A novel approach for automatic visualization and activation detection of evoked potentials induced by epidural spinal cord stimulation in individuals with spinal cord injury

Abstract

Voluntary movements and the standing of spinal cord injured patients have been facilitated using lumbosacral spinal cord epidural stimulation (scES). Identifying the appropriate stimulation parameters (intensity, frequency and anode/cathode assignment) is an arduous task and requires extensive mapping of the spinal cord using evoked potentials. Effective visualization and detection of muscle evoked potentials induced by scES from the recorded electromyography (EMG) signals is critical to identify the optimal configurations and the effects of specific scES parameters on muscle activation. The purpose of this work was to develop a novel approach to automatically detect the occurrence of evoked potentials, quantify the attributes of the signal and visualize the effects across a high number of scES parameters. This new method is designed to automate the current process for performing this task, which has been accomplished manually by data analysts through observation of raw EMG signals, a process that is laborious and time-consuming as well as prone to human errors. The proposed method provides a fast and accurate five-step algorithms framework for activation detection and visualization of the results including: conversion of the EMG signal into its 2-D representation by overlaying the located signal building blocks; de-noising the 2-D image by applying the Generalized Gaussian Markov Random Field technique; detection of the occurrence of evoked potentials using a statistically optimal decision method through the comparison of the probability density functions of each segment to the background noise utilizing log-likelihood ratio; feature extraction of detected motor units such as peak-to-peak amplitude, latency, integrated EMG and Min-max time intervals; and finally visualization of the outputs as Colormap images. In comparing the automatic method vs. manual detection on 700 EMG signals from five individuals, the new approach decreased the processing time from several hours to less than 15 seconds for each set of data, and demonstrated an average accuracy of 98.28% based on the combined false positive and false negative error rates. The sensitivity of this method to the signal-to-noise ratio (SNR) was tested using simulated EMG signals and compared to two existing methods, where the novel technique showed much lower sensitivity to the SNR.

Conflict of interest statement

Competing Interests: The authors have declared that no competing interests exist.

Figures

Fig 1. Schematic representation of the epidural stimulation unit (16-electrode array, IPG unit and wireless programmer) and its connections to the EMG recording system.

(The Motion Lab EMG System and EMG electrodes illustrations are from Motion Lab System Inc. Manual).

Fig 2. Block diagram of the proposed framework for visualization and activation detection of evoked potentials induced by scES.

Fig 3. The steps for converting raw EMG signals into 2-D and 3-D images.

(A) Raw EMG signal, (B) Signal segmentation using stimulation time intervals, (C) Overlaying all the segments and building the 3-D graph where X-axis is the evoked potentials duration (ms), the Y-axis is the stimulation voltage (V), and the Z-axis is the amplitude of the signals (μV) and (D) Converting the 3-D graphs into 2-D images using Colormap.

Fig 4. EMG denoising using GGMRF.

(A) Applying GGMRF method to 2D image (B) An example of evoked potential before (black) and after (red) applying GGMRF method.

Fig 5. Calculation steps for activation detection algorithm.

(A) Sample evoked potential (one segment of the EMG signal), (B) Histogram of the sample evoked potential (black) and its estimated Gaussian distribution (red), (C) Comparing the Gaussian pdf of evoked potential signal (red) to pdf of background noise (black), (D) Plotting the calculated LLR for all segments of the EMG signal and detect the activation threshold (arrow).

Fig 6. Selected feature parameters for EMG activation signal.

(A) Visual inspection: Number of peaks of the evoked potential, Activation onset and Latency, (B) Computer-based feature extraction of peak-to-peak amplitude (Vpp), Activation latency, Time interval between minimum and maximum values (Tpp) and Integrated EMG (summation of absolute values of all gray areas).

Fig 7. Examples of converting 14 EMG signals into colormap images for voltage ramp-up and frequency ramp-up experiments.

(A) Raw EMG signals of 14 ploximal and distal left and right leg muscles during voltage ramp-up from 0.1 to 10 V. (B) Raw EMG signals of same muscles during frequency ramp-up from 2 to 60 Hz. (C) Colormap image shows the corresponding peak-to-peak amplitudes (μV) with respect to each muscle and stimulation voltage after stimulation threshold detection. The gray area is presenting the pre-threshold part of the experiment where no activation was induced. (D) Colormap image shows the corresponding integrated EMG values with respect to 14 muscles and stimulation frequencies.

Fig 8. Boxplot representation of performace measurements for comparing automated activation detection method with the ground truth.

(A) Accuracy, (B) Sensitivity, (C) Specificity, (D) Dice Similarity.

Fig 9. Examples of recorded EMG signals with different SNR levels and the performance comparison between three activation detection methods.

(A) High SNR signal from right MH; (B) Medium SNR signal from right GL and (C) Low SNR signal from L GL. Detected activation windows for AGLR + MMGRF, AGLR and TKEO from left to right are shown as continues and dashed black lines. De-noised signal is shown in light red and manually detected activation window as dashed red line.

Fig 10. Comparison of three activation detection methods on simulated EMG signals as a function of SNR(dB).