Feasibility Study of BCI-Based Adaptive Sensing and Feedback Training for Chronic Pain Management

Project IDRRG1/16013
PartnerTan Tock Seng Hospital
FocusPrecision Rehabilitation
Clinical PI
Dr YANG Su-Yin
Clinical Lead, Principal Pyschologist, Pain Management Clinic
Tan Tock Seng Hospital
Technical PI
Dr ZHANG Hai Hong
Lab Head, Neural & Biomedical Technology
Institute for Infocomm Research (I2R)

 

The challenge

Chronic pain is defined as persistent pain lasting more than three months.

Chronic pain is a significant healthcare burden and is difficult and costly to treat. Research has shown that chronic pain not only affects the central nervous system (CNS) but can also arise from CNS dysfunction. Yet, current treatment approaches for chronic pain concentrate more on peripheral pain pathways and mechanisms rather than central ones. It has been hypothesised that directly manipulating brain regions could improve pain modulatory systems and, thereby, reverse the abnormalities in the CNS.

 

The proposed solution

In this project, we proposed that an innovative solution using a Brain Computer Interface (BCI)-based tool could be used to monitor and improve CNS conditions related to chronic pain. We designed a three-phase preliminary clinical trial to investigate the feasibility of this method in patients with chronic pain.

  • 22 participants (n = 11, healthy; n = 7, low back pain; n = 4, lower limb pain) who met study criteria were recruited via advertisements in Tan Tock Seng Hospital (TTSH) and via a weblink on the TTSH Pain Management Clinic website.
  • Participants were aged 22 to 63 years old (M = 39.4, SD = 10.7).

Phase 1

The primary objective in phase 1 was to collect, explore and identify discriminative and consistent patterns in chronic-pain related EEG. A wearable brain and peripheral sensing device was developed that integrated an EEG amplifier, a single-lead ECG sensor and a Galvanic Skin Response (GSR) sensor for simultaneous sensing and recording of the multiple signals.

A computer-based protocol procedure control and data logging device was then built. The protocol included a series of 15 physical movement tasks, and a series of 15 video watching tasks. For the former, the team set up an array of task stations. For the latter, the team engaged professionals for the making of daily activities videos related to chronic pain.

EEG brain signals were compared between healthy controls and pain patients at rest. Pain patients exhibited a higher level of activation in the mid and high beta bands at 13 electrodes (Fz, F4, F8, FC3, FCz, FT8, C3, C4, CP3, CP4, P4, PO1, PO2), localised to the frontal, central, right temporal, parietal and occipital regions. The findings are consistent with previous reports where chronic pain patients were found to demonstrate increased band power at rest.

Associations between EEG recordings during pain events and self-reported pain intensity among pain patients were also analysed. Positive correlations between pain intensity and high beta band (p=.17); and negative correlations between pain intensity and theta band (p=-.01) at electrode P3 were found. This suggests that pain may be associated with arousal of the neuronal networks in the left posterior regions linked to sensory and somatosensory processing.

Furthermore, in comparison with healthy controls, pain patients were found to report significantly higher pain scores (p=.001) and pain interference (p=.01). Pain patients also reported lower physical functioning (p=.02), with higher role limitation due to physical health status (p=.04). Contrary to expectations, there were no significant differences between pain patients and healthy controls on measures of fear avoidance of physical movements and depressive symptoms.

Phase 2

Based on the results obtained in Phase 1, Phase 2 focused on developing and proving the feasibility of a novel feedback training method based on pain EEG pattern detection and cognitive function computing.

Specifically, the technical team developed a new pain neuromatrix modelling and decoding mechanism called Constrained Discriminative Current Density Algorithm (cDCD). Current density, or more specifically, radial current density, represents the 2nd derivative of the electrical potential field on the scalp surface. It measures the physical quantity of current flowing along the radial direction at a particular coordinate. In the detection accuracy test, the new cDCD mechanism was able to produce significantly higher accuracy than the state-of-the-art EEG temporal-spectral method.

The team used this new neurocomputing method for pain neuromatrix activity detection, in combination with I2R’s proprietary EEG-based attention detection algorithm, to create a BCI based pain neuromodulatory therapy in the form of a fun game that enabled patients to practise an individualised pain management strategy.

The participants were instructed to focus and sustain their concentration to improve the attention score. When the system detects that the participant’s pain neuromatrix is activated, the wellness score will go down and the participant is advised to practise pain-management strategies taught by their clinician or therapist. These strategies may include cognitive behaviour therapy, breathing techniques, and other tailored management strategies. This visual feedback mechanism provides the subject with a more direct and informative response, one that informs them whether such pain-management strategies work and allow them to translate or use these strategies more effectively in their day-to-day activities.

In the Phase 2 study:

  • 5 pain participants (n = 3, low back pain; n = 2, low limb pain) were selected to undergo 18 sessions of BCI training over a period of 6-11 weeks.
  • Before the start of the BCI training in session 1 and at the end of the BCI training in session 18, participants were required to complete a set of questionnaires of pain intensity, pain-interference, fear-avoidance, disability, pain catastrophising and depression, and a 6-minute shuttle walk test.
  • The participants were all able to play the BCI feedback game, and completed the sessions in compliance with the training programme.

Phase 3

In phase 3, the follow-up phase, a total of two visits to the study site were conducted. Both sessions involve a repeat of the procedures completed in Phase 1.

From analysis of the averaged user ratings during BCI training, it was found that there were general trends of lowered pain rating score in three out of the five subjects. There was no significant trend in distress rating scores. One subject exhibited large fluctuations in rating scores during game play, which calls for further analysis.

Wilcoxon signed rank test indicates shows a decrease in pain rating score from 3.69 (2.25) to 2.3 (2.07) with a p-value of .042. The interesting findings from the five subjects require a much larger scale study to confirm.