| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Developments |
1 From the Thayer School of Engineering, Dartmouth College, 8000 Cummings Hall, Hanover, NH 03755-8000 (B.W.P., T.O.M., K.S.O., U.L.O., K.D.P.); and the Departments of Radiology (S.P.P.) and Pathology (W.A.W.), Dartmouth-Hitchcock Medical Center, Lebanon, NH. Received December 28, 1999; revision requested February 4, 2000; revision received April 1; accepted May 26. Supported by the National Institutes of Health through National Cancer Institute grants RO1CA69544 and P01CA80139. Address correspondence to K.D.P. (e-mail: keith.paulsen@dartmouth.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
Index terms: Breast neoplasms, 00.311, 00.324 • Breast neoplasms, radiography, 00.111, 00.112, 00.113 • Breast radiography, 00.111, 00.112, 00.113 • Breast radiography, technology, 00.111, 00.112, 00.113 • Infrared and near-infrared spectroscopy • Tomography, 00.19
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
Since Jobsis (5) observed that changes in hemoglobin and myoglobin concentrations could be monitored with near-infrared (NIR) spectroscopy, there has been considerable interest in developing methods to noninvasively assess hemoglobin concentration in tissues by using this portion of the electromagnetic spectrum. Transillumination or diaphanography precursors to NIR breast imaging have yielded disappointing clinical results (6–8). Images were formed with direct projection of the NIR intensity onto a pixilated map, which represented a qualitative display of the amount of transmission that propagated through the breast. No attempts were made to exploit the spectral signatures of known breast constituents that absorb in the NIR wavelength region—for example, hemoglobin, water, and lipids (9)—or to account for the substantial amount of scattering that occurs in tissue.
Investigators have sought to overcome these limitations by either using spatially averaged spectroscopy with quantitative sampling of individual points directly on the malignancy (10) or neglecting the nonlinearity from multiple scattering and proceeding with spatially resolved but qualitative image reconstruction of the breast. NIR imaging studies based on intensity recorded at multiple detector positions also have been described: In these investigations, photon propagation was modeled as a diffusive process to form qualitative interaction coefficient images (11). Although these advances represent major steps forward in NIR breast characterization and imaging, quantitative spectroscopy has not been merged with spatially resolved imaging to produce quantitative spatial maps of functional parameters such as hemoglobin concentration, in part because of the hardware and software complexities required to incorporate both tissue spectroscopy and imaging within a single system.
To achieve quantitative spatially resolved spectroscopy, multispectral data acquisition from multiple source-detector fibers arranged in a tomographic configuration must be coupled with an image reconstruction methodology that accounts for both tissue scattering and the nonlinear relationship between the optical properties intrinsic to tissue and the measured detector responses. The purpose of our study was to report what is, to the best of our knowledge, the first quantitative hemoglobin concentration images obtained in the female breast that were formed with nonlinear model-based reconstruction of NIR tomographic data.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
|
During image reconstruction, absolute intensity-modulated measurements are matched to calculations from a finite element solution of the diffusion equation. A calibration of the system is completed daily by using homogeneous tissue phantoms with known optical properties. Small variations in the data that are assumed to be caused by systematic errors that result from individual fiber characteristics occur daily and are subtracted from patient data during the reconstruction process. An initial estimate of the tissue optical properties is required for iterative image reconstruction, and this estimate is derived for each patient data set by fitting the measurements to a homogeneous finite element diffusion equation calculation on a circular geometry scaled to the relevant breast size.
The reconstructed absorption coefficient images obtained at a series of wavelengths are then converted to a hemoglobin concentration image by using the known spectra for hemoglobin and water. The dominant absorbing molecules of breast tissue at NIR wavelengths are hemoglobin, water, and lipids (9), so the quantification of these can be obtained from absorption coefficient images at wavelengths that allow separation of the molecular spectral features (12,14). Reconstructed hemoglobin concentration images are displayed as a circular section in the coronal plane, with the gray scale corresponding to the hemoglobin concentration, in micromolars.
