Time-frequency analysis of biomagnetic signals measured by ..., Lecture notes of Engineering

Download Time-frequency analysis of biomagnetic signals measured by ... and more Lecture notes Engineering in PDF only on Docsity! Time-frequency analysis of biomagnetic signals measured by a 3-D vector gradiometer K. Kobayashi1, Y. Uchikawa1,2, I. Izumi2, M. Kawakatsu1, and M. Saito1,3 1Applied Superconductivity Research Laboratory, Tokyo Denki University, 2-1200 Chiba, Japan; 2Dept. of Electronic and Computer Engineering, Tokyo Denki University, 350-0394 Saitama, Japan; 3Dept. of Information and Communication Engineering, Tokyo Denki University, 101-8457 Tokyo, Japan 1 Introduction The SQUID magnetometer system with ultra-high magnetic sensitivity has made it possible to measure the very weak magnetic field generated by the activity of the brain and the heart. Biomagnetic measurement of the magnetic field perpendicular to the body surface is widely used. We had developed a 39-channel SQUID system consisting of thirteen 3-D second-order gradiometers (12 measurement positions (4×3) and a noise measurement) for vector measurement of the magnetoencephalogram (MEG) and the magnetocardiogram (MCG). This SQUID magnetometer can detect magnetic field components perpendicular to the body and tangential to the body simultaneously [1]. We carried out 3-D vector measurements of MEG with a mixed auditory evoked field (AEF) and somatosensory evoked field (SEF) overlapping in time, and MCG with normal subjects and an IRBBB (incomplete right bundle branch block) patient were also measured [2]. In order to discuss the frequency characteristics with time course, a short-term spectrum of biomagnetic signals were calculated by an autoregressive (AR) model, and the power spectrum was analyzed (time- frequency analysis). 2 Methods 2.1 Biomagnetic measurement We carried out 3-D measurement of MEG and MCG. Fig. 1 a) showed a 3-D second order gradiometer, which was orthogonal wound with NbTi-Cu wire on a rectangular solid 3x3x6 cm. 1. MEG measurement was the mixed AEF and SEF overlapping in time. AEF was elicited by 1 kHz tone bursts of 20 ms duration to the right ear. SEF was elicited by electric pulses of 0.2 ms duration with 5 mA to the median nerve of the right wrist. In the mixed AEF and SEF, stimulation to the median nerve was delivered 80 ms later than to the right ear. Fig. 1 b) showed coordinate system and measurement positions for MEG. There were forty-eight measurement positions (4 scanning, fixed circle in Fig. 1 b)) on the scalp and a band-pass filter was used in the range of 0.1 to 40 Hz. All magnetic data were averaged for 300 measurements at each position. 2. MCG measurements were carried out to normal subjects and an IRBBB patient, and ECG (lead II) was measured simultaneously. Fig. 1 c) showed coordinate system and measurement positions for MCG. There were forty-eight measurement positions (4 scanning, fixed circle in Fig. 1 c)) on the chest wall and a band-pass filter was used in the range of 0.1 to 80 Hz. All magnetic data were averaged for 50 measurements at each position. The reference signal for averaging was R-wave of ECG. 2.2 Time-Frequency analysis Time-frequency analysis of the biomagnetic data was calculated by the AR model [3],[4]. This AR model is effective in respect of frequency resolution in short-time spectrum. The analyzing time was 50ms and/or 100ms. The order of AR model was decided by final prediction error criterion (FPE) between 15 and 20. The short-time spectrum is calculated by equation (1). 2 1 2 )2exp(1 )( ∑ = ∆−+ ∆ = p k p k e tfkja tfB π σ (1) Where B(f): power spectrum of magnetic field, 2 eσ : variance of white noise, p ka : coefficient of the AR model, p: order of AR model, t∆ : sampling interval. 3 Results 3.1 MEG measurement Fig. 2 shows the waveform (Fig. 2 a)) of the mixed AEF and SEF overlapping in time measured at position F2 (Br component), the time-frequency analysis (Fig. 2 b)), and the intensity distributions of 0 50 100 150 200 250 300 0 10 20 30 40 50 0 0.5 1 Frequency[Hz] Po w er A B C D E F G H 1 2 3 4 5 6 A B C D E F G H 1 2 3 4 5 6 Tim e[m s] 1000 -150 [fT] Median Nerve Auditory Tim e[m s] a) c)b) e)d) f) h)g) Po w erTim e[m s] 1000 -150 [fT] Tim e[m s] 1000 -150 [fT] Tim e[m s] Figure 2: a) The waveform of the mixed AEF and SEF overlapping in time measured at position F2 (Br component). b) The time-frequency analysis. c)-d) The intensity distributions of power spectrum of 15 Hz at latency of 38 ms ( c):Br, d):Bθ, e):Bφ ). f)-h) The intensity distributions of power spectrum of 15 Hz at latency of 150 ms ( f):Br, g):Bθ, h):Bφ ). 2 . 9 cm 2.8cm 5.8 cm B r B φ B θ 1.4cm z B ƒ Ó B r B ƒ Æ x y A B C D E F G H 1 2 3 4 5 6 X Y Z 1 A B C D E G F H 2 3 4 5 6 4 cm 4 cm Figure 1: a) A 3-D second order gradiometer. b) Coordinate system and measurement positions for MEG. c) Coordinate system and measurement positions for MCG. power spectrum. Fig. 2 c)-e) and f)-h) show the intensity distributions of 15Hz power spectrum at latencies of 38 ms (analysis time: 38-88 ms) and 150 ms (analysis time: 150-200 ms) respectively. These figures show the intensity distribution of power spectrum obtained from each measurement position. In Fig. 2 c)-e), it can be seen the response at G4 by auditory stimulation before median nerve is stimulated. Bθ component of Fig. 2 d) shows the peak at G4. The peak of tangential components (Bθ, Bφ) corresponds to the source location [5]. Namely, the source can be estimated at G4. The position G4 corresponds to the location of auditory area at the left hemisphere [6]. In Fig. 2 f)-h), it can be seen the response at D3 by median nerve stimulation. The position D3 corresponds to the location of somatosensory area at the left hemisphere [7]. Fig. 3 also shows the waveform of the mixed AEF and SEF overlapping in time measured at position F2 (Br component), the time-frequency analysis, and the intensity distributions of distributions of 35 Hz and 15 Hz power spectrum at latency of 50 ms (analysis time: 50-100 ms) respectively. Bθ component of Fig. 3 d) shows the peak at D3. Therefore, the source of 35 Hz component can be estimated at D3. In Fig. 3 c)-e), it can be seen the response by median nerve stimulation, by referring 35 Hz component shown in Fig. 2 c)-e). Bθ component of Fig. 3 g) and Bφ component of Fig. 3 h) show the peak of 15 Hz component at G4 receptivity. In Fig. 3 f)-h), these can be seen the response by auditory stimulation, by referring 15 Hz component shown in Fig. 2 c)-e). By considering time-frequency analysis, the separation of the multiple sources is possible. These results show that time-frequency analysis is a useful method for estimating multiple sources. a) b) c)