Using scanning time-domain instrumentation we recorded fluorescence projection mammograms on few breast cancer patients prior, during and after infusion of indocyanine green (ICG), while monitoring arterial ICG concentration by transcutaneous pulse densitometry. Late-fluorescence mammograms recorded after ICG had been largely cleared from the blood by the liver, showed invasive carcinomas at high contrast over a rather homogeneous background, whereas benign lesions did not produce (focused) fluorescence contrast. During infusion, tissue concentration contrast and hence fluorescence contrast is determined by intravascular contributions, whereas late-fluorescence mammograms are dominated by contributions from protein-bound ICG extravasated into the interstitium, reflecting relative microvascular permeabilities of carcinomas and normal breast tissue. We simulated intravascular and extravascular contributions to ICG tissue concentration contrast within a two-compartment unidirectional pharmacokinetic model.
© 2009 OSA
During the past decade considerable efforts were undertaken to develop various methods of diffuse optical imaging and spectroscopy for breast cancer detection and characterization. Recently, in-vivo broadband diffuse optical spectroscopy of known breast cancers was used to determine their physiologic status by measuring the concentration of intrinsic chromophores in tissue, e.g. the concentration of deoxy- and oxy-hemoglobin as well as water and lipid content in tissue . Such information was then used to predict and monitor the response to breast cancer neoadjuvant chemotherapy at an early stage . In addition, measurements of blood flow in tumors by diffuse correlation spectroscopy during early stages of neoadjuvant chemotherapy might be employed for this purpose . Optical mammography [4–10], on the other hand, aims at detecting breast tumors and at differentiating cancers from benign lesions. A retrospective clinical study  on time-domain projection optical mammography with limited spectral coverage (670 nm up to 884 nm) allowed to detect and distinguish carcinomas from normal breast tissue at a sensitivity and specificity of about 81% and 84%, respectively . However, sensitivity and specificity for discrimination of malignant and benign lesions were found to be well below 70%, being too low to be of clinical relevance. Broader spectral coverage will certainly improve the accuracy at which intrinsic chromophore concentrations can be deduced and will likely enhance specificity and sensitivity, although a retrospective clinical study on time-domain projection optical mammography that covered the range from 637 nm up to 985 nm yielded essentially the same sensitivity for cancer detection with respect to normal breast tissue . There are indications that broad spectral coverage together with tomographic recording may allow to discriminate benign from malignant lesions based on total hemoglobin concentration, at least for larger size tumors [10,12]. Another route to improve breast cancer detection is the use of (non-targeted) contrast agents which absorb and fluoresce in the near infrared (NIR) spectral range. Apart from the experimental dye Omocyanine [13,14], mostly indocyanine green (ICG) has been employed for contrast-enhanced mammography. However, ICG is the only dye that is FDA-approved at present and has been used since 50 years for various diagnostic purposes, e.g. to assess cardiac and hepatic function.
After administering an ICG bolus for absorption-based breast cancer detection, concurrent diffuse optical tomography and magnetic resonance imaging (MRI) was performed on three patients by Ntziachristos et al. . Contrast enhancement over intrinsic absorption was observed for an invasive ductal carcinoma and to some smaller extent for a fibroadenoma as well as for healthy breast tissue. Following administration of a bolus of ICG, Intes et al.  monitored dye absorption in breast tissue and therefore ICG inflow and outflow on three breast cancer patients. The same wash-in and wash-out rates were observed for a benign lesion (fibroadenoma) and for normal breast tissue. In contrast, both carcinomas investigated exhibited slower wash-in and wash-out rates compared to normal breast tissue, presumably caused by an increased geometric resistance of the chaotic tumor vasculature to blood flow. More recently, the absorption coefficient of ICG in carcinomas and normal breast tissue was obtained from time-resolved measurements recorded on eight breast cancer patients prior and after application of a bolus of ICG using a scanning time-domain mammograph . It was observed that during the entire wash-out phase the absorption coefficient of ICG normalized to the tissue total hemoglobin concentration was the same for normal and cancerous breast tissue.
Besides imaging or monitoring the enhanced absorption by ICG, fluorescence measurements have been used for breast cancer detection as well. Despite the low fluorescence quantum efficiency of ICG, fluorescence imaging is attractive since tissue autofluorescence background is low in the NIR spectral range and fluorescence signals can be detected with high sensitivity. Lately, three-dimensional fluorescence diffuse optical tomography was carried out by Corlu et al.  on three breast cancer patients, following application of a bolus of ICG. Fluorescence images of the breast carcinomas were in good agreement with those of MRI and of diffuse optical tomography, based on endogenous (absorption) contrast.
Common to all reports on ICG-enhanced breast imaging published so far is that absorption or fluorescence was detected during or shortly after administration of a bolus. At that time most of the dye is confined to the vascular compartment due to the binding of ICG to plasma proteins and the recorded signals basically represent the fractional volume of blood plasma in the local breast tissue. Although an improved contrast between tumor and normal breast tissue over pure absorption contrast was achieved in this way, no additional information is likely to be gained over that available from intrinsic absorption contrast by hemoglobin that would allow to discriminate benign from malignant lesions. On the other hand, it is known that newly grown microvessels in carcinomas tend to be hyperpermeable enabling even large molecules such as albumin to extravasate into the interstitium at an increased rate . Dynamic, contrast enhanced magnetic resonance imaging (DCE-MRI)  permits to assess microvessel permeability of tumors non-invasively by measuring the rate constant of extravasation of the particular contrast agent used. Similarly, ICG, which strongly and non-covalently binds to plasma proteins, may pass through large intercellular openings between endothelial cells  of hyperpermeable tumor capillaries or through vesicular-vacuolar organelles  present in the endothelial lining of tumor venules but may not extravasate at all or at much smaller rates through the capillary walls of benign lesions and normal breast tissue, thus allowing to discriminate benign and malignant lesions based on their microvessel permeability. However, in order to assess the permeability of the capillary wall, extravasated ICG must significantly contribute to the fluorescence signal recorded. During and shortly after application of the dye, fluorescence is dominated by ICG contained in the blood (“vascular phase”). Therefore, in order to estimate capillary permeability fluorescence imaging has to be carried out after ICG has been mostly cleared from the vascular compartment by the liver (“extravascular phase”).
