April 1, 2015 / Vol. 40, No. 7 / OPTICS LETTERS
1193
Single-trial estimation of the cerebral metabolic rate of oxygen with imaging photoplethysmography and laser speckle contrast imaging Hongyang Lu,1,2 Yao Li,1,2 Hangdao Li,2 Lu Yuan,2 Qi Liu,2 Yu Sun,3 and Shanbao Tong1,2,* 1
Med-X Research Institute, Shanghai Jiao Tong University, Shanghai 200030, China
2
School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
3
Singapore Institute for Neurotechnology, National University of Singapore, Singapore 117456, Singapore *Corresponding author: [email protected] Received December 18, 2014; revised February 3, 2015; accepted February 16, 2015; posted February 17, 2015 (Doc. ID 230888); published March 17, 2015
Cortical cerebral metabolic rate of oxygen (CMRO2 ) could conventionally be measured by combining laser Doppler flowmetry and multispectral reflectance imaging across multiple trials of stimulation, which compromises the realtime capacity. Monitoring transient change of CMRO2 has been challenging. In this Letter, imaging photoplethysmography (iPPG) and laser speckle contrast imaging were combined into a multi-modal optical imaging system for single-trial estimation of CMRO2 . In a physiologically stable experiment, the iPPG-based method showed a less than 4% variance in comparison with the conventional method over 20 trials, and its temporal stability could be comparable to that by conventional method over 6 trials. While the oxygen supply was decreased deliberately, the new method was able to detect the transient changes of CMRO2 in real time, which could not be revealed by the conventional method. © 2015 Optical Society of America OCIS codes: (170.2655) Functional monitoring and imaging; (170.3880) Medical and biological imaging; (170.0110) Imaging systems. http://dx.doi.org/10.1364/OL.40.001193
As an important measure of oxygen metabolism, the cortical cerebral metabolic rate of oxygen (CMRO2 ) is of great importance for studying the brain functions. For example, CMRO2 could be used as a biomarker in animal experiments for neural activation [1], neurovascular coupling [2], and cerebral ischemia studies [3,4]. Clinically, alteration of CMRO2 has been found to be related to Huntington’s disease [5], Alzheimer’s disease [6], and natural aging [7]. So far, different techniques have been developed to quantify the relative CMRO2 (rCMRO2 ) changes in vivo, such as positron emission tomography (PET) [8], functional magnetic resonance imaging (fMRI) [9], nuclear magnetic resonance (NMR) [10,11], two-photon microscopy [12], and multispectral reflectance imaging (MSRI) [1,4]. However, the temporal resolutions of current PET and fMRI are insufficient to detect the transient changes of CMRO2 , such as the hemodynamic change several seconds immediately after the cerebral artery occlusion [13]. Therefore, PET and fMRI have been mostly used when the subjects are in physiologically stable condition. Two-photon microscopy has also been used for mapping the oxygen consumption with superior spatial resolution in a particular structure or neuronal circuit. However, it is unable to get the profile of global oxygen metabolism [12]. Therefore, in both clinical and experimental research, we need a full-field real-time in vivo technique to monitor rCMRO2 in response to transient events, e.g., acute ischemia or functional stimulation. In this Letter, inspired by multi-wavelength imaging photoplethysmography (imaging PPG, iPPG) [14,15], which offers contactless assessment of blood oxygen saturation (sO2 ) in real time [16], we propose a multi-modal optical imaging methodology for single-trial rCMRO2 estimation by simultaneous sO2 and cerebral blood flow (CBF) measurement. 0146-9592/15/071193-04$15.00/0
According to Jones et al. [17], rCMRO2 can be calculated from relative CBF (rCBF) and hemoglobin concentration as ΔHbR 1 rCMRO2 1 rCBF 1 γ R HbR0 −1 ΔHbT ; × 1 γT HbT0
(1)
where rCBF, which can be measured by laser speckle contrast imaging (LSCI) [18,19], is the fractional change of CBF compared with baseline, [HbT] is the concentration of total hemoglobin that consists of oxygenated ([HbO]) and deoxygenated ([HbR]) hemoglobin, and γ R and γ T are vascular weighting constants that are assumed to be 1 for simplification [1]. ΔHbR and ΔHbT at each pixel are solved using the modified Beer–Lambert Law, I0 lεHbR ΔHbR εHbO ΔHbO; ln It
(2)
where lnI0∕It is the attenuation of intensity over time t 0, and t, εHbR and εHbO are the molar extinction coefficients for HbR and HbO, respectively, and l is the differential path length factor. However, in practice, the periodic noise due to heart beat and respiration would introduce large vascular artifacts in [HbO] and [HbR] [20]. The traditional method for improving the signal-to-noise ratio (SNR) is to average optical intrinsic signals across multiple trials [21,22]. Such a grand-averaging approach hypothesizes a stable oxygen metabolism and ignores the variance of [HbO] and [HbR] over trials [23], which apparently also extends the duration of imaging. © 2015 Optical Society of America