Calibration and Tissue Phantom Testing
Before in vivo imaging, the reconstruction algorithm is calibrated on tissue-simulating phantoms to ensure quantitatively accurate results. We use a well-characterized tissue-simulating phantom composed of a lipid emulsion solution (Intralipid; Kabi Pharmacia, Clayton, NC) mixed with various amounts of whole blood to mimic the scattering of breast tissue (12,15). The added blood samples are measured for their hemoglobin concentration on a clinical co-oximeter (Rapidlab 800; Chiron Diagnostics, Norwood, Mass) and have typically yielded a standard value of 156 g/L of hemoglobin in the blood. Assuming a hemoglobin molecular weight of 64,500, mixtures of 0.5% of the lipid emulsion with 0.5% whole blood produce tissue-simulating solutions with a hemoglobin concentration of 12 µM (0.80 g/L). This ratio also approximates the optical transport scattering coefficient, µs', and absorption coefficient, µa, of tissue well, with µs' values of 0.5 mm-1 and µa values of 0.005 mm-1 at an 800-nm wavelength. Within an 86-mm-diameter volume of this solution, we add a volume of the same material with serial increases in the concentration of blood—0%, 1.0%, 1.5%, and 2.0%—to yield hemoglobin absorption contrasts of 1:1, 2:1, 3:1, and 4:1, respectively, between the added tissue-simulating solution (hereafter referred to as "inclusion") and the background tissue. Figure 2 shows typical hemoglobin concentration images on which the peak value of the inclusion is quantitatively accurate—to within 10% of the true value.
|
Two additional participants, who had confirmed pathologic findings, were recruited directly from our Breast Imaging Center from a group of women who had suspicious mammograms that necessitated tissue sampling and who enrolled in the study sequentially after consenting to undergo the examination. One of these two subjects had a clinical lump that corresponded to her imaging abnormality. The fiber-optic array was therefore placed at the site of the clinical and corresponding imaging abnormality. Fine-needle aspiration of the lump had been performed in this symptomatic patient 1 day before NIR imaging. In a second subject with an asymptomatic imaging abnormality, US was performed immediately before the NIR examination to localize the section of interest for hemoglobin imaging. Stereotactic core-needle biopsy was performed in this subject 16 days before NIR imaging. During NIR imaging, the subjects were placed in a prone position on a padded examination platform, with the breast to be imaged hanging pendantly within the fiber-optic array.
In the two patients with imaging abnormalities, the NIR imaging plane was specifically chosen to include the abnormal region, as described above. In contrast, the subjects without abnormalities underwent imaging in a randomly chosen plane of one breast. The contralateral breast of all subjects was imaged in an equivalent plane that was based on the distance between the chest wall and the nipple. The location of the conventional imaging abnormality was correlated with the NIR image geometry by a radiologist who specializes in mammography (S.P.P.). The coronal location of the abnormalities (described by quadrant, o’clock position, and distance from the nipple) was estimated from the lesion location on standard craniocaudal, mediolateral oblique, and 90° mediolateral or lateromedial mammographic projections and on radial and antiradial US projections.
Pathologic Studies
Subsequent excisional biopsy was performed in the two patients with imaging abnormalities 17 and 34 days after hemoglobin imaging. Immunostaining of the resected breast tumors was used to compare the magnitude of hemoglobin concentration calculated on the NIR images with the blood vessel density and the epithelium-to-stroma ratio within the sampled lesion. The endothelial cell marker CD34 was used to outline the vascular channels within the breast tissue. The epithelial cell marker cytokeratin 5D3 labeled the epithelial components of the breast parenchyma—that is, normal, fibrocystic, or tumorous—in contrast to those of the surrounding stroma—that is, fat, connective tissue, or muscle. Tumor biopsy specimens were fixed in 10% buffered formalin (Biochemical Science, Swedesboro, NJ), dehydrated with graded alcohols, and paraffin embedded per our routine laboratory protocol. The specimens were cut into 4-µm-thick sections, deparaffinized with xylene, and rehydrated with graded alcohols. The sections were immunostained, per a protocol adapted for the BioTek 1000 automated immunohistochemistry system (Ventana, Tucson, Ariz), by using the avidin-biotin-complex technique for CD34 (prediluted) (Immunotech, Westbrook, Me) and 5D3 (1:40) (Biogenix, San Ramon, Calif). By using computer-assisted imaging, multiple microscopic fields, representative of the lesion’s periphery and center, were digitized at a fixed magnification, and the blood vessel density and epithelium-to-stroma ratio were evaluated.