In order to explore this approach we carried out a feasibility study on patients with a suspicious breast lesion, and recorded fluorescence images of the slightly compressed breast at various times after application of ICG using a prototype fluorescence mammograph. As will be shown on 3 cases, we observed strongly enhanced fluorescence contrast between cancers and normal breast tissue in late-fluorescence images, i.e. after the ICG had been largely eliminated from the blood by the liver, opposed to benign lesions. In this way, we could discriminate between benign and malignant lesions of patients suspected of breast cancer.
2. Materials and methods
2.1 Examination protocol
This study was approved by the institutional ethics committee of the Charité hospital and written informed consent was obtained from each patient prior to entering the study. Only women between 40 and 80 years of age with one suspicious breast lesion graded BI-RADS 4 or BI-RADS 5 according to the Breast Imaging-Reporting and Data System were enrolled and all patients received standard x-ray mammography and ultrasound prior to inclusion into the clinical study. In selected cases contrast-enhanced T1-weighted MRI scans were recorded using Gd-DTPA as contrast agent. In all cases a biopsy was scheduled after the optical measurements for histological workup.
A sterile aqueous ICG solution (PULSION Medical Systems AG, Munich, Germany) was administered intravenously as initial bolus at 0.25 mg per kg of body weight and a rate of 40 mg/min, immediately followed by a continuous ICG infusion at constant rate of 2 mg ICG/min. The initial bolus was applied to more quickly reach the steady state. The infusion rates were controlled using a calibrated syringe pump (Injectomat Agilia, Fresenius Kabi Deutschland GmbH, Germany). After 15 min to 25 min, depending on the time needed to record two consecutive mammograms, the infusion was stopped. Each patient received between 39 mg and 75 mg of ICG, which is safe and far below the maximum allowable daily dose of 5 mg/kg body weight. No adverse reactions or side effects were observed.
During the entire procedure the arterial ICG concentration was monitored at the finger pad by transcutaneous pulse densitometry (PC5000 LiMON liver function monitor, PULSION Medical Systems AG, Munich, Germany). Figure 1 shows a typical time profile recorded for a 72-year-old patient of 80 kg body weight (case 1). The temporal course of the arterial ICG concentration can be divided into different phases: a native phase (prior to administration of ICG), the bolus phase, a vascular phase (during constant infusion of ICG), the washout phase, and an extravascular phase (arterial ICG concentration back to baseline level). The ICG concentration values shown in Fig. 1 were obtained from the output data of the LiMON liver function monitor together with the steady state condition for the vascular phase:Equation (1) assumes that ICG is exclusively eliminated from the blood plasma by the liver, neglecting any other elimination pathway like extravasation of the dye molecule into the extravascular-extracellular space (EES) of the breast tissue investigated and consecutive clearance by the lymphatic system. Although the latter pathway is decisive for the dye uptake in cancerous breast tissue and the corresponding contrast enhancement in optical mammograms, its influence on the arterial ICG concentration is negligible. The elimination rate kelim can be derived by a mono-exponential fit from the washout phase and amounts to 0.164/min for the 72-year-old patient (red solid line in Fig. 1). It has been reported previously that the disappearance of ICG from plasma follows a bi-exponential decay in humans  and pigs . For all patients, however, the decrease of the LiMON output signal after the end of the infusion could be fitted to a simple exponential and bi-exponential decays of the LiMON output signal were not observed within error limits. The disappearance rates varied between 0.10/min ≤ kelim ≤ 0.30/min, and the average taken over all patients amounted to kelim,ave = 0.195/min, being close to the value (0.179/min) reported in the literature .
Generally, optical mammograms were recorded at constant arterial ICG concentration, i.e. during the native phase prior to ICG administration (Scan 1), the vascular phase during the infusion (Scan 2, Scan 3), and the extravascular phase after ICG clearance (Scan 4, Scan 5). Scans 1 to 3 were accomplished after the patient’s breast was initially positioned and compressed by the radiologist. During the wash-out phase the breast was released and recompressed again for Scans 4 and 5, which were taken at least 25 min after the infusion was terminated. It had been observed previously, that total hemoglobin concentration in breast tissue changed with increasing compression pressure . In our case we did not observe such changes after recompressing the released breast to the same thickness. In selected cases, fluorescence mammograms were recorded even up to 24 hours after the infusion.
The principle of operation of our current apparatus is based on the PTB’s time-domain optical mammograph where a slab-like detection geometry in combination with a scanning approach is utilized . In time-domain optical mammography the collected data represent times of flights of photons which diffuse through the tissue from the source position to the detector position and are measured by time-correlated single photon counting (TCSPC). In TCSPC the delay between a synchronization trigger, delivered by a pulsed light source and the detector signal are recorded and accumulated for many photon-detection events to give a histogram of distributions of times of flights of photons also referred to as temporal point spread function (TPSF).
In Fig. 2 the principle of operation is sketched, a detailed description of the mammograph is given in Refs. 26 and 27. Briefly, the breast of a patient is gently compressed between two glass plates of the compression unit, which can be adjusted in both vertical and horizontal position and can be rotated for cranio-caudal, medio-lateral or oblique projections. Picosecond laser pulses from a diode laser module (Sepia II, PicoQuant GmbH, Berlin, Germany) were coupled into an optical fiber, the distal end of which scanned the entrance compression plate (top) in 2D in a meander-like fashion. For the excitation of ICG, laser pulses of 780 nm wavelength were used with a repetition-rate of 64 MHz and an average output power of 15 mW. Furthermore, four additional laser heads (660 nm, 8 mW; 797 nm, 2.8 mW; 934 nm, 11 mW, and 1064 nm, 7 mW, each at 32 MHz repetition rate) were available for measuring the absorption by intrinsic chromophores. An opposite (on-axis geometry), 6 mm diam. bifurcated fiber bundle (NA 0.54) at the exit compression plate (bottom) moved in tandem with the source fiber and collected diffusely transmitted laser and fluorescence photons at each scan position. The width of the instrument response function is determined by this fiber bundle and amounts to about 400 ps, independent of the laser-wavelength used.
The fiber bundles were connected to computer controlled detector boxes, which are rack mounted cassettes each containing a Peltier-cooled GaAs photomultiplier (H7422P-60 or H7422P-50, Hamamatsu Photonics Deutschland GmbH, Herrsching, Germany), a preamplifier, a variable neutral density filter and diaphragm-based attenuation optics, as well as a filter wheel equipped with various sets of interference filters. Because optical mammograms were recorded at different phases (see examination protocol) with relatively high ICG concentration (vascular phase) as well as very low ICG concentration (extravascular phase) the dynamic range of the detector boxes was large allowing a gradual light attenuation by six orders of magnitudes. The preamplified detector signals were processed by commercial TCSPC electronics (SPC 134, Becker&Hickl GmbH, Berlin, Germany).
