Biology subjects

Hebling Vieira, B.

Publications and source records attributed to Hebling Vieira, B..

4 recordsLinked to original sources

bioRxiv2026

Predicting children's literacy from task-based fMRI: Neural heterogeneity and task-dependent performance

Reading is a complex skill with a well-characterized neural basis. Multivariate fMRI analyses have deepened our neuroscientific understanding of literacy by linking neural patterns to behavioral traits. Although task-based fMRI often outperforms resting-state fMRI in predicting cognitive traits, few studies have applied it to continuous measures of childrens reading ability. To identify neural markers of literacy, we compared predictive performance across multiple fMRI tasks and reading-related measures. In this data-driven study, we predicted literacy skills in school-aged children (6.7-10.3 years) from eleven behavioral scores grouped into Reading (fluency and comprehension), Verbal (vocabulary knowledge and verbal intelligence), and Naming (object naming speed). Predictive performance was examined across four fMRI tasks completed by subgroups of children (n = 73-97): two active tasks - phonological-lexical decisions (PhonLex) and audiovisual character learning (Learn) - and two passive tasks - word and face viewing (Localizer) and character processing (CharProc). Individual activation contrast maps, categorized as simple (single condition) or subtractive (condition contrasts), were analyzed using a machine learning model with whole-brain predictors derived from principal component analysis. Results showed the highest predictive performance for Reading and Naming with PhonLex > Learn > Localizer = CharProc, and for Verbal with PhonLex = Learn > Localizer = CharProc. Simple contrasts generally outperformed subtractive contrasts in predicting behavioral scores. Key neural predictors, identified through whole-brain and region-of-interest analyses, included the left inferior frontal gyrus, supramarginal gyrus, ventral occipitotemporal cortex, insula, and default mode network regions. Together, these findings indicate that, for predicting literacy traits in children, active tasks and tasks that engage brain systems involved in multisensory learning tend to outperform both passive paradigms and simple subtractive task contrasts. This study provides a methodological benchmark for brain-based prediction of reading ability and highlights the value of activation heterogeneity across distributed regions as a potential marker for tracking literacy development over time.

neuroscience↗

bioRxiv2025

Toward Robust Neuroanatomical Normative Models: Influence of Sample Size and Covariates Distributions

Normative models are increasingly used to characterize individual-level brain deviations in neuroimaging studies, but their performance depends heavily on the reference sample used for training or adaptation. In this study, we systematically investigated how sample size and covariate composition of the reference cohort influence model fit, deviation estimates, and clinical readouts in Alzheimers disease (AD). Using a discovery dataset (OASIS-3, n = 1032), we trained models on healthy control (HC), subsamples ranging from 5 to 600 individuals, while varying age and sex distributions to simulate biases in reference populations. We further assessed the use of adaptive transfer learning by pre-training models on the UK Biobank (n = 42,747) and adapting them to the clinical dataset applying the same sub-sampling strategies. We evaluated model performance on a fixed HC test set and quantified deviation score errors, outlier detection, and classification accuracy in both the HC test set and the AD cohort. The findings were replicated in an external validation sample (AIBL, n = 463). Across all settings, model performance improved with increasing sample size, but demographic alignment of the covariates, particularly in age, was essential for reliable deviation estimates. Models trained directly within the dataset achieved stable fit with approximately 200 HCs, while adapted models reached comparable performance with as few as 50 individuals when pre-trained on large-scale data. These results show that robust individual-level modeling can be achieved using moderately sized but demographically matched cohorts, supporting broader application of normative modeling in ageing and neurodegeneration research.

neuroscience↗

bioRxiv2022

The AI Neuropsychologist: Automatic scoring of memory deficits with deep learning

BackgroundMemory deficits are a hallmark of many different neurological and psychiatric conditions. The Rey-Osterrieth complex figure (ROCF) is the state-of-the-art assessment tool for neuropsychologists across the globe to assess the degree of non-verbal visual memory deterioration. To obtain a score, a trained clinician inspects a patients ROCF drawing and quantifies deviations from the original figure. This manual procedure is time-consuming, slow and scores vary depending on the clinicians experience, motivation and tiredness. MethodsHere, we leverage novel deep learning architectures to automatize the rating of memory deficits. For this, we collected more than 20k hand-drawn ROCF drawings from patients with various neurological and psychiatric disorders as well as healthy participants. Unbiased ground truth ROCF scores were obtained from crowdsourced human intelligence. This dataset was used to train and evaluate a multi-head convolutional neural network. ResultsThe model performs highly unbiased as it yielded predictions very close to the ground truth and the error was similarly distributed around zero. The neural network outperforms both online raters and clinicians. The scoring system can reliably identify and accurately score individual figure elements in previously unseen ROCF drawings, which facilitates explainability of the AI-scoring system. To ensure generalizability and clinical utility, the model performance was successfully replicated in a large independent prospective validation study that was pre-registered prior to data collection. ConclusionsOur AI-powered scoring system provides healthcare institutions worldwide with a digital tool to assess objectively, reliably and time-efficiently the performance in the ROCF test from hand-drawn images.

neuroscience↗

bioRxiv2021

On the prediction of human intelligence from neuroimaging: A systematic review of methods and reporting

Human intelligence is one of the main objects of study in cognitive neuroscience. Reviews and meta-analyses have proved to be fundamental to establish and cement neuroscientific theories on intelligence. The prediction of intelligence using in vivo neuroimaging data and machine learning has become a widely accepted and replicated result. Here, we present a systematic review of this growing area of research, based on studies that employ structural, functional, and/or diffusion MRI to predict human intelligence in cognitively normal subjects using machine-learning. We performed a systematic assessment of methodological and reporting quality, using the PROBAST and TRIPOD assessment forms and 30 studies identified through a systematic search. We observed that fMRI is the most employed modality, resting-state functional connectivity (RSFC) is the most studied predictor, and the Human Connectome Project is the most employed dataset. A meta-analysis revealed a significant difference between the performance obtained in the prediction of general and fluid intelligence from fMRI data, confirming that the quality of measurement moderates this association. The expected performance of studies predicting general intelligence from fMRI was estimated to be r = 0.42 (CI95% = [0.35, 0.50]) while for studies predicting fluid intelligence obtained from a single test, expected performance was estimated as r = 0.15 (CI95% = [0.13, 0.17]). We further enumerate some virtues and pitfalls we identified in the methods for the assessment of intelligence and machine learning. The lack of treatment of confounder variables, including kinship, and small sample sizes were two common occurrences in the literature which increased risk of bias. Reporting quality was fair across studies, although reporting of results and discussion could be vastly improved. We conclude that the current literature on the prediction of intelligence from neuroimaging data is reaching maturity. Performance has been reliably demonstrated, although extending findings to new populations is imperative. Current results could be used by future works to foment new theories on the biological basis of intelligence differences.

neuroscience↗