In this paper, we propose a novel NIR imaging method for the quantitative measurement of perfusion in peripheral tissues using planar reflectance imaging, which is simpler to implement compared to the tomographic method, by introducing mathematical modeling and a computational simulation. The validity of this novel method was evaluated using a murine model of hindlimb ischemia and therapeutic interventions directed at stimulation of angiogenesis. We showed that time-series imaging of intravenously injected ICG can provide quantitative information regarding peripheral tissue perfusion precisely enough to predict the prognosis of the ischemic tissues and to verify the therapeutic effects of proangiogenic factors such as VEGF and cAng1. For analysis of ICG dynamics, we designed a biologically meaningful mathematical model.
Previous studies that used ICG dynamics to estimate blood perfusion failed to provide quantitative information that was required for interindividual comparison. In those studies, blood perfusion was estimated based on the maximal fluorescence intensities of injected ICG, which depend heavily on the heterogenous optical properties of regional tissues 
. Moreover, static reflectance images from different animals or from the same animal at different time points are generally insufficient for yielding quantitative insights, because the fluorescence intensity depends on the depth of the signal and the volume and optical properties of the tissue. Thus, we performed time-series NIR fluorescence imaging to obtain the spatiotemporal dynamics of ICG and translated the time-series images into quantitative perfusion maps. Furthermore, we compensated for idiosyncratic variations in ICG pharmacokinetics to obtain quantitative information on perfusion rates, because the idiosyncratic variation of ICG dynamics depends on the level of anesthesia, systemic hemodynamic properties, and variations in hepatic clearance that can affect the evaluation of perfusion rate. This quantitative information enabled us to predict the necrotic profile of ischemic tissues.
The necrosis profiles of the ischemic limbs were variable among subjects even when the same surgery protocol and the same strain of animals were used 
. Necrosis profiles of the ischemic hindlimbs are considered to be determined by two main factors: the preexisting collateral circulation 
and vascular genesis 
. Variability of necrosis profiles between subjects could be caused primarily by differences in the preexisting collateral circulation, because the substrates of vascular genesis are preexisting collateral arteries 
. Therefore, we made two predictions. First, if the perfusion rate measured at post-op represents the preexisting collateral circulation, we predicted that this would show a significant relationship with future necrosis of the ischemic limb. As expected, the regional perfusion rate showed a significant inverse (sigmoidal shape) relationship with the probability of necrosis of the region. Second, we predicted that aiding angiogenesis by treatment with proangiogenic factors would alter the necrosis profile based on the preexisting collateral circulation. Significantly, treatment with the proangiogenic factors VEGF and cAng1 improved the necrosis profile.
Our method has several advantages for analyzing the perfusion and prediction of necrosis compared to conventional methods. First, quantitative measurement of tissue perfusion can allow interindividual comparative analysis of tissue perfusion. Second, this highly sensitive and quantitative measurement of perfusion can be used to equally distribute ischemic hindlimb model mice in control and experimental groups, which would lessen data distortion and decrease the variability of the experimental results. We have observed that the prognosis of ischemic tissues from animals subjected to identical surgical protocols varies considerably. This results in intrinsic data heterogeneity, which could severely misrepresent the outcome of drug interventions. Third, changes in perfusion due to proangiogenic factors in ischemic tissues can be evaluated and arteriogenic processes can be differentiated from angiogenesis because our method can classify conductance through the macrovasculature. We observed that combined treatment with VEGF and cAng1 restored the bimodal histogram perfusion rate pattern in the ischemic limbs, reflecting the emergence of macrovascular components. Histological and CT-angiogram analysis also showed that combined treatment with VEGF and cAng1 induced arteriogenesis along with angiogenesis. These results suggest that ICG dynamic perfusion imaging might be applied to differentiate macroangiopathy from microangiopathy.
Our method has several potential limitations as well. First, induction of ischemia transiently changes vascular permeability and dilatation, which play roles in functional recovery from acute hindlimb ischemia 
. Our method might underestimate perfusion rates in cases of increased vascular permeability because delayed washout kinetics of ICG as a result of leakage are typically interpreted as decreased perfusion. Moreover, our method cannot distinguish between perfusion changes due to blood flow versus microvessel density. Therefore, it is important to control the physiologic factors that can affect blood flow such as blood pressure, pulse rate, and body temperature for consistent measurement of tissue perfusion. Second, even though NIR can penetrate biological tissues deeply, the depth of NIR penetration nonetheless restricts measurement of perfusion to peripheral tissues. For analysis of deep-seated organs, endoscope-coupled technology might help to overcome this limitation. Finally, it is technically difficult to estimate accurate perfusion rates for normal tissues, especially in the case of normal tissues that have very fast ICG kinetics. For accurate measurements of such fast kinetics, the images with maximal intensity should be captured, which correspond to images of the time-to-peak. Since the 20 s time-series imaging of the Kodak imaging system is too long an interval to capture the time-to-peak of normal limbs, we developed a customized NIR fluorescence imaging system that can shorten the interval to as little as 1 s. In this way, we were able to capture the peak intensity successfully even in normal limbs. In addition we confirmed that these two imaging methods provided basically the same results when estimating perfusion rates in ischemic limbs, which showed slower ICG kinetics.
In summary, our results indicate that ICG dynamic perfusion imaging can quantitatively evaluate perfusion in peripheral tissues. Moreover, this imaging technology could be used to predict future necrosis profiles and evaluate angiogenesis. ICG perfusion imaging can be a cost- and time-efficient tool for diagnosis of patients with peripheral vascular insufficiencies. Although further studies evaluating ICG imaging for clinical applications are needed, we believe that this revolutionary imaging method may provide new impulse to the multidisciplinary analysis of biomedical imaging and may contribute to the development of clinical diagnostic techniques for peripheral vascular disease.