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
Case 1: Fibroadenoma
A 40-year-old woman presented with a 3-cm sharply marginated lobular mass in the upper central portion of the right breast at mammography (Fig 3a) and US, which in retrospect was palpable at clinical examination. An excisional biopsy subsequent to NIR imaging demonstrated a fibroadenoma. Immunostaining and analysis of this excised specimen revealed an overall blood vessel density per unit area of 2.27%; the measurement at the center of the mass was slightly lower, 1.92%, relative to that at the periphery, 2.62%. The epithelium-to-stroma ratio was 0.27, reflecting the stromal overgrowth that characterizes a fibroadenoma. The corresponding NIR hemoglobin concentration image of the same breast is shown in Figure 3b. The peak hemoglobin concentration at the tumor site was 55 µM (3.66 g/L), with a mean background concentration (± SD) of 33 µM ± 5 (2.2 g/L ± 0.3) in the adjacent normal tissue. Figure 3c shows the unremarkable NIR image reconstructed from data acquired from an analogous location in the contralateral breast, which yielded a mean value of 35 µM (2.3 g/L).
|
|
|
|
|
|
|
|
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
Important questions of spatial resolution, hemoglobin sensitivity, and tumor specificity remain to be answered. In terms of spatial resolution, the full width at half maximum of the localized hemoglobin increases on the NIR images obtained in the two breasts with abnormalities measured 1.5 cm ± 0.1 and 0.8 cm ± 0.1 compared with tumor sizes of 3.4 cm and 0.8 cm, respectively, reported at pathologic analysis. This discrepancy between pathologic size and NIR image size is not understood, is the focus of ongoing studies, and may be because the regional blood perfusion of the tumor does not directly correlate with the tumor size measured at pathologic analysis.
In terms of fundamental NIR resolution, the results of recent phantom experiments (17) suggest that the minimum detectable size of a subsurface change in hemoglobin concentration is approximately 2 mm, but quantitatively accurate recovery necessitates that the inclusion size exceed 4 mm. In extrapolating these findings to the tumor cases presented here, one assumes that both lesions were of sufficient size to be accurately characterized with respect to hemoglobin concentration. Although the spatial resolution of NIR imaging does not approach that of conventional imaging modalities, including mammography, US, and MR imaging, it may provide functional information that is not obtainable with other noninvasive imaging systems.
In response to possible concerns that low signal penetration limits the signal-to-noise ratio in larger breasts, it is important to note that the breast diameters imaged in our volunteer group were larger than 60 mm but less than 100 mm. In theoretic arguments, it has been predicted that the detection sensitivity should be sufficient to allow amplitude-modulated NIR wavelengths to be transmitted through 100 mm of breast tissue. Our study results confirmed these estimates in vivo and established that the inherent tissue size constraints in NIR hemoglobin tomography should not be a major limitation in clinical practice. In the two case studies highlighted, the imaged breasts had diameters of 84 and 64 mm. The integrity of the NIR measurements recorded across these sizable distances of breast tissue is encouraging: The data were obtained within the practical constraints enforced by the in vivo clinical (ie, nonlaboratory) environment in a noncompressive geometry by using readily available optoelectronic technology. The lack of firm compression also represents a potential advantage in terms of examination acceptance, and there is sufficient tissue-fiber contact to prevent motion from substantially degrading image clarity.
These in vivo results also show that localized hemoglobin contrasts of 2:1 are readily differentiated. Variations among subjects with negative mammograms were typically plus or minus 5 µM (0.33 g/L), which suggests that tumor increases of 10 µM (0.67 g/L) or more relative to background fluctuations might be required to establish a detection threshold. The phantom study results reported here show that accurate hemoglobin concentrations can be obtained to within 10%–15% of the expected values over similar levels of contrast. In fact, localized inclusions with contrasts of 2:1, 3:1, and 4:1 relative to a 12-µM (0.80-g/L) hemoglobin concentration in the background were reconstructed with peak values of 2.1:1.0, 2.6:1.0, and 3.5:1.0, respectively, with fluctuations in the background limited to plus or minus 2 µM (0.13 g/L). These concentrations are reasonably representative of the values recovered in this pilot series of in vivo examinations and suggest that a hemoglobin sensitivity of 2–5 µM (0.13–0.33 g/L) may be achievable in vivo.
