Application of functional data analysis in classification and clustering of functional near-infrared spectroscopy signal in response to noxious stimuli

Abstract

Abstract. We introduce the application of functional data analysis (fDA) on functional near-infrared spectroscopy (fNIRS) signals for the development of an accurate and clinically practical assessment method of pain perception. We used the cold pressor test to induce different levels of pain in healthy subjects while the fNIRS signal was recorded from the frontal regions of the brain. We applied fDA on the collected fNIRS data to convert discrete samples into continuous curves. This method enabled us to represent the curves as a linear combination of basis functions. We utilized bases coefficients as features that represent the shape of the signals (as opposed to extracting defined features from signal) and used them to train a support vector machine to classify the signals based on the level of induced pain. We achieved 94% of accuracy to classify low-pain and high-pain signals. Moreover applying hierarchical clustering on the coefficients, we found three clusters in the data which represented low-pain (one cluster) and high-pain groups (two clusters) with an accuracy of 91.2%. The center of these clusters can represent the prototype fNIRS response of that pain level. © The Authors. Published by SPIE under a Creative Commons Attribution 3.0 Unported License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI.

Multi-task multiple kernel machines for personalized pain recognition from functional near-infrared spectroscopy brain signals

“…From our experimental recordings, we extracted windows of duration 20 seconds starting with the onset of the 7/10 electrical stimulus as well as randomly sampled windows from the baseline recording . This choice of window size was done by visual inspection of the responses (see Fig.2), and is in the same order of magnitude as those of other studies on pain detection from fNIRS signals [10]. From these windows, we extracted 3 feature sets: (a) b-spline coefficients, (b) statistical features, and (c) b-spline coefficients and statistical features combined, that is, all features.…”

Toward a functional near-infrared spectroscopy-based monitoring of pain assessment for nonverbal patients

“…Gram et al 20 used EEG to investigate morphine-and placebo-administered subjects after receiving stimulation using the cold pressor test, the authors used SVM to classify responders with an accuracy of 72%. In an fNIRS study, Pourshoghi et al 21 used a SVM classifier to classify low pain and high pain from healthy subjects after a cold pressor test with 94% accuracy. These results show that pain recognition and classification is plausible using neuroimaging methods, also the results from these studies advocate for the use of machine learning techniques to predict human pain.…”

“…18,21 Its basic idea is to find a linear separating boundary with the maximal margin, and new data points are then classified according to which side of the decision boundary they fall. 42 However, data in real-world scenarios are not easy to separate by a linear decision boundary; in these cases, SVM solves this problem by kernel functions to map the observations into higher-dimensional space in which the data can be separable.…”

Pain Detection with fNIRS-Measured Brain Signals: A Personalized Machine Learning Approach Using the Wavelet Transform and Bayesian Hierarchical Modeling with Dirichlet Process Priors

“…After processing the fNIRS signals as described in Sec.II-B, windows of duration 20 seconds were extracted from the HbO signals. The choice of signal modality (that is, HbO) and window size was informed by previous work done on pain detection from fNIRS [14], [33]. From these windows, we extracted the D = 80 features described in Sec.II-C from the prefrontal fNIRS channels depicted in Fig.1.…”