Scholarlink

In this study, the authors focused on analyzing fNIRS (functional near-infrared spectroscopy) signals with the use of a multimodal machine learning scheme to predict cognitive states and brain activation patterns. The data set consists of fNIRS recordings from 10 selected channels over a period of 1000 s, with a focus on the parameter HbO (oxygenated hemoglobin). Two machine learning models were used: a Deep Learning LSTM (Long Short-Term Memory) model and a KNN (K-Nearest Neighbors) algorithm. The results indicate that the LSTM model outperforms KNN in predicting brain activation events, achieving an accuracy of \(88.52 \%\) , compared to KNN’s accuracy of \(82.35 \%\) . The model is continually being refined to enhance its predictive accuracy and multimodality. This work highlights the potential of deep learning models in data analysis and underscores the importance of multimodal approaches for understanding complex brain functions.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multimodal Machine Learning Analysis of fNIRS Signals Using LSTM and KNN Models for Cognitive States and Brain Activation Patterns Prediction

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Adrian Luckiewicz,
  • Dariusz Mikolajewski,
  • Radoslaw Roszczyk,
  • Krzysztof Siwek,
  • Tomasz Kajdanowicz,
  • Mariusz Pelc,
  • Piotr Sterniuk,
  • Edward J. Gorzelanczyk,
  • Aleksandra Kawala-Sterniuk

摘要

In this study, the authors focused on analyzing fNIRS (functional near-infrared spectroscopy) signals with the use of a multimodal machine learning scheme to predict cognitive states and brain activation patterns. The data set consists of fNIRS recordings from 10 selected channels over a period of 1000 s, with a focus on the parameter HbO (oxygenated hemoglobin). Two machine learning models were used: a Deep Learning LSTM (Long Short-Term Memory) model and a KNN (K-Nearest Neighbors) algorithm. The results indicate that the LSTM model outperforms KNN in predicting brain activation events, achieving an accuracy of \(88.52 \%\) , compared to KNN’s accuracy of \(82.35 \%\) . The model is continually being refined to enhance its predictive accuracy and multimodality. This work highlights the potential of deep learning models in data analysis and underscores the importance of multimodal approaches for understanding complex brain functions.