Kaustubh R. Patil

This study links human brain functional connectivity to neurotransmitter systems using resting-state fMRI and MEG across multiple cohorts, showing consistent patterns and clinical relevance.

The paper compares brain activity during self-caught and probe-caught mind blanking events using fMRI, finding behavioral convergence but neural divergence in the dorsal anterior cingulate cortex.

97% of 184 high-impact biomedical AI studies compare cross-validated models with tests assuming fold independence; in 69% of re-analyzable studies a significant comparison does not survive correction.

The paper predicts brain volumes from anthropometric and demographic features using UK Biobank data, finding strong sex-based predictions for TIV and age-based predictions for other volumes.

OSCAR detects condition-related deviations in brain connectivity using a normative modeling framework, identifying task-evoked and clinical network reorganization with high accuracy.

This study uses machine learning on UK Biobank data to predict sleep health traits from brain imaging, finding limited predictive power despite extensive model optimization.

This study investigates the off-target antibacterial effects of 1076 industrial and agricultural chemicals on 22 human gut bacteria, revealing 588 previously unreported interactions.

This paper identifies nine functionally distinct subregions within the left prefrontal cortex targeted by TMS, revealing their network connectivity patterns and behavioral associations to guide individualized TMS interventions.

This study identifies genetic variants influencing brain geometry using GWAS on Laplace-Beltrami spectrum eigenvalues from UK Biobank data.

This study compares voxel-wise and region-wise denoising in functional connectivity analyses, finding region-level denoising offers equal or better performance with reduced computational cost.

This study examines how the number of gradients used in Procrustes alignment affects identification accuracy and differential identifiability in functional connectivity analyses across multiple fMRI datasets.

The paper compares age-related brain deterioration between humans and chimpanzees, finding a unique positive relationship between cerebral aging and cortical expansion in humans.

This paper examines how functional connectivity patterns from task and resting states predict individual behavior in cognitive, social, and affective domains, finding limited success overall.

This study evaluates how sex classification models generalize across brain connectivity datasets, finding that combining multiple datasets improves generalization performance.

This study identifies a latent clinical-anatomical dimension linking metabolic syndrome severity to brain morphology and cognition in 40,087 individuals.

This commentary critiques the use of geometric eigenmodes as a model for brain function, arguing that they may overfit and fail to explain functional mechanisms beyond geometry.

This study investigates the prediction of executive functioning from brain networks in young and old adults, finding low prediction accuracy and modality-specific effects.

The paper develops non-TIV-biased models for accurate sex classification of cisgender and transgender individuals using brain morphology data.

This study systematically compares five VBM pipelines for age prediction using three large datasets, finding significant differences in their ability to quantify GMV and predict age.

This paper introduces a predictive framework for analyzing naturalistic viewing fMRI data to understand individual differences in brain activity and behavior.

This study compares the predictive power of resting-state fMRI features and individual characteristics for cognitive phenotypes in 20,000 UK Biobank subjects, finding individual characteristics generally perform better.

This paper compares 128 machine learning workflows for brain-age prediction using neuroimaging data, finding that feature representation and algorithm choice affect performance, and that brain-age delta correlates differently within-site and cross-site.

Low phenotypic measurement reliability severely limits out-of-sample brain-behaviour prediction accuracy and prevents models from fully benefiting when neuroimaging sample sizes scale from hundreds to thousands of participants.

This study uses simulated whole-brain connectomes to classify Parkinson's disease patients against healthy controls, showing improved classification when combining empirical and simulated data.

The paper evaluates how different levels of co-fluctuation in functional connectivity predict individual-level phenotypes using machine learning.

The paper demonstrates that regional homogeneity (ReHo) fMRI metrics can uniquely identify individuals and correlate with intelligence, showing high stability and potential for precision medicine.

This paper introduces a connectivity-based framework to predict psychometric variables from brain connectivity and evaluates its performance compared to activation-based approaches.

The paper predicts individual schizophrenia symptom dimensions using functional brain networks and links these networks to molecular features like dopamine and serotonin systems.

This study uses machine learning to classify sex based on resting state brain connectivity across three samples, finding reliable classification across samples and identifying key brain regions associated with sex differences.