At this early stage of clinical investigation, it is impossible to assess the sensitivity and specificity of NIR imaging. It is interesting to note that the breast with carcinoma had a higher hemoglobin concentration—by approximately 10 µM (0.67 g/L)—than did the breast with the fibroadenoma. Because tissue sampling (ie, fine-needle aspiration biopsy) occurred before NIR imaging in both cases, we could not differentiate the hemoglobin concentration changes related to the intrinsic properties of the tumors from the changes due to a host response to biopsy. Although the procedural order of examination events confounds any statements about the potential for specificity and should be changed in future studies (ie, NIR imaging before biopsy), it does not detract from the fact that localized hemoglobin concentrations were clearly detected and characterized at the site of the pathologic entity with the NIR technique, and the NIR-derived values agreed relatively with the pathologically determined vessel densities. More extensive studies with larger numbers of accruals in which documented abnormalities exist and prebiopsy NIR imaging is performed are needed to provide more definitive data on tumor differentiation, hemoglobin concentration sensitivity, and spatial resolution. Nonetheless, the results presented herein are exciting because they demonstrate a substantial transition from laboratory phantom studies to in vivo clinical examinations with an NIR technique that can provide quantitative spatially resolved hemoglobin concentration images of the female breast.
It is attractive to conceptually compare the type of information obtained with this fundamentally new imaging modality with that obtainable with color Doppler US or contrast-enhanced MR imaging. With color Doppler US, blood flow is imaged, and with MR imaging, a composite of blood flow and tissue-vessel permeability are indirectly probed. In comparison, the NIR imaging system measures the total hemoglobin concentration, which is effectively equal to the total blood volume fraction localized in the tissue, assuming that the hematocrit level is known. Furthermore, NIR imaging is most sensitive to microvessel density rather than larger blood vessel density; this suggests that a different region of the tumor is sampled relative to the region sampled with contrast-based US or MR imaging. Although there is likely to be correlation between blood flow and blood volume fraction, these vascular parameters may exhibit characteristics associated with specific tumors that are independent of each other. Thus, it is possible that hemoglobin-based imaging with NIR tomography can provide fundamentally new information on breast disease without the need for intravenous contrast material.
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, B.W.P., S.P.P., K.D.P.; study concepts, B.W.P., S.P.P., K.D.P.; study design, B.W.P., K.D.P., U.L.O.; definition of intellectual content, B.W.P.; literature research, B.W.P.; clinical studies, S.P.P.; experimental studies, K.S.O., T.O.M., W.A.W., B.W.P.; data acquisition, K.S.O., T.O.M., W.A.W.; data analysis, B.W.P., S.P.P., K.D.P.; manuscript preparation, B.W.P.; manuscript editing, review, and final approval, B.W.P., K.D.P., S.P.P.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  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 PNAS, March 6, 2007; 104(10): 4014 - 4019. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Q. Zhu, E. B. Cronin, A. A. Currier, H. S. Vine, M. Huang, N. Chen, and C. Xu Benign versus Malignant Breast Masses: Optical Differentiation with US-guided Optical Imaging Reconstruction Radiology, October 1, 2005; 237(1): 57 - 66. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
V. Ntziachristos, E. A. Schellenberger, J. Ripoll, D. Yessayan, E. Graves, A. Bogdanov Jr., L. Josephson, and R. Weissleder Visualization of antitumor treatment by means of fluorescence molecular tomography with an annexin V-Cy5.5 conjugate PNAS, August 17, 2004; 101(33): 12294 - 12299. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. J. Eppstein, D. J. Hawrysz, A. Godavarty, and E. M. Sevick-Muraca Three-dimensional, Bayesian image reconstruction from sparse and noisy data sets: Near-infrared fluorescence tomography PNAS, July 23, 2002; 99(15): 9619 - 9624. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. A. Franceschini, D. A. Boas, A. Zourabian, S. G. Diamond, S. Nadgir, D. W. Lin, J. B. Moore, and S. Fantini Near-infrared spiroximetry: noninvasive measurements of venous saturation in piglets and human subjects J Appl Physiol, January 1, 2002; 92(1): 372 - 384. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Bremer, C.-H. Tung, A. Bogdanov Jr, and R. Weissleder Imaging of Differential Protease Expression in Breast Cancers for Detection of Aggressive Tumor Phenotypes Radiology, March 1, 2002; 222(3): 814 - 818. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
) with SD error bars are shown. The peak hemoglobin concentrations with the two observed abnormalities, fibroadenoma (
) and carcinoma (
), also are shown.