Classification of fNIRS Finger Tapping Data With Multi-Labeling and Deep Learning

Sommer, Natalie; Kakillioglu, Burak; Grant, Trevor; Velipasalar, Senem; Hirshfield, Leanne M.

doi:10.1109/jsen.2021.3115405

Cited by 16 publications

(16 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Using these 2D images as input into a CNN, they achieved an 8% increase in accuracy on an n-back (a classic experiment for measuring mental workload) classification task, compared to a DNN model with flattened input layer. Sommer et al arranged HD-fNIRS measurements in two areas of interest corresponding to the left and right primary motor cortices, and within these regions according to the SDS [29]. Rolling window samples of these images were used to train a CNN-LSTM model to predict finger-opposition laterality and frequency, achieving average accuracy of 0.81 over eleven subjects.…”

Section: Spatial-temporal Feature Extractionmentioning

confidence: 99%

System Derived Spatial-Temporal CNN for High-Density fNIRS BCI

Dale

O’Sullivan

Howard

et al. 2023

IEEE Open J. Eng. Med. Biol.

View full text Add to dashboard Cite

An intuitive and generalisable approach to spatial-temporal feature extraction for high-density (HD) functional Near-Infrared Spectroscopy (fNIRS) brain-computer interface (BCI) is proposed, demonstrated here using Frequency-Domain (FD) fNIRS for motor-task classification. Enabled by the HD probe design, layered topographical maps of Oxy/deOxy Haemoglobin changes are used to train a 3D convolutional neural network (CNN), enabling simultaneous extraction of spatial and temporal features. The proposed spatial-temporal CNN is shown to effectively exploit the spatial relationships in HD fNIRS measurements to improve the classification of the functional haemodynamic response, achieving an average F1 score of 0.69 across seven subjects in a mixed subjects training scheme, and improving subject-independent classification as compared to a standard temporal CNN.INDEX TERMS fNIRS, brain-computer interface, neural network, machine learning, CNN IMPACT STATEMENT The unique and often overlooked interdependence between fNIRS probe design and neural network architecture is demonstrated. A novel and effective BCI model is presented, which can be generalised to multi-wavelength, broadband, and hybrid fNIRS data.

show abstract

Section: Spatial-temporal Feature Extractionmentioning

confidence: 99%

System Derived Spatial-Temporal CNN for High-Density fNIRS BCI

Dale

O’Sullivan

Howard

et al. 2023

IEEE Open J. Eng. Med. Biol.

View full text Add to dashboard Cite

show abstract

“…Compared to other neuroimaging techniques such as electroencephalography (EEG) and functional magnetic resonance imaging (fMRI), fNIRS provides better spatial and temporal resolutions, respectively [ 3 , 4 ]. Thus, a wide range of studies in different cognitive tasks and clinical settings have employed fNIRS, e.g., [ 5 , 6 , 7 , 8 ].…”

Section: Introductionmentioning

confidence: 99%

Estimation of Respiratory Rate from Functional Near-Infrared Spectroscopy (fNIRS): A New Perspective on Respiratory Interference

Hakimi

Shahbakhti

Sappia³

et al. 2022

Biosensors

View full text Add to dashboard Cite

Objective: Respiration is recognized as a systematic physiological interference in functional near-infrared spectroscopy (fNIRS). However, it remains unanswered as to whether it is possible to estimate the respiratory rate (RR) from such interference. Undoubtedly, RR estimation from fNIRS can provide complementary information that can be used alongside the cerebral activity analysis, e.g., sport studies. Thus, the objective of this paper is to propose a method for RR estimation from fNIRS. Our primary presumption is that changes in the baseline wander of oxygenated hemoglobin concentration (O2Hb) signal are related to RR. Methods: fNIRS and respiratory signals were concurrently collected from subjects during controlled breathing tasks at a constant rate from 0.1 Hz to 0.4 Hz. Firstly, the signal quality index algorithm is employed to select the best O2Hb signal, and then a band-pass filter with cut-off frequencies from 0.05 to 2 Hz is used to remove very low- and high-frequency artifacts. Secondly, troughs of the filtered O2Hb signal are localized for synthesizing the baseline wander (S1) using cubic spline interpolation. Finally, the fast Fourier transform of the S1 signal is computed, and its dominant frequency is considered as RR. In this paper, two different datasets were employed, where the first one was used for the parameter adjustment of the proposed method, and the second one was solely used for testing. Results: The low mean absolute error between the reference and estimated RRs for the first and second datasets (2.6 and 1.3 breaths per minute, respectively) indicates the feasibility of the proposed method for RR estimation from fNIRS. Significance: This paper provides a novel view on the respiration interference as a source of complementary information in fNIRS.

show abstract

“…We also employ more appropriate evaluation strategies for the second barrier. K-fold crossvalidation (KFold-CV) is used to evaluate test results, whereas leave-one-subject-out cross-validation (LOSO-CV) is used to assess model generalization and individual differences [9], [12], [13]. For LOSO-CV, one subject's data serves as a test set, and the rest serves as a training set.…”

Section: Introductionmentioning

confidence: 99%

“…State-of-the-art visual models based on Transformers [15], [16], [20] and multi-layer perceptrons (MLPs) [21], [22] are introduced into comparison experiments. Unlike recent studies [6], [7], [11], [12], our work is more devoted to demonstrating the generality of the proposed framework. The rest of this paper is organized as follows.…”

Section: Introductionmentioning

confidence: 99%

A General and Scalable Vision Framework for Functional Near-Infrared Spectroscopy Classification

Wang

Zhang

Xia

et al. 2022

IEEE Trans. Neural Syst. Rehabil. Eng.

View full text Add to dashboard Cite

Functional near-infrared spectroscopy (fNIRS), a non-invasive optical technique, is widely used to monitor brain activities for disease diagnosis and braincomputer interfaces (BCIs). Deep learning-based fNIRS classification faces three major barriers: limited datasets, confusing evaluation criteria, and domain barriers. We apply more appropriate evaluation methods to three openaccess datasets to solve the first two barriers. For domain barriers, we propose a general and scalable vision fNIRS framework that converts multi-channel fNIRS signals into multi-channel virtual images using the Gramian angular difference field (GADF). We use the framework to train state-of-the-art visual models from computer vision (CV) within a few minutes, and the classification performance is competitive with the latest fNIRS models. In crossvalidation experiments, visual models achieve the highest average classification results of 78.68% and 73.92% on mental arithmetic and word generation tasks, respectively. Although visual models are slightly lower than the fNIRS models on unilateral finger-and foot-tapping tasks, the F1score and kappa coefficient indicate that these differences are insignificant in subject-independent experiments. Furthermore, we study fNIRS signal representations and the classification performance of sequence-to-image methods. We hope to introduce rich achievements from the CV domain to improve fNIRS classification research.

show abstract

Classification of fNIRS Finger Tapping Data With Multi-Labeling and Deep Learning

Cited by 16 publications

References 53 publications

System Derived Spatial-Temporal CNN for High-Density fNIRS BCI

System Derived Spatial-Temporal CNN for High-Density fNIRS BCI

Estimation of Respiratory Rate from Functional Near-Infrared Spectroscopy (fNIRS): A New Perspective on Respiratory Interference

A General and Scalable Vision Framework for Functional Near-Infrared Spectroscopy Classification

Contact Info

Product

Resources

About