oFVSD: a Python package of optimized forward variable selection decoder for high-dimensional neuroimaging data

• DOI: 10.3389/fninf.2023.1266713
• Publication Date: 2023-09-27T09:37:37Z
• Authors: Tung Dang; Alan S. R. Fermin; Maro G. Machizawa
• Keywords: Hyperparameter

The complexity and high dimensionality of neuroimaging data pose problems for decoding information with machine learning (ML) models because the number features is often much larger than observations. Feature selection one crucial steps determining meaningful target in decoding; however, optimizing feature from such high-dimensional has been challenging using conventional ML models. Here, we introduce an efficient high-performance package incorporating a forward variable (FVS) algorithm hyper-parameter optimization that automatically identifies best pairs both classification regression models, where total 18 are implemented by default. First, FVS evaluates goodness-of-fit across different k-fold cross-validation step subset based on predefined criterion each model. Next, hyperparameters model optimized at iteration. Final outputs highlight selected (brain regions interest) its accuracy. Furthermore, toolbox can be executed parallel environment computation typical personal computer. With decoder (oFVSD) pipeline, verified effectiveness sex age range 1,113 structural magnetic resonance imaging (MRI) datasets. Compared to without Boruta as counterpart, demonstrate oFVSD significantly outperformed all over counterpart (approximately 0.20 increase correlation coefficient, r , 8% average) 0.07 improvement 4% models). confirmed use considerably reduced computational burden MRI data. Altogether, efficiently effectively improves performance providing case example flexibility, potential many other modalities neuroimaging. This open-source freely available Python makes it valuable research communities seeking improved