Iowa Brain‐Behavior Modeling Toolkit: An Open‐Source MATLAB Tool for Inferential and Predictive Modeling of Imaging‐Behavior and Lesion‐Deficit Relationships

Human Brain Mapping, Journal Year: 2024, Volume and Issue: 45(18)

Published: Dec. 15, 2024

ABSTRACT The traditional analytical framework taken by neuroimaging studies in general, and lesion‐behavior particular, has been inferential nature focused on identifying interpreting statistically significant effects within the sample under study. While this is well‐suited for hypothesis testing approaches, achieving modern goal of precision medicine requires a different that predictive focuses maximizing power models evaluating their ability to generalize beyond data were used train them. However, few tools exist support development evaluation context or research, creating an obstacle widespread adoption modeling approaches field. Further, existing analysis are often unable accommodate categorical outcome variables impose restrictions predictor data. Researchers therefore must use software packages depending (a) whether they addressing classification versus regression problem (b) correspond binary lesion images, continuous lesion‐network connectivity matrices, other modalities. To address these limitations, we have developed MATLAB toolkit supports both frameworks, accommodates problems, does not modality features graphical user interface scripting interface, includes implementations multiple mass‐univariate, multivariate, machine learning models, built‐in customizable routines hyper‐parameter optimization, cross‐validation, model stacking, significance testing, automatically generates text‐based descriptions key methodological details results improve reproducibility minimize errors reporting methods results. Here, provide overview discussion toolkit's demonstrate its functionality applying it question how expressive receptive language impairments relate location, structural disconnection, functional network disruption large patients with left hemispheric brain lesions. We find most strongly associated lateral prefrontal posterior temporal/parietal damage, respectively. also vs. partially overlapping patterns fronto‐temporal disconnection similar networks. Importantly, location lesion‐derived measures highly types impairment, predictions from trained explaining ~30%–40% variance average when applied models. made publicly available, included comprehensive set tutorial notebooks new users studies.

Language: Английский

Iowa Brain-Behavior Modeling Toolkit: An Open-Source MATLAB Tool for Inferential and Predictive Modeling of Imaging-Behavior and Lesion-Deficit Relationships

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 1, 2024

The traditional analytical framework taken by neuroimaging studies in general, and lesion-behavior particular, has been inferential nature focused on identifying interpreting statistically significant effects within the sample under study. While this is well-suited for hypothesis testing approaches, achieving modern goal of precision medicine requires a different that predictive focuses maximizing power models evaluating their ability to generalize beyond data were used train them. However, few tools exist support development evaluation context or research, creating an obstacle widespread adoption modeling approaches field. Further, existing analysis are often unable accommodate categorical outcome variables impose restrictions predictor data. Researchers therefore must use software packages depending whether they addressing classification vs. regression problem correspond binary lesion images, continuous lesion-network connectivity matrices, other modalities. To address these limitations, we have developed MATLAB toolkit supports both frameworks, accommodates problems, does not modality features graphical user interface scripting interface, includes implementations multiple mass-univariate, multivariate, machine learning models, built-in customizable routines hyper-parameter optimization, cross-validation, model stacking, significance testing, automatically generates text-based descriptions key methodological details results improve reproducibility minimize errors reporting methods results. Here, provide overview discussion demonstrate its functionality applying it question how expressive receptive language impairments relate location, structural disconnection, functional network disruption large patients with left hemispheric brain lesions. We find most strongly associated lateral prefrontal posterior temporal/parietal damage, respectively. also partially overlapping patterns fronto-temporal networks similar. Importantly, location lesion-derived measures highly types impairment, predictions from trained explaining ~30-40% variance average when applied models. made publicly available, included comprehensive set tutorial notebooks new users studies.

Language: Английский