1194
OPTICS LETTERS / Vol. 40, No. 7 / April 1, 2015
iPPG technology has been used for real-time monitoring of oxygen saturation, heart and respiration rate, and detection of peripheral vascular diseases [15]. PPG signals consist of two major components: (i) a pulsatile variation (i.e., AC component) that contains the fundamental frequency that is dependent on the heart rate, and (ii) a constant average (i.e., quasi-DC component) that relates to the tissues and the average blood volume. The amplitude of the AC signal is sensitive to sO2 variation because of the difference in the absorption of HbO and HbR at red and near infrared wavelengths. The realtime sO2 thus can be estimated using the AC and DC ratio at different wavelengths given a sufficient sampling rate [15,16]. Here we rewrite Eq. (1) into HbR0 ΔHbR 1 rCMRO2 1 rCBF HbR0 HbT0 ΔHbT −1 . × HbT0
(3)
Substituting sO2 HbO∕HbT, we get 1 − sO2 t ; 1 rCMRO2 1 rCBF · 1 − sO2 0
(4)
which enables the single-trial estimation of rCMRO2 from real-time iPPG signal. To test the iPPG-based single-trial estimation and compare it with conventional MSRI method, we implemented a multi-modal optical imaging system in a rat hind limb electrical stimulation experiment. The schematic of the imaging system is illustrated in Fig. 1(a). Each trial consisted of a CBF imaging session and a sO2 imaging session. For rCBF imaging, coherent light from an infrared laser diode (780 nm; 10 mW; L780P010, Thorlabs, Newton, New Jersey) powered by a drive module (LDC220C, Thorlabs) illuminated the cortex, and the reflected speckle images were captured by a 12-bit CMOS camera (2 × 2 binning; 640 × 640 pixels; 40 fps; acA2040180 km, Basler, Ahrensburg, Germany) through a micro lens system (Nikon AF-S VR Micro-Nikkor 105 mm f/2.8G IF-ED, Tochigi, Japan). At each second, random process estimator method [18] was applied to obtain the rCBF velocity information. During the sO2 imaging session, the cortex was illuminated by amber (590 nm; 100 mW; M590L3; Thorlabs) and red (635 nm; 400 mW; M625L3; Thorlabs) lightemitting diodes charged alternately at a frequency of 20 Hz [Fig. 1(a)]. The wavelengths were selected according to Fig. 2(a). 590 nm is close to an isosbestic point, thus [HbT] could be readily obtained. While at 635 nm, absorption coefficients of HbO and HbR are significantly different, so that [HbO] and [HbR] could be easily calculated. Images of intrinsic signals were acquired at 20 fps for each wavelength with the same camera of LSCI. The value of sO2 could be determined by the intensity of reflected intrinsic signals at different wavelengths. To improve SNR, we applied nonoverlapping sub-windows (e.g., 8 × 8 pixels in study) to segment the frames and defined the intrinsic signals as the averaged intensities within sub-windows for each frame by compromising
Fig. 1. (a) Overview of the multi-modal optical imaging system and the strobe signals for single-trial estimation of CMRO2 . LED, 590-nm and 635-nm light-emitting diodes; LD, 780-nm laser diode; Stim, trigger signal for hind limb electrical stimulation. (b)–(d) rCMRO2 change at 3 s post-stimulation by the conventional MSRI method and iPPG-based method overlaid on corresponding baseline CBF images obtained by LSCI on the left hemisphere. A threshold of 3.0% is applied to the rCMRO2 maps. In MSRI method, 1 (b) or 20 trials (c) are used. Two regions of interest are defined in (d) for further analysis. Asterisk, location of bregma; triangle, anterior direction. The scale bar represents 1 mm that applies to (b)–(d).
the spatial resolution. The AC and DC components could be extracted for sO2 estimation [24]. A fourth-order bandpass Butterworth filter was used to extract the AC component. Cutoff frequencies were 2 Hz and 10 Hz. Using Eq. (4), we therefore could obtain a single-trial estimation of rCMRO2 . The experimental protocols were approved by the Institutional Animal Care and Use Committee of Med-X Research Institute, Shanghai Jiao Tong University. A male Sprague-Dawley rat (200 g in weight, 8 weeks in age, Slac Laboratory Animal, Shanghai, China) was anesthetized with isoflurane (5.0% initial and 1.0% for maintenance) intubated with 20% O2 . Rectal temperature was maintained at 37.0 0.2°C with a temperature control module (FHC Inc., Bowdoin, Maine). Images were acquired through a thinned 3.0 mmhorizontal × 5.0 mmvertical cranial window centered at 0.5 mm posterior, 1.5 mm lateral to the bregma over the left hemisphere. A wall of reinforced glass ionomer cements (Dental Materials Factory of Shanghai Medical Instruments Co., Shanghai, China) was built around the window, filled with mineral oil (SigmaAldrich Co. LLC., St. Louis, Missouri) to keep the skull moist and reduce the glaring during the imaging. Electrical stimulation was delivered to the right hind paw for 2 s per session (5 Hz, 1 mA) with an inter-session interval of 43 s. All procedures were performed under standard sterile precautions.
