Open Access
Add to BibSonomyAbstract
Near infrared dynamic diffuse optical tomography measurements of breast hemodynamics during fractional mammographic compression offer a novel contrast mechanism for detecting breast cancer and monitoring chemotherapy. Tissue viscoelastic relaxation during the compression period leads to a slow reduction in the compression force and reveals biomechanical and metabolic differences between healthy and lesion tissue. We measured both the absolute values and the temporal evolution of hemoglobin concentration during 25-35 N of compression for 22 stage II and III breast cancer patients scheduled to undergo neoadjuvant chemotherapy. 17 patients were included in the group analysis (average tumor size 3.2 cm, range: 1.3-5.7 cm). We observed a statistically significant differential decrease in total and oxy-hemoglobin, as well as in hemoglobin oxygen saturation in tumor areas vs. healthy tissue, as early as 30 seconds into the compression period. The hemodynamic contrast is likely driven by the higher tumor stiffness and different viscoelastic relaxation rate, as well as the higher tumor oxygen metabolism rate.
© 2013 Optical Society of America
1. Introduction
Breast cancer remains a disease with significant societal impact. In 2013, it is likely that over 230,000 women will be diagnosed with breast cancer in the United States, while nearly 40,000 will die from it [1]. Mammography is the gold standard for breast cancer screening, yet it suffers from reduced sensitivity in women with dense breasts [2], as well as low specificity that results in unnecessary biopsies 75-85% of the time. Novel techniques such as 3D digital breast tomosynthesis have resulted in increased sensitivity, but only small improvements in specificity [2,3]. Near infrared spectroscopy (NIRS) and diffuse optical tomography (DOT) have emerged as promising functional imaging techniques sensitive to cancer-induced pathological changes in tissue related to angiogenesis and flow/metabolism imbalances. Encouraging results have been reported for both cancer diagnosis and neoadjuvant chemotherapy monitoring [4–7].
The vast majority of published breast cancer optical imaging studies have focused on obtaining snapshots of the concentrations of tissue chromophores such as the tissue-average total hemoglobin concentration (HbT), hemoglobin oxygen saturation (SO2) and the water and lipid fractions, as well as the scattering cross-section. However, recent advances in instrumentation and computing power have enabled a new approach based on the dynamic imaging of time-resolved changes in breast tissue properties in response to external mechanical [8–12] or inhaled gas stimulation [13]. In particular, our group has investigated the breast tissue hemodynamics induced by fractional mammographic-like compression (we used approximately one third of the typical mammographic compression force). Such hemodynamic changes are governed by the interplay of biomechanical properties and metabolic activity. Both stiffness as well as oxygen consumption and blood flow [14,15] are known to be elevated in malignant tumors [16,17], motivating the development of a non-invasive imaging technique sensitive to these parameters. In two previous publications, we reported preliminary data in a group of healthy volunteers showing that tissue viscoelastic relaxation during compression modulates both blood volume and hemoglobin oxygen saturation. We also proposed a model for estimating volumetric blood flow and oxygen consumption from the time-course of the HbT and SO2 variation [12,18].
In this paper, we report results from a group of 17 patients with stage II and III breast cancer (invasive breast cancer that has not yet spread to other parts of the body). The patients were scanned using our recently developed dynamic diffuse optical tomography system. To the best of our knowledge, this is the largest report so far of tomographic quantification of compression-induced hemodynamics in breast cancer. We used information from both clinical MRI/x-ray scans and optical image features to define regions of interest for the tumor and healthy tissue areas, respectively, and we’ve been able to observe a significant difference in dynamic characteristics between these two domains. These results add to the increasing body of evidence demonstrating the clinical utility of optical breast cancer imaging.
2. Methods
2.1. System description
Our breast compression dynamic optical imaging system consists of a high temporal resolution transmission mode diffuse optical tomography system integrated with a previously reported compression setup [12]. The breast is placed between two parallel plates, which apply a repeated step compression/release to the breast. As detailed in Section 2.4, data is acquired continuously, and operation is controlled by a personal computer running custom software that ensures synchronization of the various optical and mechanical components (a simple scripting language is used for flexibility in defining the experimental protocol). Figure 1(a) shows the dynamic optical imaging setup as mounted on a cart that can be wheeled into an examination room, while Fig. 1(b) offers a detail view of the compression and fiber probe sub-assembly.
Fig. 1 Dynamic optical imaging instrumentation (a) Dynamic optical imaging clinical cart. (b) Detail of compression mechanism; (c) Source and (d) detector fiber co-ordinates
2.2 Optical imaging instrumentation
The optical imaging setup uses a hybrid design, combining a continuous-wave (CW) imager for spatial coverage and high acquisition rate with a frequency domain (FD) near infrared spectroscopy system for absolute quantification of tissue optical properties.
