Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bayesian Time-Series Classifier for Decoding Simple Visual Stimuli from Intracranial Neural Activity

Navid Ziaei, Reza Saadatifard, Ali Yousefi, Behzad Nazari, Sydney S. Cash, Angelique C. Paulk

TL;DR

The paper addresses decoding simple visual stimuli from intracranial neural activity under limited data conditions. It introduces the Bayesian Time-Series Classifier (BTsC), which uses time-series multivariate normal modeling of ERP and high gamma power features, along with channel subset selection and two aggregation strategies (likelihood and voting) to achieve accurate, interpretable decoding. Key findings show an average accuracy of $75.55\%$ across participants, outperforming seven strong baselines by about $3\%$, and identifying optimal time windows (typically < $0.8$ s) and informative brain regions. This approach offers a robust, interpretable tool for studying distributed neural encoding and has potential applications in brain-computer interfaces and neural prosthetics, with future work aimed at broader cognitive tasks and relaxing independence assumptions.

Abstract

Understanding how external stimuli are encoded in distributed neural activity is of significant interest in clinical and basic neuroscience. To address this need, it is essential to develop analytical tools capable of handling limited data and the intrinsic stochasticity present in neural data. In this study, we propose a straightforward Bayesian time series classifier (BTsC) model that tackles these challenges whilst maintaining a high level of interpretability. We demonstrate the classification capabilities of this approach by utilizing neural data to decode colors in a visual task. The model exhibits consistent and reliable average performance of 75.55% on 4 patients' dataset, improving upon state-of-the-art machine learning techniques by about 3.0 percent. In addition to its high classification accuracy, the proposed BTsC model provides interpretable results, making the technique a valuable tool to study neural activity in various tasks and categories. The proposed solution can be applied to neural data recorded in various tasks, where there is a need for interpretable results and accurate classification accuracy.

Bayesian Time-Series Classifier for Decoding Simple Visual Stimuli from Intracranial Neural Activity

TL;DR

The paper addresses decoding simple visual stimuli from intracranial neural activity under limited data conditions. It introduces the Bayesian Time-Series Classifier (BTsC), which uses time-series multivariate normal modeling of ERP and high gamma power features, along with channel subset selection and two aggregation strategies (likelihood and voting) to achieve accurate, interpretable decoding. Key findings show an average accuracy of across participants, outperforming seven strong baselines by about , and identifying optimal time windows (typically < s) and informative brain regions. This approach offers a robust, interpretable tool for studying distributed neural encoding and has potential applications in brain-computer interfaces and neural prosthetics, with future work aimed at broader cognitive tasks and relaxing independence assumptions.

Abstract

Understanding how external stimuli are encoded in distributed neural activity is of significant interest in clinical and basic neuroscience. To address this need, it is essential to develop analytical tools capable of handling limited data and the intrinsic stochasticity present in neural data. In this study, we propose a straightforward Bayesian time series classifier (BTsC) model that tackles these challenges whilst maintaining a high level of interpretability. We demonstrate the classification capabilities of this approach by utilizing neural data to decode colors in a visual task. The model exhibits consistent and reliable average performance of 75.55% on 4 patients' dataset, improving upon state-of-the-art machine learning techniques by about 3.0 percent. In addition to its high classification accuracy, the proposed BTsC model provides interpretable results, making the technique a valuable tool to study neural activity in various tasks and categories. The proposed solution can be applied to neural data recorded in various tasks, where there is a need for interpretable results and accurate classification accuracy.
Paper Structure (14 sections, 7 equations, 2 figures, 1 table)

This paper contains 14 sections, 7 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Material and Methods
  3. Dataset and Behavioral Task
  4. Data Preprocessing: Intracranial LFP Data
  5. Extracting Neural Features for Decoding
  6. Bayesian Time-series Classifier (BTsC)
  7. Multi-channel Classifier.
  8. Multi-channel Likelihood Method.
  9. Maximum Voting Method.
  10. Classifiers Subset Selection.
  11. Results
  12. Machine Learning Decoding
  13. Discussion
  14. Conclusion

Figures (2)

  • Figure 1: Data Overview: (A) shows the electrode placement in four participants, (B) illustrates the Flicker Task paradigm performed in 100 trials, (C) outlines the steps for feature extraction, (D) displays the preprocessed single trial signal (top), ERP features (middle), and HGP features (bottom) extracted from the LTP02-LTP03 electrode for patient P04, and (E) presents a scatter plot of ERP (top) and HGP (bottom) features at times t1 and t2 for all trials recorded from the same electrode, indicating a Gaussian distribution.
  • Figure 2: Results: (A) illustrates the performance of individual classifiers. (B) shows accuracy evolution during channel combination steps for participant P05. (C) shows the comparison between the BTsC performance using ERP, HGP, and both ERP and HGP features to assess the impact of neural features. (D) displays the accuracy of the model at different time points for participants P01 to P05.