Head Pulsation Signal Analysis for 3-Axis Head-Worn Accelerometers

Olli Lahdenoja1, Tero Hurnanen1, Juho Koskinen1, Matti Kaisti1, Kim Munck2, Samuel Emil Schmidt2, Tero Koivisto1, Mikko Pänkäälä1
1University of Turku, 2Aalborg University


AIMS: The purpose of this pre-clinical study was to validate the usability of non-invasive head-worn 3-axis accelerometers for head pulsation signal acquisition. A future application is to separate between heathy and diseased subjects (e.g. with traumatic brain injury). METHODS: A 3-axis accelerometer was attached with a double-sided tape to three locations of human head (left temple, forehead and right temple). A head pulsation signal of length 5 min was recorded from each of five male volunteers. All subjects were healthy and in supine position. Simultaneous and synchronized 3-lead chest electrocardiogram (ECG) was recorded using the same sampling rate (1 kHz). A 3-min duration uniform artifact free segment was extracted from each subject manually. Previously developed single-axis autocorrelation (AC) method was applied to estimate the HR (heart rate) and HRV (heart rate variability) from each subject, sensor and axis (the AC window size was adjusted for each subject). A reference ECG was used to obtain the ground truth for HR and HRV by extracting the R-peaks of the signal. RESULTS: As there were five subjects and three accelerometers each, in total there were 15 recordings. For each recording there were three acceleration axes (X, Y Z). The HR and HRV were calculated for each recording and axis using single-axis AC algorithm. The mean HR and HRV for each recording is reported. For all recordings (15 of 15) the deviation of the mean HR against ECG was below or equal to 1 bpm. For 11 best recordings the deviation in mean HRV against ECG was below or equal to 10 ms. CONCLUSIONS: The results warrant further investigations in utilizing head pulsation signal derived from 3-axis accelerometers.