2.3 Data acquisition and processing
The breast scans were performed in a meander-like fashion with a step size of 2.5 mm. At each scan position TPSFs were recorded simultaneously for all detection channels with integration times of 150 ms/step and stored on disk together with the corresponding motor position. Including all latency periods due to data processing and storage the times needed for a complete breast scan sum up to about 6 min to 10 min depending on breast size. Count rates in the center region of the breast were set to typically 300 kHz for the fluorescence and the excitation channel each. The reversal points of the meander trajectory were found by monitoring the photomultiplier current of the excitation channel. When a certain current threshold was exceeded due to the reduced thickness of the breast near its edge the scanner motion was stopped and the next scan line in the opposite direction was commenced. For this reason no data could be acquired at positions less than 1 cm away from the breast’s edge. To measure ICG fluorescence a 800 nm long-pass filter was used to reject laser photons. For counting photons at the excitation wavelength (780 nm) a 790 nm short-pass filter was utilized suppressing ICG fluorescence. Both, fluorescence and laser-photon measurements were performed simultaneously.
Because of point source excitation and the large number of positions sampled– 2000 to 4000 depending on breast size – a 2D projection image of the breast can be easily obtained by simply taking either integral photon counts, photons from dedicated time windows or statistical moments of the TPSFs and plotting them against the recorded source position. Thus no reconstruction is necessary and projection images can be displayed online while running an acquisition.
Figure 3a and 3b display raw images of normalized integrated photon counts at the fluorescence wavelength (3a) and laser wavelength (3b), which were recorded in transmission (on-axis channel) during the extravascular phase (case 1, Scan 5, s. Figure 1). The images were corrected for background noise by baseline subtraction of the TPSFs only. In order to visualize the center breast-region on the same gray scale as the edges of the breast, where the light intensities increase dramatically due to a decreasing tissue thickness (edge effects), reciprocal values of the normalized integral photon counts are shown. Both mammograms (Fig. 3a, 3b) are very similar in appearance and are dominated by absorption of the laser and fluorescence radiation by blood (hemoglobin). The carcinoma, however, cannot be detected in the fluorescence mammogram because of cancellation effects, i.e. the higher fluorescence intensity of the carcinoma is compensated for by the higher absorption of laser and fluorescence radiation due to its increased blood content.
To correct transmitted fluorescence for the inhomogeneous absorption of the breast tissue we calculated ratio images of (normalized) fluorescence and laser intensities. An example for a ratio image is shown in Fig. 3c, which has been generated by simply dividing Fig. 3b by Fig. 3a, taking into account that inverse photon counts are displayed in the top row of Fig. 3. The intrinsic absorption contrast has essentially been eliminated in the ratio image, displaying the carcinoma above a rather homogeneous background. In such ratio images higher (normalized) ICG fluorescence intensity corresponds to enhanced dye concentration.
Although part of the edge effects apparent in the laser and fluorescence image cancel out in the corresponding ratio image, an algorithm using the local breast thickness deduced from the first moment of the TPSFs of laser photons was used to rescale total laser and fluorescence counts to constant breast thickness . In this way mammograms were corrected for the decreasing breast thickness towards the edges of the breast. Figure 3d shows the improvement of the ratio image 3c, when applying this correction algorithm.
2.4 Pharmacokinetic model
We correlate fluorescence contrast observed in ratio images during and after the ICG infusion with tissue ICG concentration contrast calculated within a two-compartment pharmacokinetic model with unidirectional flow of ICG from the local vascular compartment to the local extravascular-extracellular space followed by slow outflow from the local EES (s. Figure 4 , right hand side). We used our model to calculate the local ICG concentration in plasma Cp(t) and in the EES Ce(t) and hence the resulting total ICG concentration in tissue Ct(t) = vpCp(t) + veCe(t), which is responsible for the fluorescence observed. Here vp and ve are the fractional volumes of blood plasma and EES in the local breast tissue, respectively. The rate equations for the ICG concentrations in both compartments readFig. 4 illustrates the open redistribution model of Ott et al.  consisting of a global (whole body) vascular space and global extravascular-extracellular space to deduce the forward (k12), and backward (k 21) exchange rates between both compartments and the elimination rate kelim from the bi-exponential plasma disappearance curve CA(t) measured in pigs. However, this interpretation of the plasma disappearance curve is not required in the present analysis.
As usual  we assume that the leakage of ICG into the EES of breast tissue including the lesion does not affect the global ICG plasma concentration, i.e. the arterial input function. Our two-compartment pharmacokinetic model is closely related to the model of Brix et al.  which was used to analyze DCE-MRI data and which takes the inflow of the contrast agent into account. Whereas the model used by these authors accounts for both the extravasation of the contrast agent (Gd-DTPA) as well as its backflow into the microvasculature due to the rapid diffusion of the extravasated small molecular contrast medium, our unidirectional model neglects backflow because of the slow diffusion of ICG bound to plasma proteins (retention effect, s. Discussion). Previously, unidirectional pharmacokinetic models were used to account for the extravasation of the macromolecular contrast medium Gd-DTPA covalently bound to plasma proteins [30,31].
We solved the rate equations during and after termination of the ICG infusion of duration τinf, assuming a constant infusion rate Sinf, but neglecting the initial bolus applied experimentally to more quickly reach the steady state. The analytical solutions to Eqs. (2) and (3) are summarized in the Appendix.
Three cases are presented: two (malignant) invasive ductal carcinomas (cases 1, 2) and one (benign) fibroadenoma (case 3). Optical mammograms shown here display ratio images of normalized integral counts of transmitted fluorescence and laser photons of the on-axis detection channel unless otherwise indicated. Measurements during the vascular phase, i.e. during ICG infusion (Scan 3, s. Figure 1), and during the extravascular phase, i.e. after infusion and clearance of ICG from the vascular system by the liver (Scan 5, s. Figure 1), are compared using global min-max gray scaling for both phases without thresholding.
