Sparse electrophysiological source imaging predicts aging-related gait speed slowing
Authors:
Vega-Hernandez,
Mayrim,
Galan-Garcia,
Lidice,
Perez-Hidalgo-Gato,
Jhoanna,
Ontivero-Ortega,
Marlis,
Garcia-Agustin,
Daysi,
Garcia-Reyes,
Ronaldo,
Bosch-Bayard,
Jorge,
Marinazzo,
Daniele,
Martinez-Montes,
Eduardo,
Valdes-Sosa,
Pedro A
Abstract:
Objective: We seek stable Electrophysiological Source Imaging (ESI) biomarkers associated with Gait Speed (GS) as a measure of functional decline. Towards this end we determine the predictive value of ESI activation and connectivity patterns of resting-state EEG Theta rhythm on physical performance decline measured by a slowing GS in aging individuals. Methods: As potential biomarkers related to G…
▽ More
Objective: We seek stable Electrophysiological Source Imaging (ESI) biomarkers associated with Gait Speed (GS) as a measure of functional decline. Towards this end we determine the predictive value of ESI activation and connectivity patterns of resting-state EEG Theta rhythm on physical performance decline measured by a slowing GS in aging individuals. Methods: As potential biomarkers related to GS changes, we estimate ESI using flexible sparse/smooth/non-negative models (NN-SLASSO), from which activation ESI (aESI) and connectivity ESI (cESI) features are selected using the Stable Sparse Classifier method. Results and Conclusions: Novel sparse aESI models outperformed traditional methods such as the LORETA family. The models combining aESI and cESI features improved the predictability of GS changes. Selected biomarkers from activation/connectivity patterns were localized to orbitofrontal and temporal cortical regions. Significance: The proposed methodology contributes to understanding the activation and connectivity of ESI complex patterns related to GS, providing potential biomarker features for GS slowing. Given the known relationship between GS decline and cognitive impairment, this preliminary work suggests it might be applied to other, more complex measures of healthy and pathological aging. Importantly, it might allow an ESI-based evaluation of rehabilitation programs.
△ Less
Submitted 22 August, 2023; v1 submitted 20 July, 2023;
originally announced July 2023.
Multiple penalized least squares and sign constraints with modified Newton-Raphson algorithms: application to EEG source imaging
Authors:
Mayrim Vega-Hernández,
José M. Sánchez-Bornot,
Agustín Lage-Castellanos,
Jhoanna Pérez-Hidalgo-Gato,
Darío Palmero-Ledón,
José E. Alvarez-Iglesias,
Daysi García-Agustin,
Eduardo Martínez-Montes,
Pedro A. Valdés-Sosa
Abstract:
Multiple penalized least squares (MPLS) models are a flexible approach to find adaptive least squares solutions required to be simultaneously sparse and smooth. This is particularly important when addressing real-life inverse problems where there is no ground truth available, such as electrophysiological source imaging. In this work we formalize a modified Newton-Raphson (MNR) algorithm to estimat…
▽ More
Multiple penalized least squares (MPLS) models are a flexible approach to find adaptive least squares solutions required to be simultaneously sparse and smooth. This is particularly important when addressing real-life inverse problems where there is no ground truth available, such as electrophysiological source imaging. In this work we formalize a modified Newton-Raphson (MNR) algorithm to estimate general MPLS models and propose its extension to perform efficient optimization over the active set of selected features (AMNR). This algorithm can be used to minimize continuously differentiable objective functions with multiple restrictions, including sign constraints. We show that these algorithms provide solutions with acceptable reconstruction in simulated scenarios that do not cope with model assumptions, and for low n/p ratios. We then use both algorithms for estimating different electroencephalography (EEG) inverse models with multiple penalties. We also show how the AMNR allows us to estimate new models in the EEG inverse problem context, such as nonnegative versions of Smooth Garrote and Smooth LASSO. Synthetic data were used for a comparison between solutions obtained using the two algorithms proposed here and the least angle regression (LARS) algorithm, according to well-known quality measures. A visual event-related EEG from healthy young subjects and a resting-state EEG study on the relationship between cognitive ageing and walking speed decline in active elders, were used to illustrate the usefulness of the proposed methods in the analysis of real experimental data.
△ Less
Submitted 24 August, 2021; v1 submitted 5 November, 2019;
originally announced November 2019.