April 1, 2015 / Vol. 40, No. 7 / OPTICS LETTERS
We first tested the rCMRO2 changes of the animal in response to hind limb electrical stimulation with both conventional MSRI method and iPPG-based estimation under a stable oxygen supply (∼20% oxygen level). The rCMRO2 profiles at 3 s post-stimulation calculated with MSRI method are illustrated in Figs. 1(b) and 1(c), using one and twenty trials, respectively. A threshold of 3.0% was applied. Figure 1(d) shows the result using the iPPG-based estimation, indicating similar spatial pattern as in Figs. 1(b) and 1(c). It could be noticed that rCMRO2 response map calculated by MSRI (r 20) was smaller than the extent measured by iPPG-based method, which might be due to the loss of inter-trial variance after averaging across multiple trials by MSRI. The time course of rCMRO2 changes within the box area [Fig. 1(d)] is plotted in Fig. 2(b). Note that rCMRO2 reliably showed a peak response at 3–4 s after stimulation, followed by an approximate 9-s recovery session in both methods [Fig. 2(b)]. Particularly, the iPPG-based estimation had similar performance to the grand-average MSRI over 20 trials with a 3.92% variance in the peak responses. Figure 2(c) shows the estimation of rCMRO2 by different methods at 3 s after stimulation along the red line crossing the bregma [Fig. 1(d)]. The performance of iPPGbased estimation was comparable to that of MRSI method using twenty trials. To compare the stability of both methods, we further calculated the standard deviations (SD) of rCMRO2 peak magnitudes by conventional MSRI method [SDMSRI r] of N continuous estimations over r trials in each estimation [N 8 and r 1; 2; …; 12 in this study, Fig. 2(d)] under a stable oxygen supply (∼20%). v u N u1 X 2 ¯ SDMSRI r t X r − Xr ; N i1 i
(5)
Fig. 2. (a) Molar absorption coefficients of HbO and HbR. Black lines indicate the wavelengths of LEDs used in the study. (b) Time courses of rCMRO2 in the box area depicted in Fig. 1(d) using both methods. Stim, hind paw stimulation starts at time 0 and lasts for 2 s. (c) The estimations of rCMRO2 by both methods at 3 s after stimulation onset along the red line crossing the bregma depicted in Fig. 1(d). (d) SD of rCMRO2 peak magnitude using MSRI across various numbers of trials, showing the SD by single-trial method (red) is comparable to MSRI method for more than 6 trials.
1195
Fig. 3. Changes of peak rCMRO2 responses to oxygen supply reduction in the box area depicted in Fig. 1(d) using both methods for one rat in a single test. Note that the transient hemodynamic change could not be revealed by the conventional MSRI method.
¯ where Xr is the mean value of N estimations of rCMRO2 peak magnitude. In each estimation, r trials were used. For comparison, we also plotted the SD in the iPPG-based method (i.e., SDiPPG ) from N estimations using the first trials in calculating each X i r. SDiPPG was 0.0041 in this study [see the red circle in Fig. 2(d)], which was comparable to SDMSRI r ≥ 6. To show the real-time rCMRO2 changes utilizing our iPPG-based single-trial estimation, we deliberately reduced the oxygen level every 30 min from 20% to 5% with a step of 5% in the rodent experiment. Figure 3 illustrates the changes of rCMRO2 peak response to oxygen supply variation in the area depicted in Fig. 1(d) for one rat in a single test. Apparently, the iPPG-based method could successfully track the real-time changes in oxygen supply, while conventional MSRI method only showed the average rCMRO2 level within the 30-min time window. Note that the rCMRO2 peak response rapidly dropped by ∼60% after the oxygen was reduced to 15%, then it remained at a relatively stable low rate while the oxygen was further reduced to 10% and 5%. Such a transient hemodynamic change could not be revealed by the conventional MSRI method. In conclusion, we proposed a method combining multiwavelength iPPG and LSCI for single-trial rCMRO2 estimation. The iPPG-based estimation allows a real-time estimation of rCMRO2 changes and could be utilized to track the transient physiological and pathological process. In a physiologically stable experiment, the iPPGbased estimation showed a less than 4% variance in rCMRO2 magnitude compared with the MSRI method using 20 trials. Its temporal stability is comparable to the MSRI method. For its simple setup and temporal stability, the new method could also be used in other stimulation paradigms with similar CMRO2 response magnitude, e.g., whisker stimulation. It should be noted that if we used two sets of cameras and micro-lens, the multi-wavelength iPPG and LSCI could be applied in separated light paths, which would further improve the real-time capability of the system. This research was supported by the National Natural Science Foundation of China (No. 61371018). Hongyang Lu was also supported by the China Scholarship Council (No. 201306230041). References 1. A. K. Dunn, A. Devor, H. Bolay, M. L. Andermann, M. A. Moskowitz, A. M. Dale, and D. A. Boas, Opt. Lett. 28, 28 (2003).
