Neural dynamics in the middle temporal gyrus predicts behavioral performance in spatial navigation
Poster No:
1973
Submission Type:
Abstract Submission
Authors:
Xichunwang Wang1,2,3, Fengxiang Zhang1,2,3, Duxiao Guo1,2,3, Yi Pu4, Xiangzhen Kong1,2,3
Institutions:
1Department of Psychology and Behavioral Sciences, Zhejiang University, Hangzhou, China, 2Department of Psychiatry of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, China, 3The State Key Lab of Brain-Machine Intelligence, Zhejiang University, Hangzhou, China, 4Department of Neuroscience, Max Planck Institute for Empirical Aesthetics, Frankfurt am Main, Germany
First Author:
Xichunwang Wang
Department of Psychology and Behavioral Sciences, Zhejiang University|Department of Psychiatry of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine|The State Key Lab of Brain-Machine Intelligence, Zhejiang University
Hangzhou, China|Hangzhou, China|Hangzhou, China
Department of Psychology and Behavioral Sciences, Zhejiang University|Department of Psychiatry of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine|The State Key Lab of Brain-Machine Intelligence, Zhejiang University
Hangzhou, China|Hangzhou, China|Hangzhou, China
Co-Author(s):
Fengxiang Zhang
Department of Psychology and Behavioral Sciences, Zhejiang University|Department of Psychiatry of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine|The State Key Lab of Brain-Machine Intelligence, Zhejiang University
Hangzhou, China|Hangzhou, China|Hangzhou, China
Department of Psychology and Behavioral Sciences, Zhejiang University|Department of Psychiatry of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine|The State Key Lab of Brain-Machine Intelligence, Zhejiang University
Hangzhou, China|Hangzhou, China|Hangzhou, China
Duxiao Guo
Department of Psychology and Behavioral Sciences, Zhejiang University|Department of Psychiatry of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine|The State Key Lab of Brain-Machine Intelligence, Zhejiang University
Hangzhou, China|Hangzhou, China|Hangzhou, China
Department of Psychology and Behavioral Sciences, Zhejiang University|Department of Psychiatry of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine|The State Key Lab of Brain-Machine Intelligence, Zhejiang University
Hangzhou, China|Hangzhou, China|Hangzhou, China
Yi Pu
Department of Neuroscience, Max Planck Institute for Empirical Aesthetics
Frankfurt am Main, Germany
Department of Neuroscience, Max Planck Institute for Empirical Aesthetics
Frankfurt am Main, Germany
Xiangzhen Kong
Department of Psychology and Behavioral Sciences, Zhejiang University|Department of Psychiatry of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine|The State Key Lab of Brain-Machine Intelligence, Zhejiang University
Hangzhou, China|Hangzhou, China|Hangzhou, China
Department of Psychology and Behavioral Sciences, Zhejiang University|Department of Psychiatry of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine|The State Key Lab of Brain-Machine Intelligence, Zhejiang University
Hangzhou, China|Hangzhou, China|Hangzhou, China
Introduction:
Spatial navigation is a fundamental cognitive process that require the integration of complex sensory information and the orchestration of precise neural dynamics. Understanding the neural substrates is a compelling and multifaceted challenge. By utilizing Stereoelectroencephalography (SEEG) recordings in patients with epilepsy and machine learning techniques, our study aimed to dissect the neural dynamics engaged during spatial navigation.
Methods:
Twenty patients (7 females, Age: 26.3±7.9 years) were recruited, all being diagnosed with pharmacoresistant epilepsy and having SEEG monitoring. Written informed consent was obtained. SEEG recordings were conducted during each patient performing a 3D pointing task.
In the task, patients navigated in the mazes and were instructed to point to the starting spot at the end of each maze (Fig. 1A). Pointing error scores were computed as the absolute difference between the pointing directions and the actual direction. The task consisted of 24 trials.
In the present work, we focused on the SEEG data right at the starting spot and at the end of the mazes (7s for each). Data with a kurtosis>5 were labeled as epileptic seizures and thus excluded from the subsequent analyses. The AAL3 template were applied for assigning electrodes to the brain regions. After excluding regions with data of less than 5 patients, 25 regions remained (Fig. 1B). For each region, the SEEG signals were decomposed into five intrinsic mode functions (IMFs), based which 108 multivariate temporal neural dynamics features were extracted (Karabiber Cura et al., 2020).
Finally, we ran machine leaning analyses to predict the pointing error using features extracted from the SEEG data. A LassoCV regression approach was applied. Cross-validation was done with 75% of the samples as training data, and 25% as testing data.
In the task, patients navigated in the mazes and were instructed to point to the starting spot at the end of each maze (Fig. 1A). Pointing error scores were computed as the absolute difference between the pointing directions and the actual direction. The task consisted of 24 trials.
In the present work, we focused on the SEEG data right at the starting spot and at the end of the mazes (7s for each). Data with a kurtosis>5 were labeled as epileptic seizures and thus excluded from the subsequent analyses. The AAL3 template were applied for assigning electrodes to the brain regions. After excluding regions with data of less than 5 patients, 25 regions remained (Fig. 1B). For each region, the SEEG signals were decomposed into five intrinsic mode functions (IMFs), based which 108 multivariate temporal neural dynamics features were extracted (Karabiber Cura et al., 2020).
Finally, we ran machine leaning analyses to predict the pointing error using features extracted from the SEEG data. A LassoCV regression approach was applied. Cross-validation was done with 75% of the samples as training data, and 25% as testing data.
Results:
Our results showed that the prediction model for the middle temporal gyrus in the left hemisphere demonstrated the best prediction performance at both the starting period (r = 0.29, p = 1.3×10-6) and the end of the mazes (r = 0.33, p = 4.7×10-8). Moreover, we found that the model based on the changes of the neural dynamics features between the end and the starting period of the maze also showed significant prediction performance for the pointing errors (r = 0.29, p = 1.3×10-6). These results suggested that the middle temporal gyrus in the left hemisphere could play a critical role in spatial cognitive map learning.
The prediction models showed relatively high contribution of Hurst exponent of the original signal and the derived IMF3 mode (whose dominant frequencies are high-theta band) (Fig. 2C). The Hurst exponent of both the original signal and the IMF mode showed significant correlation with the pointing error (Fig. 2D-E).
The prediction models showed relatively high contribution of Hurst exponent of the original signal and the derived IMF3 mode (whose dominant frequencies are high-theta band) (Fig. 2C). The Hurst exponent of both the original signal and the IMF mode showed significant correlation with the pointing error (Fig. 2D-E).
Conclusions:
Our results showed that the neural dynamics features in the middle temporal gyrus predicts the behavioral performance in the 3D pointing task. Hurst exponent of the neural activity contributed the most in the prediction. These results suggest critical role of neural dynamics in the in spatial cognitive map learning. Further investigation is warranted into the underlying mechanisms.
Higher Cognitive Functions:
Space, Time and Number Coding
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
EEG/MEG Modeling and Analysis
Multivariate Approaches 1
Physiology, Metabolism and Neurotransmission :
Neurophysiology of Imaging Signals
Keywords:
Cognition
Computational Neuroscience
Data analysis
ELECTROPHYSIOLOGY
Machine Learning
Multivariate
1|2Indicates the priority used for review
Provide references using author date format
Karabiber Cura, O. (2020), 'Epileptic seizure classifications using empirical mode decomposition and its derivative', BioMedical Engineering OnLine, 19(1), 10.