The frequency domain component consists of an ISS Imagent (model 96208, ISS Inc., Champaign, IL). Eight time-multiplexed laser diodes at 635, 670, 690, 752, 758, 810 and 830 nm modulated at 110 MHz injects light in the breast at a single central location near the chest wall using an optical fiber bundle. Three red-enhanced photomultiplier tube detectors (Hamamatsu R928 PMT) collect the back scattered light through 2.5 mm optical fiber bundles. The amplitude and phase parameters are extracted by in-phase/quadrature (I/Q) demodulation. The frequency modulated system provides absolute optical properties and is used to obtain the baseline breast bulk hemoglobin concentration and optical scattering that form the initial homogeneous guess for the 3D tomographic reconstruction based on continuous wave measurements.
The continuous wave component consists of a TechEn CW6 imager augmented by a Supplemental Source Device (SSD) (TechEn Inc., Milford, MA). The CW6 offers 32 lasers split equally among 690 and 830 nm as well as 32 avalanche photodiode (APD) detectors. The lasers are modulated at 32 individual frequencies between ~6-12 kHz allowing full simultaneous detection [19]. The SSD adds 32 lasers at additional wavelengths (equally divided between 780, 808 and 850 nm) with each laser sharing a modulation frequency with a corresponding laser in the main CW6 box. Main/supplemental laser sources were illuminated in an interleaved fashion, switching once per second. Even in conjunction with the SSD, the system is able to record all 2048 source-detectors pairs at over 10 Hz, while maintaining high signal to noise ratio (>90dB) and linear dynamic range(>60dB). Since expected breast hemodynamics are relatively slow, we averaged the CW data into 2 second windows for optimal signal quality.
2.3 Compression system and optical probes
The compression setup uses two horizontal parallel plates (equivalent to mammographic cranio-caudal (CC) compression). The lower plate is attached to the instrument cart, while the upper plate is mounted on a computer controlled translation stage by means of two S-shaped strain gauges and can apply 0-55N of compression over a range of motion in excess of 10 cm. Optical fibers are integrated into both plates. The 64 CW source fibers, as well as the FD-NIRS source and three detection bundles (FD source-detectors separations of 11.2, 16.8, 22.4 mm) are inserted in the lower plate, while 30 2.5 mm fiber bundles with a 90 degree bent end are mounted into the upper plate to collect light transmitted through the breast and deliver it to the CW detectors. Both the upper and the lower fiber arrays cover an approximately 12x7 cm half elliptical area as shown in Fig. 1(c), 1(d). The FD source power at the probe is ~2mW, while the CW source fibers deliver ~10 mW. A pressure mapping system (Tekscan I-Scan) with a Tekscan 5250 flat 10”X10” 44x44 element sensor is mounted on the lower fixed plate, and is used to monitor the breast contact patch and spatial distribution of forces during the optical measurements. Fibers in the lower plate are aligned with the transparent windows between the Tekscan sensor rows and columns to allow simultaneous optical imaging and pressure monitoring.
It is important to note that compression plate movement ends once the strain gauges sense that the desired compression force has been applied. The plates are then held in a fixed position until the end of the compression cycle, without any attempts to maintain the force level. This is done primarily to ensure a stable platform for optical tomography as any change in plate separation must be accurately accounted for, lest it should appear as an absorption variation artifact (note that even 1 mm of unaccounted motion would lead to changes in transmitted light intensity comparable to or higher than the hemodynamics driven changes being measured). The fixed plate position compression also mimics the mammography procedures and offers a way to understand the expected impact of compression for combined optical/x-ray breast imaging systems (also being developed by our group [20,21]).
2.4 Measurement protocol
All measurements were conducted under a protocol approved by the Dana Farber/Harvard Cancer Center Institutional Review Board (IRB). We have enrolled 22 female subjects diagnosed with breast cancer and scheduled to undergo neoadjuvant chemotherapy. Women with open wounds on the breast, breast implants, or breast biopsies within the previous 10 days were excluded.
Each measurement session began by placing the patient’s breast on the lower compression plate, with her chest wall resting against the plate’s edge, and the patient chair height was adjusted to maximize breast insertion. The PMT gain on the FD system was adjusted to ensure operation in the linear range, after which the breast is removed, and a separate phantom is used to calibrate the FD system. Then the patient’s breast is re-positioned on the lower compression plate, as closely as possible to the initial position and the upper plate is brought down until it lightly touches the breast. A test compression is performed to determine optode coverage and verify patient’s compression tolerance. After allowing the tissue to recover from the test compression for 2 minutes, the main measurement procedure is started. As delineated in Fig. 2, baseline optical properties (FD measurements) are recorded for 2 seconds before, during, and for 2 seconds after the application of 25 to 35 N of compression force. Then the CW imaging data is acquired for 120 seconds while under compression. After that, another FD measurement is performed for 5 seconds and the compression is released. The entire compression cycle was repeated a total of 3 times. In a subset of 16 patients, we added a partial release step at the end of the 3rd cycle, by moving the compression plate 3 mm upwards. This resulted in an approximate halving of the compression force, and we followed the tissue hemodynamics for another 90 seconds. The 25 to 35 N is approximately one third of typical mammographic compression and is comparatively comfortable for the study subjects (as reported by the subjects themselves). The Tekscan pressure monitoring system acquires data continuously during the entire breast measurement. The control software also monitors and stores inter-plate distance using a pair of linear encoders and compression force using the two strain gauges. At the end of the experiment another CW transmission measurement is taken on the tissue-like calibration phantom to compute coupling correction factors used in obtaining absolute images of hemoglobin concentration [22]. These measurements also serve as a quality check to detect deterioration of fibers, source lasers or photo detectors.