Figure 5a and 5b show cranio-caudal projection optical mammograms taken at a compression-plate distance of 6.1 cm from a 72-year-old patient (80 kg) carrying a tumor in her right breast. The tumor was identified by histology as medium differentiated invasive ductal carcinoma with a largest extent of 1.6 cm. Due to a low amount of glandular tissue (ACR index 1; American College of Radiology) the mass was clearly visible in the x-ray image (Fig. 5c) and classified as BI-RADS 5 (highly suggestive of malignancy).
Optical mammograms show the tumor at high contrast of almost 1.5 in the extravascular phase (Fig. 5b), distinctly silhouetted against a homogenous and featureless background. During the vascular phase (Fig. 5a) the tumor is also visible but at lower contrast of less than 1.2 over a rather inhomogeneous background. Different structures arising from superficial blood-vessels show up at a contrast similar to the one at the tumor site. In Fig. 6 the fluorescence ratio mammogram taken during the vascular phase with individual min-max gray scaling (Fig. 6b) is compared to an intrinsic measurement prior to ICG administration (native phase, Scan 1, s. Figure 1), showing reciprocal normalized photon counts at 660 nm of a late time window (Fig. 6a). Basically the same vascular structures (including the tumor) can be seen in both images, indicating that fluorescence observed during ICG infusion originates predominantly from ICG within the vascular compartment. Therefore, no additional information can likely be extracted from fluorescence measurements in the vascular phase that are not already available from the intrinsic absorption contrast by hemoglobin.
A more challenging case is provided by a 51-year-old patient (60 kg), who had a hypoechoic lesion in her ultrasound findings approximately 1 cm below the skin. X-ray mammograms (s. Figure 7 ) showed very dense glandular tissue (ACR index 4) and only a slight architectural distortion at approx. 3 o’clock in the left breast could be identified retrospectively and was classified as BI-RADS 4 (suspicious abnormality).
A DCE-MRI measurement finally revealed two adjacent lesions in the left breast at lateral site near the breast edge and close to the chest wall (s. Figure 8c ). After biopsy these lesions were identified as well differentiated invasive ductal carcinomas of size 1.1 x 1.0 x 0.8 cm3 joined by a low-grade ductal carcinoma in situ.
In Fig. 8a the cranio-caudal fluorescence ratio mammogram from the vascular phase (compression-plate distance: 4.6 cm) shows a contrast gradient towards the chest wall and in the lower left corner approx. 3 cm from the breast’s edge a spot at slightly enhanced local contrast can be seen. This spot is in reasonable agreement with the location of the tumors visible in the MRT image in Fig. 8c. Besides the slight contrast enhancement, the diffuse, irregular shaped region of interest lacks a sharp contour to the surrounding tissue. However, in the ratio image from the extravascular phase depicted in Fig. 8b (compression-plate distance: 5.9 cm) the contrast of the suspicious region increased to about 1.8, much higher than the minor enhancements that can be seen near the chest wall and next to the mammilla.
Because of light scattering, diffuse optical mammography has poor spatial resolution (not better than 1 cm to 2 cm) and, therefore, does not resolve both neighboring lesions seen in the MR image. Rather, a single area of enhanced fluorescence intensity appears in the fluorescence ratio image of the extravascular phase.
A cross-check with respect to the diagnostic potential of our method was provided by this 52-year old patient (67 kg) with a pathologically validated (benign) fibroadenoma in her left breast. An ultra-sound image showed the lesion approximately 2 cm below the skin with an area of 1.8 x 1.0 cm2. No irregularities were evident in the x-ray image.
None of the fluorescence ratio mammograms of Fig. 9a and 9b (cranio-caudal projection; compression-plate distance 6.8 cm) shows any localized contrast enhancement inside a well defined area. In the vascular phase (Fig. 9a) some blood-vessels with increased contrast of 1.16 may be identified, that are no longer visible in the extravascular phase (Fig. 9b). Moreover, this is the only case where the maximum contrast was found in the vascular phase and not in the extravascular phase. Apparently, the ICG uptake in the fibroadenoma does not differ substantially from that of healthy tissue.
These three examples illustrate the more general picture observed for a larger patient cohort, the results of which are presently under detailed analysis. Generally, lesions that were identified by histopathology as invasive ductal carcinomas were seen at high tumor to background contrast in the extravascular phase. In contrast, none of the fibroadenomas and mastopathic changes validated by histology produced a well-defined area of enhanced fluorescence intensity in that phase. In some cases, however, glandular tissue resulted in large, ill-focused areas of high fluorescence contrast, leading to diffuse background enhancement. In a few cases serial late-fluorescence images were recorded at selected times for several hours (up to 24 hours) without observing any changes in their fluorescence contrast.
It is well known that the permeability P of the microvascular wall and the rate constant for extravasation, PSρ, strongly depend on the molecular size of the solute [30–33], and, therefore, contrast media of appropriate molecular size might allow to discriminate benign and malignant lesions based on their microvessel permeability. The rate constant for extravasation of the macromolecular contrast medium Gd-DTPA-albumin comprising about 30 Gd-DTPA moieties covalently linked to the protein was measured by DCE-MRI in an animal model for breast cancer . Based on their higher rate constant for extravasation of Gd-DTPA-albumin, high grade cancers could be discriminated non-invasively from benign lesions (fibroadenomas), whereas such determination was not possible when the small molecular contrast medium Gd-DTPA was used since the latter extravasated non-selectively through capillary walls of normal and pathological breast tissue [31,34].
After systemic administration, ICG rapidly  and almost entirely  binds to various plasma proteins [35,37,38], turning into a macromolecular contrast medium (blood pool agent), with the molar fraction of the free dye estimated  to be below f free < 0.5%. The vascular hyperpermeability of carcinomas towards macromolecules explains the higher fluorescence intensity of the carcinomas observed during the extravascular phase, after ICG had been cleared from the vascular compartment by the liver. On the other hand, since none of the fibroadenomas produced a well-defined area of enhanced fluorescence intensity during this phase we conclude that the vascular permeability of these benign lesions for plasma proteins was comparable to that of normal breast tissue.
