Biology subjects

de la Vega, A.

Publications and source records attributed to de la Vega, A..

8 recordsLinked to original sources

bioRxiv2025

NiCLIP: Neuroimaging contrastive language-image pretraining model for predicting text from brain activation images

Predicting cognitive processes from brain activation maps has remained an open question within the neuroscience community for many years. Meta-analytic functional decoding methods aim to tackle this issue by providing a quantitative estimation of behavioral profiles associated with specific brain regions. Existing methods face intrinsic challenges in neuroimaging meta-analysis, particularly in consolidating textual information from publications, as they rely on limited metrics that do not capture the semantic context of the text. The combination of large language models (LLMs) with advanced deep contrastive learning models (e.g., CLIP) for aligning text with images has revolutionized neuroimaging meta-analysis, potentially offering solutions to functional decoding challenges. In this work, we present NiCLIP, a contrastive language-image pretrained model that predicts cognitive tasks, concepts, and domains from brain activation patterns. We leveraged over 23,000 neuroscientific articles to train a CLIP model for text-to-brain association. Evaluation of NiCLIP predictions revealed that performance is optimized when using full-text articles instead of abstracts, as well as a curated cognitive ontology with precise task-concept-domain mappings. Furthermore, domain-specific fine-tuned LLMs (e.g., BrainGPT models) show numerically similar performance to their base LLM counterparts. Our results indicated that NiCLIP accurately predicts cognitive tasks from group-level activation maps provided by the Human Connectome Project across multiple domains (e.g., emotion, language, motor) and precisely characterizes the functional roles of specific brain regions, including the amygdala, hippocampus, and temporoparietal junction. However, NiCLIP showed limitations with noisy subject-level activation maps. NiCLIP represents a significant advancement in quantitative functional decoding for neuroimaging, offering researchers a powerful tool for hypothesis generation and scientific discovery.

neuroscience↗

bioRxiv2025

Advancing image-based meta-analysis for fMRI: A framework for leveraging NeuroVault data

Image-based meta-analysis (IBMA) is a powerful method for synthesizing results from various fMRI studies. However, challenges related to data accessibility and the lack of available tools and methods have limited its widespread use. This study examined the current state of the NeuroVault repository and developed a comprehensive framework for selecting and analyzing neuroimaging statistical maps within it. By systematically assessing the quality of NeuroVault's data and implementing novel selection and meta-analysis techniques, we demonstrated the repository's potential for IBMA. We created a multi-stage selection framework that includes preliminary, heuristic, and manual image selection methods. We conducted meta-analyses for three distinct domains: working memory, motor, and emotion processing. The results from the three manual IBMA approaches closely resembled reference maps from the Human Connectome Project. Importantly, we found that while manual selection provides the most precise results, heuristic methods can serve as robust alternatives, especially for domains that include a heterogeneous set of fMRI tasks and contrasts, such as emotion processing. Additionally, we evaluated five different meta-analytic estimator methods to assess their effectiveness in handling spurious images. For domains characterized by heterogeneous tasks, employing a robust estimator (e.g., median) is essential. This study is the first to present a systematic approach for implementing IBMA using the NeuroVault repository. We introduce an accessible and reproducible methodology that allows researchers to make the most of NeuroVault's extensive neuroimaging resources, potentially fostering greater interest in data sharing and future meta-analyses utilizing NeuroVault data.

neuroscience↗

bioRxiv2025

Meta all the way down: An overview of neuroimaging meta-analyses

Meta-analyses are invaluable tools for navigating the rapidly expanding scientific literature. Given their high value, ensuring the quality of meta-analyses is paramount. We conducted a multifaceted overview, examining each step in a manual neuroimaging meta-analysis on a large scale. We used four novel datasets comprising over 14,000 papers, including fMRI meta-analyses, fMRI studies, studies included in meta-analyses, and studies associated with image data on NeuroVault. Regarding successes, two-thirds of meta-analyses stated that they followed PRISMA guidelines, and 65% included a flowchart describing their inclusion process. We point out several areas for improvement. Pre-registration was fairly rare (20%), and only half listed their exact search strategy. There could be a location bias in which papers are included, and many did not include enough studies to be robust against publication bias (68% of meta analyses have less than 30 studies included). We also offer ideas for future directions. As image based meta-analysis is the gold standard, we have indicated which topics have the most image data available. The potential redundancy of topics can be visualized in our paper, and we recommend future meta-analyses be in conversation with past ones by citing and discussing previous similar work. By addressing these findings, the neuroimaging community can collectively improve the field of neuroimaging meta-analyses.