Fig. 2 Schematic of the measurement protocol. The Tekscan pressure mapping system acquires data continuously during the entire measurement session (for each breast), while each the FD system is used to capture absolute optical properties from 2 seconds before to 2 seconds after the compression is applied in each cycle, and also for 5 seconds at end of each cycle, just before compression is released. The CW imager acquires data for 120 seconds during the compression, from 2 seconds after the upper plate stopped moving until 5 seconds before compression is released.
2.5 Data processing and image reconstruction
The first step after collecting optical and compression data is to use the coordinate mapping between the Tekscan pressure sensors and the optical probe coordinates to select the optodes covered by the breast. From that, an unstructured tetrahedral mesh is generated using a simple algorithm described in Ref. 20, and a finite element (FEM) implementation of the diffusion approximation is employed to model light transport [20]. A spectrally constrained reconstruction approach [23,24] is implemented to recover oxy (HbO) and deoxy hemoglobin (HbR) concentration images directly from the multi-wavelength measurements. First, bulk HbO and HbR concentration as well as scattering amplitude and power are estimated from the frequency domain data. The bulk concentration estimates are then refined by combining the FD scattering data with the transmission CW measurements for more representative whole breast values. We assumed 30% water and 50% lipid fractions, respectively [22,25]. Finally, an iterative tomographic reconstruction is obtained from the CW data starting from a uniform initial guess based on the bulk hemoglobin concentration values. For this purpose, the forward solution at all valid source and detectors are used to construct the Jacobian matrix using the adjoint method [20]. A dual-mesh scheme, a forward mesh for diffusion modeling and a separate reconstruction mesh to represent the optical properties [26] was used for improved computation efficiency. Hemoglobin concentration values are reconstructed by solving a regularized nonlinear optimization problem using 9 Gauss Newton iterations (we found 9 iterations to be sufficient to obtain a stable solution in all our image reconstructions). Scattering was fixed at the FD-derived bulk value.
2.6 Image analysis
Tomographic distributions of HbO and HbR are reconstructed for every 2 seconds of measurements, for a total of 60 frames in each compression cycle. Relative changes in all optical parameters were also computed by subtracting the first frame as a reference, and are denoted with a Δ below. We use the clinical imaging reports and the radiologist image annotations to identify the tumor location on the patient’s X-ray or MRI scans. A tumor region-of-interest (ROI) is defined based on the radiological images, as well as the expected optical contrast (increased total hemoglobin (HbT) concentration, and a persistent dynamic HbT reduction). We primarily consider clinical imaging in defining the tumor ROI, followed by the relative change in HbT, and finally the absolute HbT contrast (we do this because our system is not optimized for absolute imaging, as discussed later). In addition to the tumor ROI, another ROI is defined to encompass the rest of the breast up to the edge of the optical imaging sensitivity area (defined as the outer limits of the fiber array selected for tomographic reconstruction). For either ROI we analyzed the absolute values and the time-resolved changes (Δ) in HbT, HbO, HbR and SO2 (computed as HbO/HbT). While considering the optical contrast in the ROI selection has the potential to introduce bias, this is the best method available for standalone optical imaging, and has been used by the majority of the groups that have conducted optical breast imaging. If the location of the expected optical contrast does not match the position predicted from MRI/X-ray within 1.5 cm, we have instead used the position suggested by the radiological images using the relative position with respect to the tissue boundaries as a reference (irrespective of the optical features at that location).
2.7 Statistical analysis
To assess whether any of the optical parameters is significantly different between the tumor region and the surrounding healthy tissues, we performed a two-tailed, paired t-test between the corresponding value of each optical parameters in the tumor and healthy ROI across the subject group (normality of the data at the 95% confidence level was verified using the Jarque-Bera test as implemented by the Matlab jbtest function). A description of the specific quantities tested is given in the Results section. To assess the diagnostic value of a given parameter, we pooled all tumor and healthy values for that parameter, and computed the true and false positive rates (TPR/FPR) for a range of parameter thresholds between the minimum and maximum value. We then assessed the area under the receiver operating characteristic (ROC) curve from the scatter plot of TPR vs. FPR as an overall metric of performance.