1196
OPTICS LETTERS / Vol. 40, No. 7 / April 1, 2015
2. A. Devor, I. Ulbert, A. K. Dunn, S. N. Narayanan, S. R. Jones, M. L. Andermann, D. A. Boas, and A. M. Dale, Proc. Natl. Acad. Sci. USA 102, 3822 (2005). 3. J. Mayhew, D. Johnston, J. Martindale, M. Jones, J. Berwick, and Y. Zheng, Neuroimage 13, 975 (2001). 4. P. B. Jones, H. K. Shin, D. A. Boas, B. T. Hyman, M. A. Moskowitz, C. Ayata, and A. K. Dunn, J. Biomed. Opt. 13, 044007 (2008). 5. K. Leenders, A. Palmer, N. Quinn, J. Clark, G. Firnau, E. Garnett, C. Nahmias, T. Jones, and C. Marsden, J. Neurol. Neurosurg. Psychiatry. 49, 853 (1986). 6. R. L. Buckner, A. Z. Snyder, B. J. Shannon, G. LaRossa, R. Sachs, A. F. Fotenos, Y. I. Sheline, W. E. Klunk, C. A. Mathis, and J. C. Morris, J. Neurosci. 25, 7709 (2005). 7. M. D’Esposito, L. Y. Deouell, and A. Gazzaley, Nat. Rev. Neurosci. 4, 863 (2003). 8. J. C. Baron and T. Jones, Neuroimage 61, 492 (2012). 9. Q. Shen, H. Ren, and T. Q. Duong, J. Magn. Reson. Imaging 27, 599 (2008). 10. X.-H. Zhu, Y. Zhang, R.-X. Tian, H. Lei, N. Zhang, X. Zhang, H. Merkle, K. Ugurbil, and W. Chen, Proc. Natl. Acad. Sci. USA 99, 13194 (2002). 11. J. Pekar, L. Ligeti, Z. Ruttner, R. Lyon, T. Sinnwell, P. van Gerderen, D. Fiat, C. Moonen, and A. McLaughlin, Magn. Reson. Med. 21, 313 (1991). 12. C. Mathiesen, K. Caesar, K. Thomsen, T. M. Hoogland, B. M. Witgen, A. Brazhe, and M. Lauritzen, J. Neurosci. 31, 18327 (2011).
13. H. Li, Y. Li, L. Yuan, C. Wu, H. Lu, and S. Tong, Rev. Sci. Instrum. 85, 094301 (2014). 14. Y. Sun, S. Hu, V. Azorin-Peris, S. Greenwald, J. Chambers, and Y. Zhu, J. Biomed. Opt. 16, 077010 (2011). 15. J. Allen, Physiol. Meas. 28, R1 (2007). 16. F. P. Wieringa, F. Mastik, and A. F. van der Steen, Ann. Biomed. Eng. 33, 1034 (2005). 17. M. Jones, J. Berwick, D. Johnston, and J. Mayhew, Neuroimage 13, 1002 (2001). 18. P. Miao, N. Li, N. V. Thakor, and S. Tong, Opt. Express 18, 218 (2010). 19. H. Li, Q. Liu, H. Lu, Y. Li, H. F. Zhang, and S. Tong, Opt. Express 22, 21079 (2014). 20. J. E. Mayhew, S. Askew, Y. Zheng, J. Porrill, G. W. Westby, P. Redgrave, D. M. Rector, and R. M. Harper, Neuroimage 4, 183 (1996). 21. M. Nemoto, Y. Hoshi, C. Sato, Y. Iguchi, I. Hashimoto, E. Kohno, T. Hirano, and S. Terakawa, Neuroimage 59, 3325 (2012). 22. C. C. Lay, M. F. Davis, C. H. Chen-Bee, and R. D. Frostig, J. Neurosci. 31, 11495 (2011). 23. T. J. Huppert, R. D. Hoge, S. G. Diamond, M. A. Franceschini, and D. A. Boas, Neuroimage 29, 368 (2006). 24. J. G. Webster, Design of Pulse Oximeters (CRC Press, 2002).