The apparent PSρ value of ICG is the sum of the permeability-surface-area products of ICG bound to various plasma proteins and possibly of free ICG each weighted by its molar fraction. The PSρ values of free ICG and ICG bound to plasma proteins are presently unknown and, therefore, we may take for the ratio PSρ free/PSρ bound = 100 which is the corresponding value measured by DCE-MRI for free Gd-DTPA and Gd-DTPA-albumin . This assumption is reasonable because of the close match of molecular weights of the corresponding free and bound contrast agents involved. Taking into account this ratio and the small molar fraction f free of free ICG we conclude that the apparent PSρ value is dominated by ICG bound to plasma proteins, consistent with the observed discrimination of carcinomas and benign lesions in the extravascular phase. This discrimination would not be expected if ICG extravasated predominantly as free dye. To account for the possibility that ICG may also bind to plasma proteins that are larger than albumin, the PSρ value of plasma-bound ICG that enters our pharmacokinetic model was adopted to 0.2 µl/(min cm3 tissue) for tumors (s. Table 1 ) which corresponds to the lower limit of the values reported for Gd-DTPA-albumin [30,31,39]. For healthy parenchyma a twofold lower extravasation rate (PSρ value) was assumed. Previously, permeability-surface-area products being larger by three to four orders of magnitude than the value adopted by us were reported for breast cancer patients and for an adenocarcinoma tumor rat model using ICG as contrast agent [28,40]. Such high rates of extravasation and backflow of ICG from the local EES into the local plasma compartment cannot account for ICG to persist in breast tissue for 24 hours as observed by us, and possibly the high rates reported reflect plasma tissue perfusion.
After extravasation, the protein bound ICG hardly diffuses because of its large effective molecular weight, and stays within the EES close to the capillary wall . This effect is also known from radioactively labeled albumin [41,42], which accumulated in cancers of animal tumor models to a considerable extent and was retained in carcinomas for several days. The same behavior was observed for Evans blue  which after i.v. injection strongly binds non-covalently to albumin, similar to the binding of ICG to plasma proteins. This situation which is known as enhanced permeability and retention (EPR) effect of cancers and which is presently exploited to target solid carcinomas by macromolecular chemotherapeutics , is similar to our observation of enhanced and persistent fluorescence contrast of invasive carcinomas in late fluorescence ratio images. The retention effect explains that ICG fluorescence could be still detected by us even twenty four hours after application of the dye. A possible slow component  of the plasma disappearance curve of ICG was not observed for the patients included in our study and cannot explain the persistence of ICG fluorescence for 24 hours. Correspondingly, the wash-out rate was estimated to be on the order of kout ≈(10 hours)−1 (s. Table 1). For simplicity, we have taken the same wash-out rate for all three kinds of tissue (breast carcinoma, parenchyma, fat). Previously, ICG fluorescence had been observed in a canine breast cancer model up to 72 hours after application of a bolus of the dye but late-fluorescence had not been explained by a pharmacokinetic model .
Table 1 summarizes the numerical values of the various parameters that we used for the simulation of ICG tissue concentration and concentration contrast. Since a mono-exponential decay of the arterial ICG concentration was observed in each patient, we set the amplitude (concentration) of the slow component of the arterial input function A2 = 0. In this case, the analytical solutions of the rate equations simplify, since q = 1, m1 = kelim and the slow decay rate m2 no longer enters the Eqs. (A1) - (A6), given in the Appendix (see Table A1). The amplitude A1 and the constant infusion rate Sinf were calculated from the total amount of ICG administered, taking for the plasma volume the (average) value of 2.13 l reported in the literature  for a body weight of 70 kg. Although blood perfusion in fat, breast parenchyma and breast tumors [46,47] will most likely be different, we have assumed a common value of Fρ = 0.067 ml blood/(min cm3 tissue) for the sake of simplicity and converted to plasma perfusion using the small vessel hematocrit (Hct)sv ≈0.25. However, the value chosen for Fpρ does not affect ICG concentration contrast between the various kinds of breast tissue (breast carcinoma, parenchyma, fat) during the infusion and during the extravascular phase. The parameter ve is not included in Table 1 since all amplitudes appearing in the solutions Ce(t) given in Eq. (A4) and Eq. (A6) are proportional to kePSρ = PSρ/ve and, therefore, the contribution veCe(t) of the EES to the total tissue concentration Ct(t) = vpCp(t) + veCe(t) does not depend on the fractional volume ve.
With these parameter values entering our pharmacokinetic model, we calculated time courses of the vascular and extravascular contributions to the total ICG concentration in carcinoma, parenchyma and fat (s. Figure 10a , 10b). The ICG concentrations in carcinoma and breast parenchyma coincide in Fig. 10a since we have assumed the same fractional plasma volume for both kinds of tissue being two times higher than that of fat. Figure 10b illustrates the higher concentration of extravasated ICG in the carcinoma compared to parenchyma and fat because of its higher assumed permeability-surface-area product. After termination of the infusion, the vascular contribution rapidly decays in each case, essentially with the plasma disappearance rate of ICG (m1 ≈0.18 min−1, s. Table 1), allowing us to detect the small extravasated ICG contribution. At sufficiently long times the extravascular component disappears with the small outflow rate kout. Figure 10c and 10d display the vascular contribution, the extravascular contribution and the total tissue concentration of ICG, each normalized to the total (tissue) concentration in fat, illustrating the time course of the ICG concentration contrast that determines fluorescence contrast observed in mammograms. Similar to Fig. 1, the ICG concentration contrast Ct/Ct(fat) can be separated into an early vascular phase and a late extravascular phase, dominated by vascular contributions and extravascular contributions, respectively. During infusion, normalized vascular contributions vpCp/Ct(fat) and hence tissue concentration contrast Ct/Ct(fat) reach a value close to the ratio of the corresponding fractional plasma volumes, whereas normalized extravascular contributions veCe/Ct(fat) nearly linearly increase (s. Figure 10c, 10d). After the infusion has been stopped, normalized extravascular contributions rapidly increase because ICG is eliminated from the blood (plasma) in fat, lowering Ct(fat). Subsequently the ICG tissue concentration contrast approaches the saturation value Ct/Ct(fat) = PSρ /(PSρ)fat of the extravascular contrast veCe/veCe(fat), provided the outflow rate kout is the same for the three kinds of tissue considered. In contrast, normalized vascular contributions drop to zero after ICG is cleared from the vascular compartment (s. Figure 10c), consistent with Fig. 1. In summary, Fig. 10c and 10d explain the contrast between carcinoma and normal breast tissue observed in ratio fluorescence mammograms recorded during the vascular and the extravascular phases. Furthermore, Figs. 10c, 10d support our observation that contrast in late-fluorescence ratio images does essentially not change within hours. It follows from our simulations that during infusion fluorescence intensity of ratio images reflects ICG contained in the vascular compartment whereas contributions from extravasated ICG can be neglected. Hence, fluorescence intensity is a measure of the relative plasma (blood) volume in this situation described by the parameter vp. However, this parameter is also accessible from laser mammograms reflecting intrinsic absorption of laser radiation by blood (total hemoglobin) and recorded without contrast agent applied. It is known that carcinomas and benign breast lesions generally cannot be discriminated by their (relative) blood or plasma volume . The late-fluorescence mammograms recorded by us during the extravascular phase predominantly detect extravasated ICG, since the dye has been largely cleared from the vascular compartment by the liver. Therefore late-fluorescence ratio mammograms reflect the permeability of the microvasculature towards macromolecules such as plasma proteins and hence allow us to discriminate breast carcinomas from benign lesions.
