skip to main content

Attention:

The NSF Public Access Repository (NSF-PAR) system and access will be unavailable from 11:00 PM ET on Thursday, May 23 until 2:00 AM ET on Friday, May 24 due to maintenance. We apologize for the inconvenience.


Title: Exploration of physiological sensors, features, and machine learning models for pain intensity estimation
In current clinical settings, typically pain is measured by a patient’s self-reported information. This subjective pain assessment results in suboptimal treatment plans, over-prescription of opioids, and drug-seeking behavior among patients. In the present study, we explored automatic objective pain intensity estimation machine learning models using inputs from physiological sensors. This study uses BioVid Heat Pain Dataset. We extracted features from Electrodermal Activity (EDA), Electrocardiogram (ECG), Electromyogram (EMG) signals collected from study participants subjected to heat pain. We built different machine learning models, including Linear Regression, Support Vector Regression (SVR), Neural Networks and Extreme Gradient Boosting for continuous value pain intensity estimation. Then we identified the physiological sensor, feature set and machine learning model that give the best predictive performance. We found that EDA is the most information-rich sensor for continuous pain intensity prediction. A set of only 3 features from EDA signals using SVR model gave an average performance of 0.93 mean absolute error (MAE) and 1.16 root means square error (RMSE) for the subject-independent model and of 0.92 MAE and 1.13 RMSE for subject-dependent. The MAE achieved with signal-feature-model combination is less than 1 unit on 0 to 4 continues pain scale, which is smaller than the MAE achieved by the methods reported in the literature. These results demonstrate that it is possible to estimate pain intensity of a patient using a computationally inexpensive machine learning model with 3 statistical features from EDA signal which can be collected from a wrist biosensor. This method paves a way to developing a wearable pain measurement device.  more » « less
Award ID(s):
1838796
NSF-PAR ID:
10289264
Author(s) / Creator(s):
; ;
Editor(s):
Le, Khanh N.Q.
Date Published:
Journal Name:
PLOS ONE
Volume:
16
Issue:
7
ISSN:
1932-6203
Page Range / eLocation ID:
e0254108
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. The ability to monitor mental effort during a task using a wearable sensor may improve productivity for both work and study. The use of the electrodermal activity (EDA) signal for tracking mental effort is an emerging area of research. Through analysis of over 92 h of data collected with the Empatica E4 on a single participant across 91 different activities, we report on the efficacy of using EDA features getting at signal intensity, signal dispersion, and peak intensity for prediction of the participant’s self-reported mental effort. We implemented the logistic regression algorithm as an interpretable machine learning approach and found that features related to signal intensity and peak intensity were most useful for the prediction of whether the participant was in a self-reported high mental effort state; increased signal and peak intensity were indicative of high mental effort. When cross-validated by activity moderate predictive efficacy was achieved (AUC = 0.63, F1 = 0.63, precision = 0.64, recall = 0.63) which was significantly stronger than using the model bias alone. Predicting mental effort using physiological data is a complex problem, and our findings add to research from other contexts showing that EDA may be a promising physiological indicator to use for sensor-based self-monitoring of mental effort throughout the day. Integration of other physiological features related to heart rate, respiration, and circulation may be necessary to obtain more accurate predictions. 
    more » « less
  2. Automatic pain intensity assessment from physiological signals has become an appealing approach, but it remains a largely unexplored research topic. Most studies have used machine learning approaches built on carefully designed features based on the domain knowledge available in the literature on the time series of physiological signals. However, a deep learning framework can automate the feature engineering step, enabling the model to directly deal with the raw input signals for real-time pain monitoring. We investigated a personalized Bidirectional Long short-term memory Recurrent Neural Networks (BiLSTM RNN), and an ensemble of BiLSTM RNN and Extreme Gradient Boosting Decision Trees (XGB) for four-category pain intensity classification. We recorded Electrodermal Activity (EDA) signals from 29 subjects during the cold pressor test. We decomposed EDA signals into tonic and phasic components and augmented them to original signals. The BiLSTM-XGB model outperformed the BiLSTM classification performance and achieved an average F1-score of 0.81 and an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.93 over four pain states: no pain, low pain, medium pain, and high pain. We also explored a concatenation of the deep-learning feature representations and a set of fourteen knowledge-based features extracted from EDA signals. The XGB model trained on this fused feature set showed better performance than when it was trained on component feature sets individually. This study showed that deep learning could let us go beyond expert knowledge and benefit from the generated deep representations of physiological signals for pain assessment. 
    more » « less
  3. Abstract