1193
Single-trial estimation of the cerebral metabolic rate of oxygen with imaging photoplethysmography and laser speckle contrast imaging Hongyang Lu,1,2 Yao Li,1,2 Hangdao Li,2 Lu Yuan,2 Qi Liu,2 Yu Sun,3 and Shanbao Tong1,2,* 1
Med-X Research Institute, Shanghai Jiao Tong University, Shanghai 200030, China
2
School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
3
Singapore Institute for Neurotechnology, National University of Singapore, Singapore 117456, Singapore *Corresponding author: [email protected] Received December 18, 2014; revised February 3, 2015; accepted February 16, 2015; posted February 17, 2015 (Doc. ID 230888); published March 17, 2015
Cortical cerebral metabolic rate of oxygen (CMRO2 ) could conventionally be measured by combining laser Doppler flowmetry and multispectral reflectance imaging across multiple trials of stimulation, which compromises the realtime capacity. Monitoring transient change of CMRO2 has been challenging. In this Letter, imaging photoplethysmography (iPPG) and laser speckle contrast imaging were combined into a multi-modal optical imaging system for single-trial estimation of CMRO2 . In a physiologically stable experiment, the iPPG-based method showed a less than 4% variance in comparison with the conventional method over 20 trials, and its temporal stability could be comparable to that by conventional method over 6 trials. While the oxygen supply was decreased deliberately, the new method was able to detect the transient changes of CMRO2 in real time, which could not be revealed by the conventional method. © 2015 Optical Society of America OCIS codes: (170.2655) Functional monitoring and imaging; (170.3880) Medical and biological imaging; (170.0110) Imaging systems. http://dx.doi.org/10.1364/OL.40.001193
As an important measure of oxygen metabolism, the cortical cerebral metabolic rate of oxygen (CMRO2 ) is of great importance for studying the brain functions. For example, CMRO2 could be used as a biomarker in animal experiments for neural activation [1], neurovascular coupling [2], and cerebral ischemia studies [3,4]. Clinically, alteration of CMRO2 has been found to be related to Huntington’s disease [5], Alzheimer’s disease [6], and natural aging [7]. So far, different techniques have been developed to quantify the relative CMRO2 (rCMRO2 ) changes in vivo, such as positron emission tomography (PET) [8], functional magnetic resonance imaging (fMRI) [9], nuclear magnetic resonance (NMR) [10,11], two-photon microscopy [12], and multispectral reflectance imaging (MSRI) [1,4]. However, the temporal resolutions of current PET and fMRI are insufficient to detect the transient changes of CMRO2 , such as the hemodynamic change several seconds immediately after the cerebral artery occlusion [13]. Therefore, PET and fMRI have been mostly used when the subjects are in physiologically stable condition. Two-photon microscopy has also been used for mapping the oxygen consumption with superior spatial resolution in a particular structure or neuronal circuit. However, it is unable to get the profile of global oxygen metabolism [12]. Therefore, in both clinical and experimental research, we need a full-field real-time in vivo technique to monitor rCMRO2 in response to transient events, e.g., acute ischemia or functional stimulation. In this Letter, inspired by multi-wavelength imaging photoplethysmography (imaging PPG, iPPG) [14,15], which offers contactless assessment of blood oxygen saturation (sO2 ) in real time [16], we propose a multi-modal optical imaging methodology for single-trial rCMRO2 estimation by simultaneous sO2 and cerebral blood flow (CBF) measurement. 0146-9592/15/071193-04$15.00/0
According to Jones et al. [17], rCMRO2 can be calculated from relative CBF (rCBF) and hemoglobin concentration as ΔHbR 1 rCMRO2 1 rCBF 1 γ R HbR0 −1 ΔHbT ; × 1 γT HbT0
(1)
where rCBF, which can be measured by laser speckle contrast imaging (LSCI) [18,19], is the fractional change of CBF compared with baseline, [HbT] is the concentration of total hemoglobin that consists of oxygenated ([HbO]) and deoxygenated ([HbR]) hemoglobin, and γ R and γ T are vascular weighting constants that are assumed to be 1 for simplification [1]. ΔHbR and ΔHbT at each pixel are solved using the modified Beer–Lambert Law, I0 lεHbR ΔHbR εHbO ΔHbO; ln It
(2)
where lnI0∕It is the attenuation of intensity over time t 0, and t, εHbR and εHbO are the molar extinction coefficients for HbR and HbO, respectively, and l is the differential path length factor. However, in practice, the periodic noise due to heart beat and respiration would introduce large vascular artifacts in [HbO] and [HbR] [20]. The traditional method for improving the signal-to-noise ratio (SNR) is to average optical intrinsic signals across multiple trials [21,22]. Such a grand-averaging approach hypothesizes a stable oxygen metabolism and ignores the variance of [HbO] and [HbR] over trials [23], which apparently also extends the duration of imaging. © 2015 Optical Society of America