3. Results
Among the 22 stage II and III breast cancer patients recruited between January 2010 and August 2012, four subjects had tumors out of the optical field of view (either too close to the chest wall, or outside the optical coverage area for large breasts). Another had a large diffuse tumor according to the dynamic contrast enhanced (DCE)-MRI scan, which showed scattered enhancement throughout the breast – thus we could not define separate tumor and healthy ROIs. The data from the remaining 17 cases are reported below (a summary of patient characteristics is given in Table 1). The lesions ranged from 1.3 to 5.7 cm, with an average of 3.2 cm and a standar deviation of 1.3 cm.
Table 1. Summary of Patient Characteristics
Figure 3(a) shows a CC (transversal) maximum intensity projection of a 3D Gd-DTPA dynamic contrast enhanced MRI scan for both breasts of a 39 year old patient. The MR images show a 3.4x2.9x3.2 cm enhancing mass in the right breast approximately at 12 o’clock, that was proven by biopsy to be an invasive ductal carcinoma. Figures 3(b) and 3(c) show the reconstructed absolute HbT concentration image in the same view, and the corresponding changes in HbT at t = 90s after the compression is established. The tumor is characterized by a higher total hemoglobin concentration value than the rest of the tissue, coupled with an HbT dynamic decrease. We use both of these characteristics in conjunction with guidance from clinical imaging to assign the tumor ROI to minimize the impact of breast physiological variability in our ROI assignment process.
Fig. 3 (a) Maximum intensity projection of Gd-DTPA Dynamic Contrast Enhancement (DCE) MRI scan for both breasts of a 39 year old patient. MRI images show a 3.4x2.9x3.2 cm enhancing mass in the right breast at 12 o’clock; (b) Total hemoglobin concentration CC slice, half way between the compression plates, from the lesion breast of the same patient; (c) Corresponding total hemoglobin differential image (t = 60s).
Table 2 shows a summary of the mean absolute values of optical parameters and their corresponding standard deviations, in the tumor and healthy tissue, across the entire patient group. Also, tumor to healthy ratios were computed for each individual subject, and the means and standard deviations of the ratios are also reported in the table. These measurements show that the mean absolute HbT, HbO and HbR are higher in the tumor region than in the healthy region of the same breast. However, there is no significant SO2 contrast. While there is some variation in the hemoglobin concentrations across compressions, the tumor/healthy ratio values are fairly stable. Table 3 shows the results of a paired two-sided t-test performed between the values of absolute optical parameters in the tumor vs. the healthy region. HbT, HbO, are significantly higher in tumors at the p<0.01 level, while HbR is significantly higher at the p<0.05 level. No significant difference was found in SO2 values.
Table 2. Summary of absolute optical properties of breast tissue across compression cycles
Table 3. p-values for the statistical analysis of tumor vs. healthy contrast in absolute HbT/HbO/HbR/SO2
Figure 4 shows example dynamic measurements in one of our subjects. Figure 4(a) contains the effective compression force measurement from the strain gauges that support the upper compression plate. This is meant to illustrate the viscoelastic relaxation undergone by the tissue during the compression period, which is one of the primary drivers of the hemodynamic effects being investigated. Figures 4(b) and 4(c) display changes in HbT and SO2 respectively during the compression, referenced to the initial state. The tumor ROI timecourse is shown as solid lines (colors indicate the compression cycle), while the healthy ROI is shown as dashed lines. While there is significant variation among the compression cycles, in all cases the tumor ROI shows a notable HbT decrease followed by a limited blood volume recovery, while the healthy ROI has a notable increasing trend from 30 to 120 sec. For SO2 the differential relationship is similar; while both ROIs show a decrease, a stronger decrease is observed in the tumor area. Figure 5 displays a representative scatter plot of tumor vs. normal tissue changes in (a) HbT and (b) SO2 across the patient group at 60 seconds into the first compression cycle. The 1:1 ratio line is included to help interpret the data. While a wide range of responses are present, all but two patients exhibit the characteristics seen in the Fig. 4 example, with the normal tissue showing a relatively more positive HbT and SO2 trend than the corresponding lesion.
Fig. 4 Example dynamic measurements (a) strain gauge force data for the entire measurement on the tumor breast; gray areas highlight the three main compression periods; (b,c) HbT/SO2 variation vs. initial state for the tumor (solid) and healthy (dashed) ROIs over the three compression cycles (blue, green, red for cycles 1, 2, and 3)
Fig. 5 Scatter plot of changes in (a) HbT and (b) SO2 in the tumor area vs. the surrounding normal tissue across the patient group (one symbol per patient) at t = 60 seconds into the first compression cycle.
