Fast construction of interpretable whole-brain decoders

ABSTRACT: Summary Researchers often seek to decode mental states from brain activity measured with functional MRI. Rigorous decoding requires the use of formal neural prediction models, which are likely to be the most accurate if they use the whole brain. However, the computational burden and lack of interpretability of off-the-shelf statistical methods can make whole-brain decoding challenging. Here, we propose a method to build whole-brain neural decoders that are both interpretable and computationally efficient. We extend the partial least squares algorithm to build a regularized model with variable selection that offers a unique “fit once, tune later” approach: users need to fit the model only once and can choose the best tuning parameters post hoc. We show in real data that our method scales well with increasing data size and yields interpretable predictors. The algorithm is publicly available in multiple languages in the hope that interpretable whole-brain predictors can be implemented more widely in neuroimaging research. Graphical abstract Highlights • T-PLS offers interpretable whole-brain multivariate decoders with minimal computation• “Fit once, tune later” model tuning via cross-validation has nearly zero computation• In real data, T-PLS shows highest predictive performance and fitting speed• Users can decide post hoc the balance between predictive power and model parsimony Motivation Creating whole-brain predictors using functional MRI data can be challenging, especially in large datasets due to the computational burden of large number of features, large numbers of observations, and cross-validation. Our method exploits the analytical properties of the partial least squares algorithm to significantly reduce model fitting time as well as provide cross-validation-based tuning with nearly zero computational overhead. Lee et al. propose a thresholded partial least squares (T-PLS) algorithm for building interpretable whole-brain decoders using high-dimensional fMRI data. T-PLS achieves fast, scalable computation times while offering variable selection via cross-validation to achieve interpretability. In real data, T-PLS shows higher predictive performance than extant methods.

SUBMITTER: Lee S 

PROVIDER: S-EPMC9243546 | biostudies-literature |

REPOSITORIES: biostudies-literature