1194
OPTICS LETTERS / Vol. 40, No. 7 / April 1, 2015
iPPG technology has been used for real-time monitoring of oxygen saturation, heart and respiration rate, and detection of peripheral vascular diseases [15]. PPG signals consist of two major components: (i) a pulsatile variation (i.e., AC component) that contains the fundamental frequency that is dependent on the heart rate, and (ii) a constant average (i.e., quasi-DC component) that relates to the tissues and the average blood volume. The amplitude of the AC signal is sensitive to sO2 variation because of the difference in the absorption of HbO and HbR at red and near infrared wavelengths. The realtime sO2 thus can be estimated using the AC and DC ratio at different wavelengths given a sufficient sampling rate [15,16]. Here we rewrite Eq. (1) into HbR0 ΔHbR 1 rCMRO2 1 rCBF HbR0 HbT0 ΔHbT −1 . × HbT0
(3)
Substituting sO2 HbO∕HbT, we get 1 − sO2 t ; 1 rCMRO2 1 rCBF · 1 − sO2 0
(4)
which enables the single-trial estimation of rCMRO2 from real-time iPPG signal. To test the iPPG-based single-trial estimation and compare it with conventional MSRI method, we implemented a multi-modal optical imaging system in a rat hind limb electrical stimulation experiment. The schematic of the imaging system is illustrated in Fig. 1(a). Each trial consisted of a CBF imaging session and a sO2 imaging session. For rCBF imaging, coherent light from an infrared laser diode (780 nm; 10 mW; L780P010, Thorlabs, Newton, New Jersey) powered by a drive module (LDC220C, Thorlabs) illuminated the cortex, and the reflected speckle images were captured by a 12-bit CMOS camera (2 × 2 binning; 640 × 640 pixels; 40 fps; acA2040180 km, Basler, Ahrensburg, Germany) through a micro lens system (Nikon AF-S VR Micro-Nikkor 105 mm f/2.8G IF-ED, Tochigi, Japan). At each second, random process estimator method [18] was applied to obtain the rCBF velocity information. During the sO2 imaging session, the cortex was illuminated by amber (590 nm; 100 mW; M590L3; Thorlabs) and red (635 nm; 400 mW; M625L3; Thorlabs) lightemitting diodes charged alternately at a frequency of 20 Hz [Fig. 1(a)]. The wavelengths were selected according to Fig. 2(a). 590 nm is close to an isosbestic point, thus [HbT] could be readily obtained. While at 635 nm, absorption coefficients of HbO and HbR are significantly different, so that [HbO] and [HbR] could be easily calculated. Images of intrinsic signals were acquired at 20 fps for each wavelength with the same camera of LSCI. The value of sO2 could be determined by the intensity of reflected intrinsic signals at different wavelengths. To improve SNR, we applied nonoverlapping sub-windows (e.g., 8 × 8 pixels in study) to segment the frames and defined the intrinsic signals as the averaged intensities within sub-windows for each frame by compromising
Fig. 1. (a) Overview of the multi-modal optical imaging system and the strobe signals for single-trial estimation of CMRO2 . LED, 590-nm and 635-nm light-emitting diodes; LD, 780-nm laser diode; Stim, trigger signal for hind limb electrical stimulation. (b)–(d) rCMRO2 change at 3 s post-stimulation by the conventional MSRI method and iPPG-based method overlaid on corresponding baseline CBF images obtained by LSCI on the left hemisphere. A threshold of 3.0% is applied to the rCMRO2 maps. In MSRI method, 1 (b) or 20 trials (c) are used. Two regions of interest are defined in (d) for further analysis. Asterisk, location of bregma; triangle, anterior direction. The scale bar represents 1 mm that applies to (b)–(d).
the spatial resolution. The AC and DC components could be extracted for sO2 estimation [24]. A fourth-order bandpass Butterworth filter was used to extract the AC component. Cutoff frequencies were 2 Hz and 10 Hz. Using Eq. (4), we therefore could obtain a single-trial estimation of rCMRO2 . The experimental protocols were approved by the Institutional Animal Care and Use Committee of Med-X Research Institute, Shanghai Jiao Tong University. A male Sprague-Dawley rat (200 g in weight, 8 weeks in age, Slac Laboratory Animal, Shanghai, China) was anesthetized with isoflurane (5.0% initial and 1.0% for maintenance) intubated with 20% O2 . Rectal temperature was maintained at 37.0 0.2°C with a temperature control module (FHC Inc., Bowdoin, Maine). Images were acquired through a thinned 3.0 mmhorizontal × 5.0 mmvertical cranial window centered at 0.5 mm posterior, 1.5 mm lateral to the bregma over the left hemisphere. A wall of reinforced glass ionomer cements (Dental Materials Factory of Shanghai Medical Instruments Co., Shanghai, China) was built around the window, filled with mineral oil (SigmaAldrich Co. LLC., St. Louis, Missouri) to keep the skull moist and reduce the glaring during the imaging. Electrical stimulation was delivered to the right hind paw for 2 s per session (5 Hz, 1 mA) with an inter-session interval of 43 s. All procedures were performed under standard sterile precautions.