    In vivo fluorometers use chlorophyllafluorescence (Fchl) as a proxy to monitor phytoplankton biomass. However, the fluorescence yield ofFchlis affected by photoprotection processes triggered by increased irradiance (nonphotochemical quenching; NPQ), creating diurnal reductions inFchlthat may be mistaken for phytoplankton biomass reductions. Published correction methods are mostly designed for pelagic oceans and are ill suited for inland waters or for high‐frequency data collection. A machine learning‐based method was developed to correct vertical profiler data from an oligotrophic lake. NPQ was estimated as a percent reduction inFchlby comparing daytime values to mean, unquenched values from the previous night. A random forest regression was trained on sensor data collected coincident withFchl; including solar radiation, water temperature, depth, and dissolved oxygen saturation. The accuracy of the model was assessed using a grouped 10‐fold cross validation (mean absolute error [MAE]: 7.6%; root mean square error [RMSE]: 10.2%), which was then used to correctFchlprofiles. The model also predicted NPQ and corrected unseenFchlprofiles from a future period with excellent results (MAE: 9.0%; RMSE: 14.4%).Fchlprofiles were then correlated to laboratory results, allowing corrected profiles to be compared directly to collected samples. The correction reduced error (RMSE) due to NPQ from 0.67 μg L−1to 0.33 μg L−1when compared to uncorrectedFchldata. These results suggest that the use of machine learning models may be an effective way to correct for NPQ and may have universal applicability.

     
    more » « less
  4. This study aims to identify the most significant features in physiological signals representing a biphasic pattern in the menstrual cycle using circular statistics which is an appropriate analytic method for the interpretation of data with a periodic nature. The results can be used empirically to determine menstrual phases. A non-uniform pattern was observed in ovulating subjects, with a significant periodicity (p<0.05) in mean temperature, heart rate (HR), Inter-beat Interval (IBI), mean tonic component of Electrodermal Activity (EDA), and signal magnitude area (SMA) of the EDA phasic component in the frequency domain. In contrast, non-ovulating cycles displayed a more uniform distribution (p>0.05). There was a significant difference between ovulating and non-ovulating cycles (p<0.05) in temperature, IBI, and EDA but not in mean HR. Selected features were used in training an Autoregressive Integrated Moving Average (ARIMA) model, using data from at least one cycle of a subject, to predict the behavior of the signal in the last cycle. By iteratively retraining the algorithm on a per-day basis, the mean temperature, HR, IBI and EDA tonic values of the next day were predicted with root mean square error (RMSE) of 0.13 ± 0.07 (C°), 1.31 ± 0.34 (bpm), 0.016 ± 0.005 (s) and 0.17 ± 0.17 (μS), respectively.

     
    more » « less
  5. Although pain is widely recognized to be a multidimensional experience, it is typically measured by unidimensional patient self-reported visual analog scale (VAS). However, self-reported pain is subjective, difficult to interpret and sometimes impossible to obtain. Machine learning models have been developed to automatically recognize pain at both the frame level and sequence (or video) level. Many methods use or learn facial action units (AUs) defined by the Facial Action Coding System (FACS) for describing facial expressions with muscle movement. In this paper, we analyze the relationship between sequence-level multidimensional pain measurements and frame-level AUs and an AU derived pain-related measure, the Prkachin and Solomon Pain Intensity (PSPI). We study methods that learn sequence-level metrics from frame-level metrics. Specifically, we explore an extended multitask learning model to predict VAS from human-labeled AUs with the help of other sequence-level pain measurements during training. This model consists of two parts: a multitask learning neural network model to predict multidimensional pain scores, and an ensemble learning model to linearly combine the multidimensional pain scores to best approximate VAS. Starting from human-labeled AUs, the model achieves a mean absolute error (MAE) on VAS of 1.73. It outperforms provided human sequence-level estimates which have an MAE of 1.76. Combining our machine learning model with the human estimates gives the best performance of MAE on VAS of 1.48. 
    more » « less