neuroscience↗

bioRxiv2023

Meta-analytic decoding of the cortical gradient of functional connectivity

Macroscale gradients have emerged as a central principle for understanding functional brain organization. Previous studies have demonstrated that a principal gradient of connectivity in the human brain exists, with unimodal primary sensorimotor regions situated at one end and transmodal regions associated with the default mode network and representative of abstract functioning at the other. The functional significance and interpretation of macroscale gradients remains a central topic of discussion in the neuroimaging community, with some studies demonstrating that gradients may be described using meta-analytic functional decoding techniques. However, additional methodological development is necessary to fully leverage available meta-analytic methods and resources and quantitatively evaluate their relative performance. Here, we conducted a comprehensive series of analyses to investigate and improve the framework of data-driven, meta-analytic methods, thereby establishing a principled approach for gradient segmentation and functional decoding. We found that a two-segment solution determined by a k-means segmentation approach and an LDA-based meta-analysis combined with the NeuroQuery database was the optimal combination of methods for decoding functional connectivity gradients. Finally, we proposed a method for decoding additional components of the gradient decomposition. The current work aims to provide recommendations on best practices and flexible methods for gradient-based functional decoding of fMRI data.

neuroscience↗

bioRxiv2022

Neuroscout, a unified platform for generalizable and reproducible fMRI research

Functional magnetic resonance imaging (fMRI) has revolutionized cognitive neuroscience, but methodological barriers limit the generalizability of findings from the lab to the real world. Here, we present Neuroscout, an end-to-end platform for analysis of naturalistic fMRI data designed to facilitate the adoption of robust and generalizable research practices. Neuroscout leverages state-of-the-art machine learning models to automatically annotate stimuli from dozens of naturalistic fMRI studies, allowing researchers to easily test neuroscientific hypotheses across multiple ecologically-valid datasets. In addition, Neuroscout builds on a robust ecosystem of open tools and standards to provide an easy-to-use analysis builder and a fully automated execution engine that reduce the burden of reproducible research. Through a series of meta-analytic case studies, we validate the automatic feature extraction approach and demonstrate its potential to support more robust fMRI research. Owing to its ease of use and a high degree of automation, Neuroscout makes it possible to overcome modeling challenges commonly arising in naturalistic analysis and to easily scale analyses within and across datasets, democratizing generalizable fMRI research.

neuroscience↗

bioRxiv2019

Analysis of task-based functional MRI data preprocessed with fMRIPrep

Functional magnetic resonance imaging (fMRI) is a standard tool to investigate the neural correlates of cognition. fMRI noninvasively measures brain activity, allowing identification of patterns evoked by tasks performed during scanning. Despite the long history of this technique, the idiosyncrasies of each dataset have led to the use of ad-hoc preprocessing protocols customized for nearly every different study. This approach is time-consuming, error-prone, and unsuitable for combining datasets from many sources. Here we showcase fMRIPrep (http://fmriprep.org), a robust tool to prepare human fMRI data for statistical analysis. This software instrument addresses the reproducibility concerns of the established protocols for fMRI preprocessing. By leveraging the Brain Imaging Data Structure (BIDS) to standardize both the input datasets --MRI data as stored by the scanner-- and the outputs --data ready for modeling and analysis--, fMRIPrep is capable of preprocessing a diversity of datasets without manual intervention. In support of the growing popularity of fMRIPrep, this protocol describes how to integrate the tool in a task-based fMRI investigation workflow.

bioinformatics↗

bioRxiv2017

Heterogeneity Within the Frontoparietal Control Network and its Relationship to the Default and Dorsal Attention Networks

The frontoparietal control network (FPCN) plays a central role in executive control. It has been predominantly viewed as a unitary domain general system. Here, we examined patterns of FPCN functional connectivity (FC) across multiple conditions of varying cognitive demands, in order to test for FPCN heterogeneity. We identified two distinct subsystems within the FPCN based on hierarchical clustering and machine learning classification analyses of within-FPCN FC patterns. These two FPCN subsystems exhibited distinct patterns of FC the default network (DN) and the dorsal attentional network (DAN). This 2-fold FPCN differentiation was observed across four independent data sets, across 9 different conditions (rest and 8 tasks), as well as in meta-analytic co-activation patterns. The extent of FPCN differentiation varied across conditions, suggesting flexible adaptation to task demands. These findings reveal a flexible and heterogeneous FPCN organization that may in part emerge from separable DN and DAN processing streams.

neuroscience↗

bioRxiv2016

Large-scale meta-analysis suggests low regional modularity in lateral frontal cortex.

Extensive fMRI study of human lateral frontal cortex (LFC) has yet to yield a consensus mapping between discrete anatomy and psychological states, partly due to the difficulty of inferring mental states in individual studies. Here, we used a data-driven approach to generate a comprehensive functional-anatomical mapping of LFC from 11,406 neuroimaging studies. We identified putatively separable LFC regions on the basis of whole-brain co-activation, revealing 14 clusters organized into three whole-brain networks. Next, we used multivariate classification to identify the psychological states that best predicted activity in each sub-region, resulting in preferential psychological profiles. We observed large functional differences between networks, suggesting brain networks support distinct modes of processing. Within each network, however, we observed low functional specificity, suggesting discrete psychological states are not modularly organized. Our results are consistent with the view that individual LFC regions work as part of highly parallel, distributed networks to give rise to flexible, adaptive behavior.

neuroscience↗