brain models and may be relevant for their further development and application. In particular, our findings suggest that the application of the dynamical whole-brain modeling should be tightly connected with an estimate of the reliability of the results.

Feature selection framework for functional connectome fingerprinting

The ability to uniquely characterize individual subjects based on their functional connectome (FC) is a key requirement for progress toward precision psychiatry. FC fingerprinting is increasingly studied in the neuroimaging community for this purpose, where a variety of approaches have been developed for effective FC fingerprinting. Recent independent studies showed that fingerprinting accuracy suffers at large sample sizes and when coarser parcellations are used for computing the FC. Quantifying this problem and understanding the reasons these factors impact fingerprinting accuracy is crucial to develop more accurate fingerprinting methods for large sample sizes. Part of the challenge in fingerprinting is that FC captures both generic and subject‐specific information. A systematic approach for identifying subject‐specific FC information is crucial for making progress in addressing the fingerprinting problem. In this study, we addressed three gaps in our understanding of the FC fingerprinting problem. First, we studied the joint effect of sample size and parcellation granularity. Second, we explained the reason for reduced fingerprinting accuracy with increased sample size and reduced parcellation granularity. To this end, we used a clustering quality metric from the data mining community. Third, we developed a general feature selection framework for systematically identifying resting‐state functional connectivity (RSFC) elements that capture information to uniquely identify subjects. In sum, we evaluated six different approaches from this framework by quantifying both subject‐specific fingerprinting accuracy and the decrease in accuracy with an increase in sample size to identify which approach improved quality metrics the most.