April 1, 2015 / Vol. 40, No. 7 / OPTICS LETTERS
We first tested the rCMRO2 changes of the animal in response to hind limb electrical stimulation with both conventional MSRI method and iPPG-based estimation under a stable oxygen supply (∼20% oxygen level). The rCMRO2 profiles at 3 s post-stimulation calculated with MSRI method are illustrated in Figs. 1(b) and 1(c), using one and twenty trials, respectively. A threshold of 3.0% was applied. Figure 1(d) shows the result using the iPPG-based estimation, indicating similar spatial pattern as in Figs. 1(b) and 1(c). It could be noticed that rCMRO2 response map calculated by MSRI (r 20) was smaller than the extent measured by iPPG-based method, which might be due to the loss of inter-trial variance after averaging across multiple trials by MSRI. The time course of rCMRO2 changes within the box area [Fig. 1(d)] is plotted in Fig. 2(b). Note that rCMRO2 reliably showed a peak response at 3–4 s after stimulation, followed by an approximate 9-s recovery session in both methods [Fig. 2(b)]. Particularly, the iPPG-based estimation had similar performance to the grand-average MSRI over 20 trials with a 3.92% variance in the peak responses. Figure 2(c) shows the estimation of rCMRO2 by different methods at 3 s after stimulation along the red line crossing the bregma [Fig. 1(d)]. The performance of iPPGbased estimation was comparable to that of MRSI method using twenty trials. To compare the stability of both methods, we further calculated the standard deviations (SD) of rCMRO2 peak magnitudes by conventional MSRI method [SDMSRI r] of N continuous estimations over r trials in each estimation [N 8 and r 1; 2; …; 12 in this study, Fig. 2(d)] under a stable oxygen supply (∼20%). v u N u1 X 2 ¯ SDMSRI r t X r − Xr ; N i1 i
(5)
Fig. 2. (a) Molar absorption coefficients of HbO and HbR. Black lines indicate the wavelengths of LEDs used in the study. (b) Time courses of rCMRO2 in the box area depicted in Fig. 1(d) using both methods. Stim, hind paw stimulation starts at time 0 and lasts for 2 s. (c) The estimations of rCMRO2 by both methods at 3 s after stimulation onset along the red line crossing the bregma depicted in Fig. 1(d). (d) SD of rCMRO2 peak magnitude using MSRI across various numbers of trials, showing the SD by single-trial method (red) is comparable to MSRI method for more than 6 trials.
1195
Fig. 3. Changes of peak rCMRO2 responses to oxygen supply reduction in the box area depicted in Fig. 1(d) using both methods for one rat in a single test. Note that the transient hemodynamic change could not be revealed by the conventional MSRI method.
¯ where Xr is the mean value of N estimations of rCMRO2 peak magnitude. In each estimation, r trials were used. For comparison, we also plotted the SD in the iPPG-based method (i.e., SDiPPG ) from N estimations using the first trials in calculating each X i r. SDiPPG was 0.0041 in this study [see the red circle in Fig. 2(d)], which was comparable to SDMSRI r ≥ 6. To show the real-time rCMRO2 changes utilizing our iPPG-based single-trial estimation, we deliberately reduced the oxygen level every 30 min from 20% to 5% with a step of 5% in the rodent experiment. Figure 3 illustrates the changes of rCMRO2 peak response to oxygen supply variation in the area depicted in Fig. 1(d) for one rat in a single test. Apparently, the iPPG-based method could successfully track the real-time changes in oxygen supply, while conventional MSRI method only showed the average rCMRO2 level within the 30-min time window. Note that the rCMRO2 peak response rapidly dropped by ∼60% after the oxygen was reduced to 15%, then it remained at a relatively stable low rate while the oxygen was further reduced to 10% and 5%. Such a transient hemodynamic change could not be revealed by the conventional MSRI method. In conclusion, we proposed a method combining multiwavelength iPPG and LSCI for single-trial rCMRO2 estimation. The iPPG-based estimation allows a real-time estimation of rCMRO2 changes and could be utilized to track the transient physiological and pathological process. In a physiologically stable experiment, the iPPGbased estimation showed a less than 4% variance in rCMRO2 magnitude compared with the MSRI method using 20 trials. Its temporal stability is comparable to the MSRI method. For its simple setup and temporal stability, the new method could also be used in other stimulation paradigms with similar CMRO2 response magnitude, e.g., whisker stimulation. It should be noted that if we used two sets of cameras and micro-lens, the multi-wavelength iPPG and LSCI could be applied in separated light paths, which would further improve the real-time capability of the system. This research was supported by the National Natural Science Foundation of China (No. 61371018). Hongyang Lu was also supported by the China Scholarship Council (No. 201306230041). References 1. A. K. Dunn, A. Devor, H. Bolay, M. L. Andermann, M. A. Moskowitz, A. M. Dale, and D. A. Boas, Opt. Lett. 28, 28 (2003).