In order to quantitatively derive (relative) microvascular permeabilities of normal breast tissue, benign lesions and carcinomas, ICG concentration contrast needs to be reconstructed in these tissues from late-fluorescence ratio images. To this end tomographic data have to be recorded at the laser wavelength and over the fluorescence band rather than projection mammograms as in the present study.
Our present examination protocol needs to be optimized for clinical applications. At the dose applied presently, signal-to-noise ratios in fluorescence mammograms were generally high. Future investigations must determine the minimal dose of ICG required that yields sufficient signal-to-noise ratios in fluorescence ratio images even during the extravascular phase. Furthermore, for clinical use the examination time must be reduced, e.g. by shortening the infusion time. Although mammograms recorded during the native and vascular phase add some information, it is conceivable that for clinical routine recording of fluorescence mammograms during the extravascular phase only is sufficient to detect and discriminate carcinomas from benign lesions.
Within an exploratory clinical study we recorded projection laser and fluorescence mammograms of breast cancer patients prior, during and after an ICG infusion. Results on three cases were presented that illustrate the general picture observed for a larger patient cohort. Late-fluorescence ratio images, that were taken after the infusion and after clearance of ICG from the vascular compartment by the liver, showed invasive ductal carcinomas at high tumor-to-background contrast, permitting cancers to be distinguished from benign lesions, such as fibroadenomas and mastopathic changes, which did not produce a (focused) fluorescence contrast. Simulations carried out within a two-compartment pharmacokinetic model suggest that during infusion fluorescence largely originates from ICG within the vascular compartment, whereas late-fluorescence mammograms and fluorescence ratio images predominantly display fluorescence of protein-bound ICG that extravasated into the (local) extravascular-extracellular space, thus assessing microvascular permeability. In order to detect and differentiate carcinomas in late-fluorescence ratio images, breast cancer microvascular leakiness for plasma proteins must be higher compared to benign lesions, parenchyma and fat. However, permeability of tumor vasculature varies considerably both spatially and temporally and depends on the specific kind of tumor and on the host tissue . Therefore, this requirement may not be met in each case. The sensitivity of current diagnostic breast imaging modalities (x-ray mammography, breast MR imaging, ultrasound) for detecting breast lesions is high, yet limited specificity leads to large numbers of unnecessary biopsies. By providing information on tumor capillary permeability for plasma proteins, late-fluorescence mammography using ICG as contrast agent may become an adjunct clinical modality to improve specificity of current diagnostic breast imaging.
In the following we summarize the solutions of the rate Eq. (2),(3) during and after termination of the ICG infusion of duration τinf, assuming a constant infusion rate Sinf. Based on the bi-exponential plasma disappearance curve [22,23] of ICG after a bolus
the time dependence of the arterial input function during the infusion (0 ≤ t ≤ τinf) is expressed as
Here m 1 and m 2 denote the rates of the fast and the slow decay component of the ICG plasma disappearance curve, the expressions for q and kelim are given in Table A1. From the rate equations the dye concentrations in the local vascular compartment and in the local EES are obtained for 0 ≤ t ≤ τinf as
and for times t ≥ τinf as
The amplitudes of the various exponential terms and other abbreviations used are summarized in Table A1. For Fpρ → ∞, the expression in Eq. (A4) agrees with the permeability-limited results reported by Tofts and Berkowitz .
As was mentioned in Section 4, ICG binds to various plasma proteins (e.g. low and high density lipoproteins, immunoglobulin G and albumin) of different sizes and hence different rate constants for extravasation. It was observed in animal experiments that saturation of the binding sites of some of the proteins occurs at higher ICG concentrations . In that case, the apparent PSρ constant depends on the ICG concentration and the pharmacokinetic model becomes non-linear. Such effects have been neglected in the present analysis.
This work was supported in part by the German Ministry of Education and Research, grant 13N8774.
References and links
1. A. Cerussi, N. Shah, D. Hsiang, A. Durkin, J. Butler, and B. J. Tromberg, “In vivo absorption, scattering, and physiologic properties of 58 malignant breast tumors determined by broadband diffuse optical spectroscopy,” J. Biomed. Opt. 11(4), 044005 (2006). [PubMed]
2. A. Cerussi, D. Hsiang, N. Shah, R. Mehta, A. Durkin, J. Butler, and B. J. Tromberg, “Predicting response to breast cancer neoadjuvant chemotherapy using diffuse optical spectroscopy,” Proc. Natl. Acad. Sci. U.S.A. 104(10), 4014–4019 (2007). [PubMed]
3. C. Zhou, R. Choe, N. Shah, T. Durduran, G. Yu, A. Durkin, D. Hsiang, R. Mehta, J. Butler, A. Cerussi, B. J. Tromberg, and A. G. Yodh, “Diffuse optical monitoring of blood flow and oxygenation in human breast cancer during early stages of neoadjuvant chemotherapy,” J. Biomed. Opt. 12(5), 051903 (2007). [PubMed]
4. M. A. Franceschini, K. T. Moesta, S. Fantini, G. Gaida, E. Gratton, H. Jess, W. W. Mantulin, M. Seeber, P. M. Schlag, and M. Kaschke, “Frequency-domain techniques enhance optical mammography: initial clinical results,” Proc. Natl. Acad. Sci. U.S.A. 94(12), 6468–6473 (1997). [PubMed]
5. L. Götz, S. H. Heywang-Köbrunner, O. Schütz, and H. Siebold, “Optical mammography on preoperative patients (Optische Mammographie an präoperativen Patientinnen),” Akt. Radiol. 8, 31–33 (1998).