Figure 6 shows ΔHbT, ΔHbO, ΔHbR, and ΔSO2 referenced to the initial state in both tumor and healthy tissue ROIs averaged for the 17 patients that were analyzed. The group averages display the same trends seen in the Fig. 4 single subject example, with predominantly decreasing tendencies for HbT and HbO in the tumor, in contrast with a generally increasing average trend in the healthy area. Less but still significant differentiation is seen in the evolution of SO2, with more pronounced decreases in the tumor than in surrounding tissue. HbR changes are small though a weak increasing tendency is seen, and are fairly similar in both the lesion and the normal tissue. The group average HbR timecourse in tumors appear slightly below the healthy tissue one, but the difference is not significant (see next paragraph and Table 4).
Fig. 6 Group averages of changes in HbT, HbO, HbR, and SO2, respectively, in the tumor (blue) and healthy (red) regions over the three compression cycles. Error bars (displayed only every 4th timepoint for clarity) show the standard error across the group.
Table 4. p-values for the statistical significance test of differential hemodynamics in lesions vs normal tissue. Significant values (p<0.05) are shown in bold font.
To assess whether differences in the lesion vs. normal timecourses are statistically significant, we performed a two-tail, paired t-test between the values of each of the optical parameters in the tumor and healthy ROIs across the subject group, at t = 30, 60, and 90s, respectively after the beginning of compression. Table 4 summarizes these results for each of the three compression cycles. The variations in lesion vs. normal HbT and HbO are significantly different at the p<0.01 level for all cycles at all timepoints, except for t = 90s into the first cycle, where the difference is significant at the p<0.05 level. The statistical significance of tumor vs. normal SO2 changes is highest in the second and third cycle especially at t = 60 and 90s in the 3rd cycle (p<0.01), and remains significant at the p<0.05 level for cycle 1. HbR differential changes are not significant at any of the cycles or timepoints, except at t = 30 s in the first cycle. The data suggests that HbT or HbO differential changes at 30 seconds in the first cycle strongly correlate with the presence of a malignant tumor. Consequently, we developed a simple classification scheme that assigns tissue as either normal or malignant depending on the amount of HbT/HbO change. The best results were obtained by using a threshold for ΔHbT at 30 seconds into the first cycle, as described by the receiver operating characteristic (ROC) curve in Fig. 7. The area under the curve (AUC) is 0.79, with a best case of 88% sensitivity and 70% specificity.
Fig. 7 ROC curve derived by setting a sliding ΔHbT threshold at t = 30s into cycle 1, to classify tissue type.
Figure 8 displays group average hemodynamic changes induced by a partial release of the compression force at the end of the 3rd cycle (this data was available and evaluable only in a subset of 10 patients). The 3 mm upward motion of the compression plate resulted in a significant reduction in compression force, down to 25-50% of the level at the end of the 3rd cycle. As can be seen in Fig. 4(a) in the portion of the force measurements just after the third shaded period, after the immediate pressure drop due to plate motion, the tissue “catches up” with the plate after a viscoelastic delay, resulting in a slight increase of the effective force during this period, but still remaining well below the force at the end of the 3rd cycle. The normal and lesion areas had very similar behavior, with a temporary HbT increase, a more persistent HbO increase, a decrease in HbR, and a ramp up followed by a plateau in SO2. These appear to correspond to a washout dynamic due the inflow of arterial blood as compression is reduced. However, there was no statistically significant difference between the tumors and their corresponding healthy surrounding areas.
Fig. 8 Hemodynamic response to partial compression release (tumor area in blue, normal tissue in red).
4. Discussion
The goal of this study has been to characterize biomechanical and metabolic contrast in breast cancer lesions versus the surrounding healthy tissues by monitoring variations in hemoglobin concentration and oxygen saturation following fractional mammographic compression. In this group of fairly large tumors, the primary feature we’ve seen is a marked difference in the total hemoglobin, oxy-hemoglobin and hemoglobin oxygenation timecourses in the tumor area versus the rest of the breast. As the effective compression force decreases due to tissue viscoelastic relaxation, blood begins to return to tissue, and we see a general increasing trend in the normal areas. The tumor on the other hand shows a persistent decrease in blood volume with a weak reperfusion towards the end of the respective compression periods. There was little variation of these characteristics over the three compression cycles, and as shown by the statistical tests, significant differences are seen at least as early as 30 seconds into the first compression cycle. This is encouraging because a diagnostic test based on this dynamic compression procedure could be conducted quite quickly, without the need for repeated compression and lengthy monitoring.