1196
OPTICS LETTERS / Vol. 40, No. 7 / April 1, 2015
2. A. Devor, I. Ulbert, A. K. Dunn, S. N. Narayanan, S. R. Jones, M. L. Andermann, D. A. Boas, and A. M. Dale, Proc. Natl. Acad. Sci. USA 102, 3822 (2005). 3. J. Mayhew, D. Johnston, J. Martindale, M. Jones, J. Berwick, and Y. Zheng, Neuroimage 13, 975 (2001). 4. P. B. Jones, H. K. Shin, D. A. Boas, B. T. Hyman, M. A. Moskowitz, C. Ayata, and A. K. Dunn, J. Biomed. Opt. 13, 044007 (2008). 5. K. Leenders, A. Palmer, N. Quinn, J. Clark, G. Firnau, E. Garnett, C. Nahmias, T. Jones, and C. Marsden, J. Neurol. Neurosurg. Psychiatry. 49, 853 (1986). 6. R. L. Buckner, A. Z. Snyder, B. J. Shannon, G. LaRossa, R. Sachs, A. F. Fotenos, Y. I. Sheline, W. E. Klunk, C. A. Mathis, and J. C. Morris, J. Neurosci. 25, 7709 (2005). 7. M. D’Esposito, L. Y. Deouell, and A. Gazzaley, Nat. Rev. Neurosci. 4, 863 (2003). 8. J. C. Baron and T. Jones, Neuroimage 61, 492 (2012). 9. Q. Shen, H. Ren, and T. Q. Duong, J. Magn. Reson. Imaging 27, 599 (2008). 10. X.-H. Zhu, Y. Zhang, R.-X. Tian, H. Lei, N. Zhang, X. Zhang, H. Merkle, K. Ugurbil, and W. Chen, Proc. Natl. Acad. Sci. USA 99, 13194 (2002). 11. J. Pekar, L. Ligeti, Z. Ruttner, R. Lyon, T. Sinnwell, P. van Gerderen, D. Fiat, C. Moonen, and A. McLaughlin, Magn. Reson. Med. 21, 313 (1991). 12. C. Mathiesen, K. Caesar, K. Thomsen, T. M. Hoogland, B. M. Witgen, A. Brazhe, and M. Lauritzen, J. Neurosci. 31, 18327 (2011).
13. H. Li, Y. Li, L. Yuan, C. Wu, H. Lu, and S. Tong, Rev. Sci. Instrum. 85, 094301 (2014). 14. Y. Sun, S. Hu, V. Azorin-Peris, S. Greenwald, J. Chambers, and Y. Zhu, J. Biomed. Opt. 16, 077010 (2011). 15. J. Allen, Physiol. Meas. 28, R1 (2007). 16. F. P. Wieringa, F. Mastik, and A. F. van der Steen, Ann. Biomed. Eng. 33, 1034 (2005). 17. M. Jones, J. Berwick, D. Johnston, and J. Mayhew, Neuroimage 13, 1002 (2001). 18. P. Miao, N. Li, N. V. Thakor, and S. Tong, Opt. Express 18, 218 (2010). 19. H. Li, Q. Liu, H. Lu, Y. Li, H. F. Zhang, and S. Tong, Opt. Express 22, 21079 (2014). 20. J. E. Mayhew, S. Askew, Y. Zheng, J. Porrill, G. W. Westby, P. Redgrave, D. M. Rector, and R. M. Harper, Neuroimage 4, 183 (1996). 21. M. Nemoto, Y. Hoshi, C. Sato, Y. Iguchi, I. Hashimoto, E. Kohno, T. Hirano, and S. Terakawa, Neuroimage 59, 3325 (2012). 22. C. C. Lay, M. F. Davis, C. H. Chen-Bee, and R. D. Frostig, J. Neurosci. 31, 11495 (2011). 23. T. J. Huppert, R. D. Hoge, S. G. Diamond, M. A. Franceschini, and D. A. Boas, Neuroimage 29, 368 (2006). 24. J. G. Webster, Design of Pulse Oximeters (CRC Press, 2002).