6. S. B. Colak, M. B. van der Mark, G. W. 't Hooft, J. H. Hoogenraad, E. S. van der Linden, and F. A. Kuijpers, “Clinical optical tomography and NIR spectroscopy for breast cancer detection,” IEEE J. Sel. Top. Quantum Electron. 5(4), 1143–1158 (1999).
7. H. Dehghani, B. W. Pogue, S. P. Poplack, and K. D. Paulsen, “Multiwavelength three-dimensional near-infrared tomography of the breast: initial simulation, phantom, and clinical results,” Appl. Opt. 42(1), 135–145 (2003). [PubMed]
8. P. Taroni, A. Torricelli, L. Spinelli, A. Pifferi, F. Arpaia, G. Danesini, and R. Cubeddu, “Time-resolved optical mammography between 637 and 985 nm: clinical study on the detection and identification of breast lesions,” Phys. Med. Biol. 50(11), 2469–2488 (2005). [PubMed]
9. D. Grosenick, K. Th. Moesta, M. Möller, J. Mucke, H. Wabnitz, B. Gebauer, Ch. Stroszczynski, B. Wassermann, P. M. Schlag, and H. Rinneberg, “Time-domain scanning optical mammography: I. Recording and assessment of mammograms of 154 patients,” Phys. Med. Biol. 50(11), 2429–2449 (2005). [PubMed]
10. S. P. Poplack, T. D. Tosteson, W. A. Wells, B. W. Pogue, P. M. Meaney, A. Hartov, C. A. Kogel, S. K. Soho, J. J. Gibson, and K. D. Paulsen, “Electromagnetic breast imaging: results of a pilot study in women with abnormal mammograms,” Radiology 243(2), 350–359 (2007). [PubMed]
11. H. Rinneberg, D. Grosenick, K. T. Moesta, H. Wabnitz, J. Mucke, G. Wübbeler, R. Macdonald, and P. Schlag, “Detection and characterization of breast tumours by time-domain scanning optical mammography,” Opto-Electronics Review 16(2), 147–162 (2008).
12. B. W. Pogue, S. C. Davis, X. Song, B. A. Brooksby, H. Dehghani, and K. D. Paulsen, “Image analysis methods for diffuse optical tomography,” J. Biomed. Opt. 11(3), 033001 (2006).
13. C. Perlitz, K. Licha, F. D. Scholle, B. Ebert, M. Bahner, P. Hauff, K. T. Moesta, and M. Schirner, “Comparison of two tricarbocyanine-based dyes for fluorescence optical imaging,” J. Fluoresc. 15(3), 443–454 (2005). [PubMed]
14. R. Ziegler, Modeling photon transport and reconstruction of optical properties for performance assessment of laser und fluorescence mammographs and analysis of clinical data, Dissertation, Department of Physics, Free University of Berlin, Germany, 2008, (http://www.diss.fu-berlin.de/diss/receive/FUDISS_thesis_000000005928).
15. V. Ntziachristos, A. G. Yodh, M. Schnall, and B. Chance, “Concurrent MRI and diffuse optical tomography of breast after indocyanine green enhancement,” Proc. Natl. Acad. Sci. U.S.A. 97(6), 2767–2772 (2000). [PubMed]
16. X. Intes, J. Ripoll, Y. Chen, S. Nioka, A. G. Yodh, and B. Chance, “In vivo continuous-wave optical breast imaging enhanced with Indocyanine Green,” Med. Phys. 30(6), 1039–1047 (2003). [PubMed]
17. A. Corlu, R. Choe, T. Durduran, M. A. Rosen, M. Schweiger, S. R. Arridge, M. D. Schnall, and A. G. Yodh, “Three-dimensional in vivo fluorescence diffuse optical tomography of breast cancer in humans,” Opt. Express 15(11), 6696–6716 (2007). [PubMed]
18. P. Carmeliet and R. K. Jain, “Angiogenesis in cancer and other diseases,” Nature 407(6801), 249–257 (2000). [PubMed]
19. A. L. Baert, K. Sartor, A. Jackson, D. L. Buckley, and G. J. M. Parker, eds., Dynamic Contrast-Enhanced Magnetic Resonance Imaging in Oncology, Springer, Berlin, Heidelberg, Germany (2005).
20. H. Hashizume, P. Baluk, S. Morikawa, J. W. McLean, G. Thurston, S. Roberge, R. K. Jain, and D. M. McDonald, “Openings between defective endothelial cells explain tumor vessel leakiness,” Am. J. Pathol. 156(4), 1363–1380 (2000). [PubMed]
21. D. Feng, J. A. Nagy, H. F. Dvorak, and A. M. Dvorak, “Ultrastructural studies define soluble macromolecular, particulate, and cellular transendothelial cell pathways in venules, lymphatic vessels, and tumor-associated microvessels in man and animals,” Microsc. Res. Tech. 57(5), 289–326 (2002). [PubMed]
22. D. K. F. Meijer, B. Weert, and G. A. Vermeer, “Pharmacokinetics of biliary excretion in man. VI. Indocyanine green,” Eur. J. Clin. Pharmacol. 35(3), 295–303 (1988). [PubMed]
23. P. Ott, L. Bass, and S. Keiding, “The kinetics of continuously infused indocyanine green in the pig,” J. Pharmacokinet. Biopharm. 24(1), 19–44 (1996). [PubMed]
24. S. A. Carp, J. Selb, Q. Fang, R. Moore, D. B. Kopans, E. Rafferty, and D. A. Boas, “Dynamic functional and mechanical response of breast tissue to compression,” Opt. Express 16(20), 16064–16078 (2008). [PubMed]
25. D. Grosenick, H. Wabnitz, H. H. Rinneberg, K. Th. Moesta, and P. M. Schlag, “Development of a time-domain optical mammograph and first in vivo applications,” Appl. Opt. 38(13), 2927–2943 (1999).
26. A. Hagen, O. Steinkellner, D. Grosenick, M. Möller, R. Ziegler, T. Nielsen, K. Lauritsen, R. Macdonald, and H. Rinneberg, “Development of a multi-channel time-domain fluorescence mammograph,” Proc. SPIE 6434, 64340Z (2007).
27. M. Möller, H. Wabnitz, A. Kummrow, D. Grosenick, A. Liebert, B. Wassermann, R. Macdonald, and H. Rinneberg, “A four-wavelength multi-channel scanning time-resolved optical mammograph,” Proc. SPIE 5138, 290–297 (2003).
