Acute Deep Pain Detection Utilizing Heart Rhythm Analysis by Sparse Representation

It has been previously demonstrated that heart rhythm analysis using sparse representation can detect painful events in volunteers performing a cold pressor test. The current study aimed to detect deep pain events by objective analysis of the heart rhythm signal of subjects undergoing painful capsaicin injection by using 3 different analysis methods: the Fourier transform (FT), the wavelet transform (WT), and sparse representation (SR) of the signal by an over-complete dictionary of combined Fourier and wavelet bases using the orthogonal matching pursuit (OMP) algorithm. To compare their results and correlate serological levels of pain-related cytokines to the same painful event. Methods: In all, 16 adult volunteers participated in a study conducted in the pain clinic at the Veterans Affairs Center, San Diego. All subjects were injected with intramuscular capsaicin while being monitored; pain was rated using the Visual Analog Score (VAS). In all, 15 subjects who reported VAS scores of 7 and above were included in the study. Results: Compared with baseline, the WT analysis showed a significant coefficient density increase during the initial 60 seconds of the pain incline period ( P <.01) and over the overall 5-minute period ( P <.01). OMP analysis showed a significant increase of wavelet coefficients density during the initial 60 seconds ( P <.001) and at 5 minutes ( P <.001), with superior results compared with the WT analysis. Both methods showed significantly higher wavelet coefficients’ amplitudes during pain. Comparison of both methods showed that during baseline, there was a significant reduction of wavelet coefficients’ density using the OMP algorithm analysis ( P <0.01). Spectral analysis using the Fourier transform showed significant variations between the different subjects with various patterns during pain, thus it was unable to detect pain consistently. Serological pain-related cytokine levels were also not consistent with pain occurrence. Discussion: Heart rhythm analysis using SR and WT proved feasible for acute pain detection in healthy conscious subjects, with SR having superior results. Further studies in patients under general anesthesia are underway to evaluate this model for objective pain detection in that setting.