Degradation-aware neural imputation: Advancing decoding stability in brain machine interfaces DOI Creative Commons

Yun-Ting Kuo,

Han-Lin Wang,

Bo-Wei Chen

et al.

APL Bioengineering, Journal Year: 2025, Volume and Issue: 9(2)

Published: April 16, 2025

Neural signal degradation poses a significant challenge in maintaining stable performance when decoding motor tasks using multiunit activity (MUA) and local field potential (LFP) signals the implantable brain machine interface (iBMI) applications. Effective methods for imputing degraded or missing are essential to restore neural integrity, thereby improving accuracy system robustness over long-term recordings with fluctuating quality. This study introduces confidence-weighted Bayesian linear regression (CW-BLR) approach impute affected by degradation, enhancing consistency of decoding. The CW-BLR was compared traditional methods—mean imputation (Mean-imp) Gaussian-mixture-model-based expectation–maximization (GMM-EM)—using kernel-sliced inverse (kSIR) decoder evaluate outcomes. Four Wistar rats were trained perform forelimb-reaching task while (MUA LFPs) recorded 27 days. imputed during days 8–27. Decoding evaluated kSIR Mean-imp GMM-EM. demonstrated superior effectively preserving both temporal spatial dependencies within signals. CW-BLR-imputed data significantly improved methods, showing consistently higher performance, particularly quality from period. offers robust effective framework iBMI applications, addressing challenges accurate prolonged recordings. By utilizing confidence-based metrics, surpasses providing across scenarios.

Language: Английский

Degradation-aware neural imputation: Advancing decoding stability in brain machine interfaces DOI Creative Commons

Yun-Ting Kuo,

Han-Lin Wang,

Bo-Wei Chen

et al.

APL Bioengineering, Journal Year: 2025, Volume and Issue: 9(2)

Published: April 16, 2025

Neural signal degradation poses a significant challenge in maintaining stable performance when decoding motor tasks using multiunit activity (MUA) and local field potential (LFP) signals the implantable brain machine interface (iBMI) applications. Effective methods for imputing degraded or missing are essential to restore neural integrity, thereby improving accuracy system robustness over long-term recordings with fluctuating quality. This study introduces confidence-weighted Bayesian linear regression (CW-BLR) approach impute affected by degradation, enhancing consistency of decoding. The CW-BLR was compared traditional methods—mean imputation (Mean-imp) Gaussian-mixture-model-based expectation–maximization (GMM-EM)—using kernel-sliced inverse (kSIR) decoder evaluate outcomes. Four Wistar rats were trained perform forelimb-reaching task while (MUA LFPs) recorded 27 days. imputed during days 8–27. Decoding evaluated kSIR Mean-imp GMM-EM. demonstrated superior effectively preserving both temporal spatial dependencies within signals. CW-BLR-imputed data significantly improved methods, showing consistently higher performance, particularly quality from period. offers robust effective framework iBMI applications, addressing challenges accurate prolonged recordings. By utilizing confidence-based metrics, surpasses providing across scenarios.

Language: Английский

Citations

0