As a fast moving and strong scattering sample

As a fast-moving and strong scattering sample, the fingertip of a test subject was used in this study. Non-invasive blood testing has been reported for several decades; however, no actual device to obtain blood samples non-invasively has been proposed yet. One of the main reasons that no such device has been developed is that it is difficult to obtain reliable blood spectra from in vivo measurements, because the thickness and content of skin vary appreciably from person to person. Certainly diffuse reflectance spectroscopy can be used to achieve sufficient light intensity for the measurement, but diffuse reflections are easily affected by the skin conditions. Hence, we decided to use transmitted light spectroscopy to reduce the measurement uncertainty. The problem of signal weakness was solved by employing McFT spectroscopy with a high optical throughput. We measured blood fat as a specific example of non-invasive blood testing. Two blood components change shortly after a meal: neutral fat and glucose [17]. Although glucose can be monitored to obtain important medical information, the key technology for extracting glucose includes the use of an analysis algorithm, such as multi-component analysis software, which is beyond the scope of the measuring equipment focused on in this paper [7]. Thus, to evaluate the proposed device, we measured neutral fat. Robust neutral fat detection with our developed device would be useful for the measurement of glucose. We considered the carboxypeptidase a peak corresponding to a wavelength of 1200nm to detect neutral fat. This wavelength is known to occur in molecular oscillations due to the second overtones of C–H vibrations, which occur predominantly in neutral fat and protein in the human body [16]. Unfortunately, most fat and protein absorption signals obtained from fingertip measurements are due to skin components, subcutaneous fat, and skin cell protein. However, real-time spectroscopy enables blood components to be distinguished from subcutaneous fat and skin cell protein by detecting the fingertip pulsations [8,9]. This kind of plethysmography is known to be useful for hemoglobin detection [13], because human blood is composed mostly of hemoglobin and water. Since lipids and glucose each only constitute 0.1wt% of blood, optical non-invasive detection of these substances has not been realized directly.
Results and discussion The in-phase and anti-phase interferograms were obtained using the upper and lower halves of the area sensor, respectively, and are shown in Fig. 2. These images were obtained after correcting the distortions using the geometrical warping transform. After the distortions were corrected, the in-phase and anti-phase interferograms, which each corresponded to 190 lines, were integrated into a single interferogram by subtracting the anti-phase interferogram from the in-phase interferogram to increase the signal-to-noise ratio. This procedure was performed at a sufficiently high speed so that it was completed within the frame rate of the area sensor. We applied this technique to measure the absorbance of the fingertip. When calculating the sample absorbance, we use a neutral density filter with an absorbance of 3 as the reference. Fig. 3a shows one of the absorbance spectra, which was continuously measured at a scan rate of 20Hz. The wavelength range was 900–1700nm, which was limited by the area sensor used. The 1200nm spectral peak, which is attributable to the second overtones of C–H vibrations, is clearly observable. The signals at wavelengths greater than 1500nm were not usable for quantitative analysis because the irradiation power of the halogen lamp was insufficient. Time-varying absorption was observed because of the peripheral arterial pulse, as shown in Fig. 3b. The high-frequency oscillation (approximately 80bpm), which occurs once or twice per second, is due to peripheral vasodilation and vasoconstriction, whereas the low-frequency oscillation, which occurs approximately once every 8s, is due to respiration. The concentration of neutral fat within blood can be determined by extracting the periodically varying signal from the 1200nm peak. The calibration equation was obtained by in vitro triglyceride absorption spectroscopy (Wako Pure Chemical Industries Ltd., Japan). According to the calibration equation, absorbance is directly proportional to triglyceride weight with a proportionality constant of 0.012. Therefore, the observed absorption amplitude change 0.0045±0.0010 was determined to correspond to 0.37mg of triglyceride. To determine the corresponding triglyceride concentration, we estimated the blood volume obtained from a 133mm2 detection area on a fingertip. The blood volume was calculated to be 41μL based on four arteries, which were assumed to be cylindrical tubes that had a 1-mm diameter and 13-mm length [14]. The volumetric change was then estimated to be one tenth of the total blood volume, 4.1μL under systolic/diastolic pressure fluctuations [6], by neglecting the pulsations of the peripheral arteries. Thus, the 0.37mg of triglyceride measured in 4.1μL of blood yielded a concentration of 900mg/dL. Considering that neutral fat is generally one sixth of the weight of albumin, its concentration was estimated to be 130mg/dL, which may be in good agreement with the neutral fat content of normal human blood. However, it would be inaccurate to assume that the ratio of neutral fat to albumin is the same for everyone, as individuals have different blood component ratios. To avoid the effects of albumin concentration, it would be preferable to measure the time series variation after meals because the concentration of albumin remains unchanged, whereas the neutral fat concentration varies significantly after meals. Therefore, the index value of neutral fat in the blood was estimated by measuring the absorption changes after a meal. Fig. 4 shows an example blood fat index time series for a subject whose blood was sampled once every hour from 10:15 to 18:15. The subject ate uncontrolled meals at 7:30 and 12:30, and each measurement took less than 10min. Peaks can be observed 4–5h after each meal, which agrees with the results obtained in other invasive studies of blood fat concentration [5].