This differential behavior may be due to the higher stiffness and longer viscoelastic relaxation time of the tumor tissue. We hypothesize that during the application of compression (prior to our imaging period) healthy tissue is progressively “protected” by the stiffer tumor area which becomes more and more load bearing. Once the plate motion is completed, the tumor experiences a disproportionate amount of the effective compression force and also takes longer (possible on the order of 20-40 seconds, as shown by the bending of the ΔHbT curve) to come into mechanical equilibrium with its surroundings and begin the relaxation/reperfusion process. Both behaviors have been suggested by previous publications. Darling et al., using simulations and phantoms studies noted that higher forces can be transmitted at depth in the presence of stiff inclusions [27], while Al Abdi et al. observed similar relaxation dynamics on a patient examined using a dynamic optical tomography system equipped with articulated actuators that could apply various pressure patterns to the breast [8]. It is interesting to note that the majority of blood volume change (as seen through total hemoglobin) appears to be driven by the influx/efflux of oxy-hemoglobin. This suggests changes in the arterial compartment dominate our measurements. However, the faster decrease in SO2 in the tumor area also suggests a higher tumor oxygen metabolic rate. One would expect to see an increased rate of HbR accumulation due to the higher metabolic rate expected in the tumor area. However, the lack of statistically significant trends could be explained by the competing biomechanical and metabolic factors: while the blood volume reduction due to the stress/strain in the tumor area is driving a corresponding reduction in HbR, the higher tumor metabolism is producing HbR at a higher rate vs. normal tissue. Unfortunately, these two factors nearly cancel each other, leading to the lack of observed tumor vs normal contrast. Of note, a study published using a dynamic breast compression imaging system that uses 640 nm light reported that the application of slight compression increases optical absorption at this wavelength in malignant breast tumors [10]. This wavelength is primarily impacted by the deoxy-Hb concentration (though the authors do not draw any explicit conclusions regarding this link). However, the compression protocol is different, with imaging done with respect to the uncompressed baseline, thus the results are not directly comparable with our study.
While not a focus of this study, we have also observed a slight hyperemic effect with repeated compression, as well as increased blood volume in malignant tumors in comparison to normal tissue, though with little contrast in hemoglobin saturation. This trend matches well numerous published studies that used optical spectroscopy or tomography to characterize breast lesions [4,5]. Though still statistically significant, the contrast values we report are lower than most of the other studies. The necrotic core present in large tumors may contribute to this reduced contrast but it is more likely that our results are influenced by shortcomings in our instrumentation that impact absolute imaging. Absolute optical images are highly sensitive to calibration/tissue contact and to systematic errors in light transport modeling, for example due to incorrect modeling of the air-tissue boundary. We rely on the Tekscan-derived breast contact outline on the lower compression plate to estimate the breast shape. This method is unlikely to give an accurate result for the upper surface of the breast, thus potentially leading to spatially varying systematic deviations in reconstructed absolute hemoglobin concentration, which cannot be compensated by using healthy tissue in the same breast as a reference. At the same time, difference (relative) imaging is self-calibrated, and is thus much less sensitive to both fiber contact and inaccurate light transport modeling, as these errors tend to cancel out. This is seen for example in Ref. 8, where good results are obtained even though a single finite element mesh shape is used to analyze time resolved relative optical imaging data from multiple subjects.
Finally, the investigation of step release hemodynamics was inconclusive. However, this remains an area of interest, as the tumor vasculature is functionally defective and it has been hypothesized that it is not capable of a vasodilatory response, unlike the vessels in normal tissues [28,29]. We were expecting a significantly higher reperfusion in normal areas vs. the tumor. A possible explanation for the lack of such a response may be that by the end of the 3rd cycle the compression pressure was substantially reduced and there was not enough hypoxia to generate a vasodilatory response that would enhance reperfusion in normal tissue. Perhaps a step partial decompression from a higher starting pressure would show a larger difference in hemodynamic response. Another explanatory factor is that Fig. 7 shows absolute changes in optical parameters instead of looking at variation percentages. Since the tumor baseline hemoglobin concentration is higher, the same absolute hemoglobin change in tumor vs. healthy tissue would indeed translate into a smaller percentage change in the tumor. However, as mentioned in the previous paragraph, we measured only an ~10% higher hemoglobin concentration in the tumor vs normal areas, thus the percentage change timecourses would differ by 10% as well. This is unlikely to reach statistical significance, but remains a focus of future study.
A caveat of our method is the need to use optical contrast to define the tumor location due to the lack of co-registered structural imaging. While we do our best to match the relative tumor location with respect to the tissue boundary between optical images and clinical MRI or mammography scans (see Section 2.6 for the detailed procedure), this remains a potential source of bias shared with most standalone breast optical imaging investigations. An effective solution is the use of co-registered multi-modal methods, such as the combination of optical tomography with MRI or mammography and we intend to use this approach for future work.
5. Conclusion
We have shown that compression-induced hemodynamics can differentiate malignant breast tumors from the surrounding tissue. The main feature observed is a persistent blood volume reduction in the tumor area in contrast with a slow re-perfusion in the normal tissue. This characteristic is statistically significant, and can be detected using a single compression cycle with 60 seconds of follow-up during compression.