28. D. J. Cuccia, F. Bevilacqua, A. J. Durkin, S. Merritt, B. J. Tromberg, G. Gulsen, H. Yu, J. Wang, and O. Nalcioglu, “In vivo quantification of optical contrast agent dynamics in rat tumors by use of diffuse optical spectroscopy with magnetic resonance imaging coregistration,” Appl. Opt. 42(16), 2940–2950 (2003). [PubMed]
29. G. Brix, F. Kiessling, R. Lucht, S. Darai, K. Wasser, S. Delorme, and J. Griebel, “Microcirculation and microvasculature in breast tumors: pharmacokinetic analysis of dynamic MR image series,” Magn. Reson. Med. 52(2), 420–429 (2004). [PubMed]
30. E. E. Uzgiris, “Tumor microvasculature: endothelial leakiness and endothelial pore size distribution in a breast cancer model,” Breast Cancer: Basic and Clinical Research 1, 83–90 (2008).
31. H. Daldrup, D. M. Shames, M. Wendland, Y. Okuhata, T. M. Link, W. Rosenau, Y. Lu, and R. C. Brasch, “Correlation of dynamic contrast-enhanced MR imaging with histologic tumor grade: comparison of macromolecular and small-molecular contrast media,” AJR Am. J. Roentgenol. 171(4), 941–949 (1998). [PubMed]
32. C. C. Michel and F. E. Curry, “Microvascular permeability,” Physiol. Rev. 79(3), 703–761 (1999). [PubMed]
33. M. R. Dreher, W. Liu, C. R. Michelich, M. W. Dewhirst, F. Yuan, and A. Chilkoti, “Tumor vascular permeability, accumulation, and penetration of macromolecular drug carriers,” J. Natl. Cancer Inst. 98(5), 335–344 (2006). [PubMed]
34. H. E. Daldrup-Link and R. C. Brasch, “Macromolecular contrast agents for MR mammography: current status,” Eur. Radiol. 13(2), 354–365 (2003). [PubMed]
35. P. Ott and R. A. Weisiger, “Nontraditional effects of protein binding and hematocrit on uptake of indocyanine green by perfused rat liver,” Am. J. Physiol. 273(1 Pt 1), G227–G238 (1997). [PubMed]
36. S. Keiding, P. Ott, and L. Bass, “Enhancement of unbound clearance of ICG by plasma proteins, demonstrated in human subjects and interpreted without assumption of facilitating structures,” J. Hepatol. 19(3), 327–344 (1993). [PubMed]
37. K. Sauda, T. Imasaka, and N. Ishibashi, “Determination of protein in human serum by high-performance liquid chromatography with semiconductor laser fluorometric detection,” Anal. Chem. 58(13), 2649–2653 (1986). [PubMed]
38. S. Yoneya, T. Saito, Y. Komatsu, I. Koyama, K. Takahashi, and J. Duvoll-Young, “Binding properties of indocyanine green in human blood,” Invest. Ophthalmol. Vis. Sci. 39(7), 1286–1290 (1998). [PubMed]
39. Z. M. Bhujwalla, D. Artemov, K. Natarajan, E. Ackerstaff, and M. Solaiyappan, “Vascular differences detected by MRI for metastatic versus nonmetastatic breast and prostate cancer xenografts,” Neoplasia 3(2), 143–153 (2001). [PubMed]
40. B. Alacam, B. Yazici, X. Intes, S. Nioka, and B. Chance, “Pharmacokinetic-rate images of indocyanine green for breast tumors using near-infrared optical methods,” Phys. Med. Biol. 53(4), 837–859 (2008). [PubMed]
41. Y. Matsumura and H. Maeda, “A new concept for macromolecular therapeutics in cancer chemotherapy: mechanism of tumoritropic accumulation of proteins and the antitumor agent smancs,” Cancer Res. 46(12 Pt 1), 6387–6392 (1986). [PubMed]
42. U. Schilling, E. A. Friedrich, H. Sinn, H. H. Schrenk, J. H. Clorius, and W. Maier-Borst, “Design of compounds having enhanced tumor uptake, using serum albumin as a carrier – Part II. In vivo studies,” Nucl. Med. Biol. 19(6), 685–695 (1992).
43. R. Duncan, “Polymer conjugates as anticancer nanomedicines,” Nat. Rev. Cancer 6(9), 688–701 (2006). [PubMed]
44. M. Gurfinkel, A. B. Thompson, W. Ralston, T. L. Troy, A. L. Moore, T. A. Moore, J. D. Gust, D. Tatman, J. S. Reynolds, B. Muggenburg, K. Nikula, R. Pandey, R. H. Mayer, D. J. Hawrysz, and E. M. Sevick-Muraca, “Pharmacokinetics of ICG and HPPH-car for the detection of normal and tumor tissue using fluorescence, near-infrared reflectance imaging: a case study,” Photochem. Photobiol. 72(1), 94–102 (2000). [PubMed]
45. E. Brown, J. Hopper Jr, J. L. Hodges Jr, B. Bradley, R. Wennesland, and H. Yamauchi, “Red cell, plasma, and blood volume in the healthy women measured by radiochromium cell-labeling and hematocrit,” J. Clin. Invest. 41(12), 2182–2190 (1962). [PubMed]
46. C. B. Wilson, A. A. Lammertsma, C. G. McKenzie, K. Sikora, and T. Jones, “Measurements of blood flow and exchanging water space in breast tumors using positron emission tomography: a rapid and noninvasive dynamic method,” Cancer Res. 52(6), 1592–1597 (1992). [PubMed]
47. P. Vaupel and M. Höckel, “Blood supply, oxygenation status and metabolic micromilieu of breast cancers: characterization and therapeutic relevance,” Int. J. Oncol. 17(5), 869–879 (2000) (review). [PubMed]
48. D. Grosenick, H. Wabnitz, K. Th. Moesta, J. Mucke, P. M. Schlag, and H. Rinneberg, “Time-domain scanning optical mammography: II. Optical properties and tissue parameters of 87 carcinomas,” Phys. Med. Biol. 50(11), 2451–2468 (2005). [PubMed]
49. P. S. Tofts and B. A. Berkowitz, “Measurement of capillary permeability from the Gd enhancement curve: a comparison of bolus and constant infusion injection methods,” Magn. Reson. Imaging 12(1), 81–91 (1994). [PubMed]