These encouraging results suggest dynamic optical imaging can be a useful addition to the range of optical breast imaging techniques, and motivates the development of a next-generation system with an improved breast compression interface (MLO compression for increase coverage) and curved compression plates for more uniform pressure distribution. It is expected that hemodynamic biomarkers revealed through breast compression will become useful tools for both cancer detection and chemotherapy monitoring.
Acknowledgments
We would like to acknowledge funding from NIH through grant R01-CA97305, R01-CA 142575, U54-CA105480, K99/R00EB011889; and from the Komen Breast Cancer Foundation grant KG200021. The authors would like to thank Ms. Nancy Nagda for coordinating patient scheduling and the staff of the MGH Center for Breast Cancer for their support.
References and links
1. Cancer Facts & Figs. 2013 (American Cancer Society, Atlanta, 2013).
2. E. D. Pisano, C. Gatsonis, E. Hendrick, M. Yaffe, J. K. Baum, S. Acharyya, E. F. Conant, L. L. Fajardo, L. Bassett, C. D’Orsi, R. Jong, M. Rebner, and Digital Mammographic Imaging Screening Trial (DMIST) Investigators Group, “Diagnostic performance of digital versus film mammography for breast-cancer screening,” N. Engl. J. Med. 353(17), 1773–1783 (2005). [CrossRef] [PubMed]
3. M. L. Zuley, A. I. Bandos, M. A. Ganott, J. H. Sumkin, A. E. Kelly, V. J. Catullo, G. Y. Rathfon, A. H. Lu, and D. Gur, “Digital breast tomosynthesis versus supplemental diagnostic mammographic views for evaluation of noncalcified breast lesions,” Radiology 266(1), 89–95 (2013). [CrossRef] [PubMed]
4. S. Fantini and A. Sassaroli, “Near-infrared optical mammography for breast cancer detection with intrinsic contrast,” Ann. Biomed. Eng. 40(2), 398–407 (2012). [CrossRef] [PubMed]
5. P. Taroni, A. Pifferi, G. Quarto, L. Spinelli, A. Torricelli, F. Abbate, A. Villa, N. Balestreri, S. Menna, E. Cassano, and R. Cubeddu, “Noninvasive assessment f breast cancer risk using time-resolved diffuse optical spectroscopy,” J. Biomed. Opt. 15(6), 060501 (2010). [CrossRef] [PubMed]
6. R. Choe and T. Durduran, “Diffuse Optical Monitoring of the Neoadjuvant Breast Cancer Therapy,” IEEE J. Sel. Top. Quantum Electron. 18(4), 1367–1386 (2012). [CrossRef] [PubMed]
7. A. E. Cerussi, V. W. Tanamai, D. Hsiang, J. Butler, R. S. Mehta, and B. J. Tromberg, “Diffuse optical spectroscopic imaging correlates with final pathological response in breast cancer neoadjuvant chemotherapy,” Philos Trans A Math Phys Eng. Sci. 369, 4512–4530 (2011).
8. R. Al abdi, H. L. Graber, Y. Xu, and R. L. Barbour, “Optomechanical imaging system for breast cancer detection,” J. Opt. Soc. Am. A 28(12), 2473–2493 (2011). [CrossRef]
9. R. X. Xu, D. C. Young, J. J. Mao, and S. P. Povoski, “A prospective pilot clinical trial evaluating the utility of a dynamic near-infrared imaging device for characterizing suspicious breast lesions,” Breast Cancer Res. 9(6), R88 (2007). [CrossRef] [PubMed]
10. L. S. Fournier, D. Vanel, A. Athanasiou, W. Gatzemeier, I. V. Masuykov, A. R. Padhani, C. Dromain, K. Galetti, R. Sigal, A. Costa, and C. Balleyguier, “Dynamic optical breast imaging: a novel technique to detect and characterize tumor vessels,” Eur. J. Radiol. 69(1), 43–49 (2009). [CrossRef] [PubMed]
11. S. Jiang, B. W. Pogue, A. M. Laughney, C. A. Kogel, and K. D. Paulsen, “Measurement of pressure-displacement kinetics of hemoglobin in normal breast tissue with near-infrared spectral imaging,” Appl. Opt. 48(10), D130–D136 (2009). [CrossRef] [PubMed]
12. 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). [CrossRef] [PubMed]
13. M. L. Flexman, M. A. Khalil, R. Al Abdi, H. K. Kim, C. J. Fong, E. Desperito, D. L. Hershman, R. L. Barbour, and A. H. Hielscher, “Digital optical tomography system for dynamic breast imaging,” J. Biomed. Opt. 16(7), 076014 (2011). [CrossRef] [PubMed]
14. R. P. Beaney, A. A. Lammertsma, T. Jones, C. G. McKenzie, and K. E. Halnan, “Positron emission tomography for in-vivo measurement of regional blood flow, oxygen utilisation, and blood volume in patients with breast carcinoma,” Lancet 323(8369), 131–134 (1984). [CrossRef] [PubMed]
15. 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]
16. J. M. Chang, I. A. Park, S. H. Lee, W. H. Kim, M. S. Bae, H. R. Koo, A. Yi, S. J. Kim, N. Cho, and W. K. Moon, “Stiffness of tumours measured by shear-wave elastography correlated with subtypes of breast cancer,” Eur. Radiol. 23(9), 2450–2458 (2013). [CrossRef] [PubMed]
17. M. Sridhar and M. F. Insana, “Ultrasonic measurements of breast viscoelasticity,” Med. Phys. 34(12), 4757–4767 (2007). [CrossRef] [PubMed]
18. S. A. Carp, T. Kauffman, Q. Fang, E. Rafferty, R. Moore, D. Kopans, and D. Boas, “Compression-induced changes in the physiological state of the breast as observed through frequency domain photon migration measurements,” J. Biomed. Opt. 11(6), 064016 (2006). [CrossRef] [PubMed]
19. D. K. Joseph, T. J. Huppert, M. A. Franceschini, and D. A. Boas, “Diffuse optical tomography system to image brain activation with improved spatial resolution and validation with functional magnetic resonance imaging,” Appl. Opt. 45(31), 8142–8151 (2006). [CrossRef] [PubMed]
20. Q. Fang, S. A. Carp, J. Selb, G. Boverman, Q. Zhang, D. B. Kopans, R. H. Moore, E. L. Miller, D. H. Brooks, and D. A. Boas, “Combined optical imaging and mammography of the healthy breast: optical contrast derived from breast structure and compression,” IEEE Trans. Med. Imaging 28(1), 30–42 (2009). [CrossRef] [PubMed]
21. Q. Fang, J. Selb, S. A. Carp, G. Boverman, E. L. Miller, D. H. Brooks, R. H. Moore, D. B. Kopans, and D. A. Boas, “Combined optical and X-ray tomosynthesis breast imaging,” Radiology 258(1), 89–97 (2011). [CrossRef] [PubMed]
22. D. Boas, T. Gaudette, and S. Arridge, “Simultaneous imaging and optode calibration with diffuse optical tomography,” Opt. Express 8(5), 263–270 (2001). [CrossRef] [PubMed]
23. A. Li, Q. Zhang, J. P. Culver, E. L. Miller, and D. A. Boas, “Reconstructing chromosphere concentration images directly by continuous-wave diffuse optical tomography,” Opt. Lett. 29(3), 256–258 (2004). [CrossRef] [PubMed]
24. S. Srinivasan, B. W. Pogue, S. Jiang, H. Dehghani, and K. D. Paulsen, “Spectrally constrained chromophore and scattering near-infrared tomography provides quantitative and robust reconstruction,” Appl. Opt. 44(10), 1858–1869 (2005). [CrossRef] [PubMed]
25. H. Q. Woodard and D. R. White, “The composition of body tissues,” Br. J. Radiol. 59(708), 1209–1218 (1986). [CrossRef] [PubMed]
26. K. D. Paulsen, P. M. Meaney, M. J. Moskowitz, and J. R. Sullivan, “A dual mesh scheme for finite element based reconstruction algorithms,” IEEE Trans. Med. Imaging 14(3), 504–514 (1995). [CrossRef] [PubMed]
27. A. L. Darling, P. K. Yalavarthy, M. M. Doyley, H. Dehghani, and B. W. Pogue, “Interstitial fluid pressure in soft tissue as a result of an externally applied contact pressure,” Phys. Med. Biol. 52(14), 4121–4136 (2007). [CrossRef] [PubMed]
28. A. A. Gilad, T. Israely, H. Dafni, G. Meir, B. Cohen, and M. Neeman, “Functional and molecular mapping of uncoupling between vascular permeability and loss of vascular maturation in ovarian carcinoma xenografts: the role of stroma cells in tumor angiogenesis,” Int. J. Cancer 117(2), 202–211 (2005). [CrossRef] [PubMed]
29. S. Dische, M. I. Saunders, R. Sealy, I. D. Werner, N. Verma, C. Foy, and S. M. Bentzen, “Carcinoma of the cervix and the use of hyperbaric oxygen with radiotherapy: a report of a randomised controlled trial,” Radiother. Oncol. 53(2), 93–98 (1999). [CrossRef] [PubMed]
30. D. R. White, H. Q. Woodard, and S. M. Hammond, “Average soft-tissue and bone models for use in radiation dosimetry,” Br. J. Radiol. 60(717), 907–913 (1987). [CrossRef] [PubMed]
